├── .gitee ├── ISSUE_TEMPLATE.zh-CN.md └── PULL_REQUEST_TEMPLATE.zh-CN.md ├── .gitignore ├── CONTRIBUTING.md ├── LICENSE ├── README.md ├── README_en.md ├── RELEASE.md ├── data └── datasets │ └── README.md ├── docs └── README_TEMPLATE.md ├── models ├── audio │ └── speech_recognition │ │ ├── conformer │ │ ├── igie │ │ │ ├── README.md │ │ │ ├── build_engine.py │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ ├── common.py │ │ │ ├── ixrt_inference_accuracy.py │ │ │ ├── ixrt_inference_performance.py │ │ │ ├── load_ixrt_plugin.py │ │ │ ├── requirements.txt │ │ │ ├── scripts │ │ │ │ ├── aishell_data_prepare.sh │ │ │ │ ├── infer_conformer_fp16_accuracy.sh │ │ │ │ └── infer_conformer_fp16_performance.sh │ │ │ ├── tools │ │ │ │ ├── __init__.py │ │ │ │ ├── compute_cer.py │ │ │ │ ├── filter_scp.pl │ │ │ │ ├── make_raw_list.py │ │ │ │ ├── make_shard_list.py │ │ │ │ └── text2token.py │ │ │ └── wenet │ │ │ │ ├── dataset.py │ │ │ │ ├── file_utils.py │ │ │ │ └── processor.py │ │ └── ixrt │ │ │ ├── README.md │ │ │ ├── build_engine.py │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ ├── common.py │ │ │ ├── ixrt_inference_accuracy.py │ │ │ ├── ixrt_inference_performance.py │ │ │ ├── load_ixrt_plugin.py │ │ │ ├── requirements.txt │ │ │ ├── scripts │ │ │ ├── aishell_data_prepare.sh │ │ │ ├── infer_conformer_fp16_accuracy.sh │ │ │ └── infer_conformer_fp16_performance.sh │ │ │ ├── tools │ │ │ ├── __init__.py │ │ │ ├── compute_cer.py │ │ │ ├── filter_scp.pl │ │ │ ├── make_raw_list.py │ │ │ ├── make_shard_list.py │ │ │ └── text2token.py │ │ │ └── wenet │ │ │ ├── dataset.py │ │ │ ├── file_utils.py │ │ │ └── processor.py │ │ └── transformer_asr │ │ └── ixrt │ │ ├── README.md │ │ ├── aishell_prepare.py │ │ ├── beam_search.py │ │ ├── build.sh │ │ ├── builder.py │ │ ├── ci │ │ └── prepare.sh │ │ ├── convert.py │ │ ├── ctc.py │ │ ├── faster_cat │ │ ├── __init__.py │ │ ├── build.sh │ │ ├── kernel.cu │ │ ├── setup.py │ │ ├── test.cpp │ │ └── test.py │ │ ├── faster_layer_norm │ │ ├── __init__.py │ │ ├── build.sh │ │ ├── kernel.cu │ │ ├── setup.py │ │ ├── test.cpp │ │ └── transformer_helper.cuh │ │ ├── faster_logsumexp │ │ ├── __init__.py │ │ ├── build.sh │ │ ├── kernel.cu │ │ ├── setup.py │ │ ├── test.cpp │ │ └── test.py │ │ ├── faster_stack │ │ ├── __init__.py │ │ ├── build.sh │ │ ├── kernel.cu │ │ ├── setup.py │ │ ├── test.cpp │ │ └── test.py │ │ ├── hparams │ │ └── train_ASR_transformer.yaml │ │ ├── inference.py │ │ ├── load_ixrt_plugin.py │ │ └── requirements.txt ├── benchmark │ ├── benchmark_longbench.py │ ├── benchmark_serving.py │ ├── benchmark_serving_multimodal.py │ ├── internvl2 │ │ └── README.md │ ├── llama3.1 │ │ └── README.md │ ├── llava.jinja │ ├── longbench │ │ ├── config │ │ │ ├── dataset2maxlen.json │ │ │ ├── dataset2prompt.json │ │ │ ├── model2maxlen.json │ │ │ └── model2path.json │ │ ├── metrics.py │ │ ├── result_record │ │ │ ├── llama3.1-8b-chat.jsonl │ │ │ ├── qwen2-vl-chat.jsonl │ │ │ └── qwen2.5-14b-chat.jsonl │ │ └── utils.py │ ├── mixtral │ │ └── README.md │ ├── quantize_w8a8.py │ ├── qwen2.5 │ │ └── README.md │ ├── requirements.txt │ ├── set_environment.sh │ ├── test.jpg │ ├── test_performance_server.sh │ ├── test_performance_server_multimodal.sh │ ├── test_vllm_longbench.sh │ └── yi1.5 │ │ └── README.md ├── cv │ ├── classification │ │ ├── alexnet │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_alexnet_fp16_accuracy.sh │ │ │ │ │ ├── infer_alexnet_fp16_performance.sh │ │ │ │ │ ├── infer_alexnet_int8_accuracy.sh │ │ │ │ │ └── infer_alexnet_int8_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_alexnet_fp16_accuracy.sh │ │ │ │ ├── infer_alexnet_fp16_performance.sh │ │ │ │ ├── infer_alexnet_int8_accuracy.sh │ │ │ │ └── infer_alexnet_int8_performance.sh │ │ ├── clip │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── build_engine.py │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── export.py │ │ │ │ ├── imagenet_labels.yaml │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ │ ├── infer_clip_fp16_accuracy.sh │ │ │ │ │ └── infer_clip_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── export.py │ │ │ │ ├── imagenet_labels.yaml │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_clip_fp16_accuracy.sh │ │ │ │ └── infer_clip_fp16_performance.sh │ │ ├── conformer_base │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_conformer_base_fp16_accuracy.sh │ │ │ │ └── infer_conformer_base_fp16_performance.sh │ │ ├── convnext_base │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_convnext_base_fp16_accuracy.sh │ │ │ │ │ └── infer_convnext_base_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_convnext_base_fp16_accuracy.sh │ │ │ │ └── infer_convnext_base_fp16_performance.sh │ │ ├── convnext_s │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_convnext_s_fp16_accuracy.sh │ │ │ │ └── infer_convnext_s_fp16_performance.sh │ │ ├── convnext_small │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_convnext_small_fp16_accuracy.sh │ │ │ │ │ └── infer_convnext_small_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_convnext_small_fp16_accuracy.sh │ │ │ │ └── infer_convnext_small_fp16_performance.sh │ │ ├── convnext_tiny │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_convnext_tiny_fp16_accuracy.sh │ │ │ │ └── infer_convnext_tiny_fp16_performance.sh │ │ ├── cspdarknet53 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_cspdarknet53_fp16_accuracy.sh │ │ │ │ │ └── infer_cspdarknet53_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_cspdarknet53_fp16_accuracy.sh │ │ │ │ ├── infer_cspdarknet53_fp16_performance.sh │ │ │ │ ├── infer_cspdarknet53_int8_accuracy.sh │ │ │ │ └── infer_cspdarknet53_int8_performance.sh │ │ ├── cspresnet50 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_cspresnet50_fp16_accuracy.sh │ │ │ │ │ └── infer_cspresnet50_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_cspresnet50_fp16_accuracy.sh │ │ │ │ ├── infer_cspresnet50_fp16_performance.sh │ │ │ │ ├── infer_cspresnet50_int8_accuracy.sh │ │ │ │ └── infer_cspresnet50_int8_performance.sh │ │ ├── cspresnext50 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_cspresnext50_fp16_accuracy.sh │ │ │ │ │ └── infer_cspresnext50_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_cspresnext50_fp16_accuracy.sh │ │ │ │ └── infer_cspresnext50_fp16_performance.sh │ │ ├── deit_tiny │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_deit_tiny_fp16_accuracy.sh │ │ │ │ │ └── infer_deit_tiny_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_deit_tiny_fp16_accuracy.sh │ │ │ │ └── infer_deit_tiny_fp16_performance.sh │ │ ├── densenet121 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_densenet121_fp16_accuracy.sh │ │ │ │ │ └── infer_densenet121_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── export.py │ │ │ │ └── scripts │ │ │ │ ├── infer_densenet121_fp16_accuracy.sh │ │ │ │ └── infer_densenet121_fp16_performance.sh │ │ ├── densenet161 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_densenet161_fp16_accuracy.sh │ │ │ │ │ └── infer_densenet161_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_densenet161_fp16_accuracy.sh │ │ │ │ └── infer_densenet161_fp16_performance.sh │ │ ├── densenet169 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_densenet169_fp16_accuracy.sh │ │ │ │ │ └── infer_densenet169_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_densenet169_fp16_accuracy.sh │ │ │ │ └── infer_densenet169_fp16_performance.sh │ │ ├── densenet201 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_densenet201_fp16_accuracy.sh │ │ │ │ │ └── infer_densenet201_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_densenet201_fp16_accuracy.sh │ │ │ │ └── infer_densenet201_fp16_performance.sh │ │ ├── efficientnet_b0 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_efficientnet_b0_fp16_accuracy.sh │ │ │ │ │ └── infer_efficientnet_b0_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_efficientnet_b0_fp16_accuracy.sh │ │ │ │ ├── infer_efficientnet_b0_fp16_performance.sh │ │ │ │ ├── infer_efficientnet_b0_int8_accuracy.sh │ │ │ │ └── infer_efficientnet_b0_int8_performance.sh │ │ ├── efficientnet_b1 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_efficientnet_b1_fp16_accuracy.sh │ │ │ │ │ └── infer_efficientnet_b1_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_efficientnet_b1_fp16_accuracy.sh │ │ │ │ ├── infer_efficientnet_b1_fp16_performance.sh │ │ │ │ ├── infer_efficientnet_b1_int8_accuracy.sh │ │ │ │ └── infer_efficientnet_b1_int8_performance.sh │ │ ├── efficientnet_b2 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_efficientnet_b2_fp16_accuracy.sh │ │ │ │ │ └── infer_efficientnet_b2_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_efficientnet_b2_fp16_accuracy.sh │ │ │ │ └── infer_efficientnet_b2_fp16_performance.sh │ │ ├── efficientnet_b3 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_efficientnet_b3_fp16_accuracy.sh │ │ │ │ │ └── infer_efficientnet_b3_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_efficientnet_b3_fp16_accuracy.sh │ │ │ │ └── infer_efficientnet_b3_fp16_performance.sh │ │ ├── efficientnet_b4 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_efficientnet_b4_fp16_accuracy.sh │ │ │ │ │ └── infer_efficientnet_b4_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_efficientnet_b4_fp16_accuracy.sh │ │ │ │ └── infer_efficientnet_b4_fp16_performance.sh │ │ ├── efficientnet_b5 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_efficientnet_b5_fp16_accuracy.sh │ │ │ │ │ └── infer_efficientnet_b5_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_efficientnet_b5_fp16_accuracy.sh │ │ │ │ └── infer_efficientnet_b5_fp16_performance.sh │ │ ├── efficientnet_b6 │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_efficientnet_b6_fp16_accuracy.sh │ │ │ │ └── infer_efficientnet_b6_fp16_performance.sh │ │ ├── efficientnet_b7 │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_efficientnet_b7_fp16_accuracy.sh │ │ │ │ └── infer_efficientnet_b7_fp16_performance.sh │ │ ├── efficientnet_v2 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_efficientnet_v2_fp16_accuracy.sh │ │ │ │ │ └── infer_efficientnet_v2_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── _builder.py │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── export_onnx.py │ │ │ │ └── scripts │ │ │ │ ├── infer_efficientnet_v2_fp16_accuracy.sh │ │ │ │ ├── infer_efficientnet_v2_fp16_performance.sh │ │ │ │ ├── infer_efficientnet_v2_int8_accuracy.sh │ │ │ │ └── infer_efficientnet_v2_int8_performance.sh │ │ ├── efficientnet_v2_s │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_efficientnet_v2_s_fp16_accuracy.sh │ │ │ │ │ └── infer_efficientnet_v2_s_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_efficientnet_v2_s_fp16_accuracy.sh │ │ │ │ └── infer_efficientnet_v2_s_fp16_performance.sh │ │ ├── efficientnetv2_rw_t │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── inference.py │ │ │ │ └── scripts │ │ │ │ │ ├── infer_efficientnetv2_rw_t_fp16_accuracy.sh │ │ │ │ │ └── infer_efficientnetv2_rw_t_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_efficientnetv2_rw_t_fp16_accuracy.sh │ │ │ │ └── infer_efficientnetv2_rw_t_fp16_performance.sh │ │ ├── googlenet │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_googlenet_fp16_accuracy.sh │ │ │ │ │ ├── infer_googlenet_fp16_performance.sh │ │ │ │ │ ├── infer_googlenet_int8_accuracy.sh │ │ │ │ │ └── infer_googlenet_int8_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_googlenet_fp16_accuracy.sh │ │ │ │ ├── infer_googlenet_fp16_performance.sh │ │ │ │ ├── infer_googlenet_int8_accuracy.sh │ │ │ │ └── infer_googlenet_int8_performance.sh │ │ ├── hrnet_w18 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_hrnet_w18_fp16_accuracy.sh │ │ │ │ │ └── infer_hrnet_w18_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_hrnet_w18_fp16_accuracy.sh │ │ │ │ ├── infer_hrnet_w18_fp16_performance.sh │ │ │ │ ├── infer_hrnet_w18_int8_accuracy.sh │ │ │ │ └── infer_hrnet_w18_int8_performance.sh │ │ ├── igie_common │ │ │ ├── build_engine.py │ │ │ ├── export.py │ │ │ ├── export_mmcls.py │ │ │ ├── export_timm.py │ │ │ ├── inference.py │ │ │ ├── quantize.py │ │ │ └── requirements.txt │ │ ├── inception_resnet_v2 │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── export_model.py │ │ │ │ └── scripts │ │ │ │ ├── infer_inception_resnet_v2_fp16_accuracy.sh │ │ │ │ ├── infer_inception_resnet_v2_fp16_performance.sh │ │ │ │ ├── infer_inception_resnet_v2_int8_accuracy.sh │ │ │ │ └── infer_inception_resnet_v2_int8_performance.sh │ │ ├── inception_v3 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_inception_v3_fp16_accuracy.sh │ │ │ │ │ ├── infer_inception_v3_fp16_performance.sh │ │ │ │ │ ├── infer_inception_v3_int8_accuracy.sh │ │ │ │ │ └── infer_inception_v3_int8_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_inception_v3_fp16_accuracy.sh │ │ │ │ ├── infer_inception_v3_fp16_performance.sh │ │ │ │ ├── infer_inception_v3_int8_accuracy.sh │ │ │ │ └── infer_inception_v3_int8_performance.sh │ │ ├── ixrt_common │ │ │ ├── README.md │ │ │ ├── build_engine.py │ │ │ ├── build_i8_engine.py │ │ │ ├── calibration_dataset.py │ │ │ ├── common.py │ │ │ ├── config │ │ │ │ ├── ALEXNET_CONFIG │ │ │ │ ├── CLIP_CONFIG │ │ │ │ ├── CONVNEXT_BASE_CONFIG │ │ │ │ ├── CONVNEXT_SMALL_CONFIG │ │ │ │ ├── CSPDARKNET53_CONFIG │ │ │ │ ├── CSPRESNET50_CONFIG │ │ │ │ ├── CSPRESNEXT50_CONFIG │ │ │ │ ├── DEIT_TINY_CONFIG │ │ │ │ ├── DENSENET121_CONFIG │ │ │ │ ├── DENSENET161_CONFIG │ │ │ │ ├── DENSENET169_CONFIG │ │ │ │ ├── DENSENET201_CONFIG │ │ │ │ ├── EFFICIENTNETV2_RW_T_CONFIG │ │ │ │ ├── EFFICIENTNET_B0_CONFIG │ │ │ │ ├── EFFICIENTNET_B1_CONFIG │ │ │ │ ├── EFFICIENTNET_B2_CONFIG │ │ │ │ ├── EFFICIENTNET_B3_CONFIG │ │ │ │ ├── EFFICIENTNET_B4_CONFIG │ │ │ │ ├── EFFICIENTNET_B5_CONFIG │ │ │ │ ├── EFFICIENTNET_V2_CONFIG │ │ │ │ ├── EFFICIENTNET_V2_S_CONFIG │ │ │ │ ├── GOOGLENET_CONFIG │ │ │ │ ├── HRNET_W18_CONFIG │ │ │ │ ├── INCEPTION_RESNET_V2_CONFIG │ │ │ │ ├── INCEPTION_V3_CONFIG │ │ │ │ ├── MOBILENET_V2_CONFIG │ │ │ │ ├── MOBILENET_V3_CONFIG │ │ │ │ ├── REPVGG_CONFIG │ │ │ │ ├── REPVIT_CONFIG │ │ │ │ ├── RES2NET50_CONFIG │ │ │ │ ├── RESNET101_CONFIG │ │ │ │ ├── RESNET18_CONFIG │ │ │ │ ├── RESNET34_CONFIG │ │ │ │ ├── RESNET50_CONFIG │ │ │ │ ├── RESNETV1D50_CONFIG │ │ │ │ ├── RESNEXT101_32X8D_CONFIG │ │ │ │ ├── RESNEXT101_64X4D_CONFIG │ │ │ │ ├── RESNEXT50_32X4D_CONFIG │ │ │ │ ├── SHUFFLENETV2_X0_5_CONFIG │ │ │ │ ├── SHUFFLENETV2_X1_0_CONFIG │ │ │ │ ├── SHUFFLENETV2_X1_5_CONFIG │ │ │ │ ├── SHUFFLENETV2_X2_0_CONFIG │ │ │ │ ├── SHUFFLENET_V1_CONFIG │ │ │ │ ├── SQUEEZENET_V1_0_CONFIG │ │ │ │ ├── SQUEEZENET_V1_1_CONFIG │ │ │ │ ├── VGG16_CONFIG │ │ │ │ └── WIDE_RESNET50_CONFIG │ │ │ ├── export.py │ │ │ ├── export_mmcls.py │ │ │ ├── export_timm.py │ │ │ ├── inference.py │ │ │ ├── load_ixrt_plugin.py │ │ │ ├── modify_batchsize.py │ │ │ ├── quant.py │ │ │ ├── quant_i8.py │ │ │ ├── refine_model.py │ │ │ ├── refine_utils │ │ │ │ ├── __init__.py │ │ │ │ ├── common.py │ │ │ │ ├── linear_pass.py │ │ │ │ └── matmul_to_gemm_pass.py │ │ │ ├── requirements.txt │ │ │ └── simplify_model.py │ │ ├── mlp_mixer_base │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_mlp_mixer_base_fp16_accuracy.sh │ │ │ │ └── infer_mlp_mixer_base_fp16_performance.sh │ │ ├── mnasnet0_5 │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_mnasnet0_5_fp16_accuracy.sh │ │ │ │ └── infer_mnasnet0_5_fp16_performance.sh │ │ ├── mnasnet0_75 │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_mnasnet0_75_fp16_accuracy.sh │ │ │ │ └── infer_mnasnet0_75_fp16_performance.sh │ │ ├── mnasnet1_0 │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_mnasnet1_0_fp16_accuracy.sh │ │ │ │ └── infer_mnasnet1_0_fp16_performance.sh │ │ ├── mnasnet1_3 │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_mnasnet1_3_fp16_accuracy.sh │ │ │ │ └── infer_mnasnet1_3_fp16_performance.sh │ │ ├── mobilenet_v2 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_mobilenet_v2_fp16_accuracy.sh │ │ │ │ │ ├── infer_mobilenet_v2_fp16_performance.sh │ │ │ │ │ ├── infer_mobilenet_v2_int8_accuracy.sh │ │ │ │ │ └── infer_mobilenet_v2_int8_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_mobilenet_v2_fp16_accuracy.sh │ │ │ │ ├── infer_mobilenet_v2_fp16_performance.sh │ │ │ │ ├── infer_mobilenet_v2_int8_accuracy.sh │ │ │ │ └── infer_mobilenet_v2_int8_performance.sh │ │ ├── mobilenet_v3 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_mobilenet_v3_fp16_accuracy.sh │ │ │ │ │ └── infer_mobilenet_v3_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_mobilenet_v3_fp16_accuracy.sh │ │ │ │ └── infer_mobilenet_v3_fp16_performance.sh │ │ ├── mobilenet_v3_large │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_mobilenet_v3_large_fp16_accuracy.sh │ │ │ │ └── infer_mobilenet_v3_large_fp16_performance.sh │ │ ├── mvitv2_base │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_mvitv2_base_fp16_accuracy.sh │ │ │ │ └── infer_mvitv2_base_fp16_performance.sh │ │ ├── regnet_x_16gf │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_regnet_x_16gf_fp16_accuracy.sh │ │ │ │ └── infer_regnet_x_16gf_fp16_performance.sh │ │ ├── regnet_x_1_6gf │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_regnet_x_1_6gf_fp16_accuracy.sh │ │ │ │ └── infer_regnet_x_1_6gf_fp16_performance.sh │ │ ├── regnet_x_32gf │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_regnet_x_32gf_fp16_accuracy.sh │ │ │ │ └── infer_regnet_x_32gf_fp16_performance.sh │ │ ├── regnet_x_3_2gf │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_regnet_x_3_2gf_fp16_accuracy.sh │ │ │ │ └── infer_regnet_x_3_2gf_fp16_performance.sh │ │ ├── regnet_x_400mf │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_regnet_x_400mf_fp16_accuracy.sh │ │ │ │ └── infer_regnet_x_400mf_fp16_performance.sh │ │ ├── regnet_x_800mf │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_regnet_x_800mf_fp16_accuracy.sh │ │ │ │ └── infer_regnet_x_800mf_fp16_performance.sh │ │ ├── regnet_x_8gf │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_regnet_x_8gf_fp16_accuracy.sh │ │ │ │ └── infer_regnet_x_8gf_fp16_performance.sh │ │ ├── regnet_y_16gf │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_regnet_y_16gf_fp16_accuracy.sh │ │ │ │ └── infer_regnet_y_16gf_fp16_performance.sh │ │ ├── regnet_y_1_6gf │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_regnet_y_1_6gf_fp16_accuracy.sh │ │ │ │ └── infer_regnet_y_1_6gf_fp16_performance.sh │ │ ├── regnet_y_32gf │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_regnet_y_32gf_fp16_accuracy.sh │ │ │ │ └── infer_regnet_y_32gf_fp16_performance.sh │ │ ├── regnet_y_3_2gf │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_regnet_y_3_2gf_fp16_accuracy.sh │ │ │ │ └── infer_regnet_y_3_2gf_fp16_performance.sh │ │ ├── regnet_y_400mf │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_regnet_y_400mf_fp16_accuracy.sh │ │ │ │ └── infer_regnet_y_400mf_fp16_performance.sh │ │ ├── repvgg │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_repvgg_fp16_accuracy.sh │ │ │ │ │ └── infer_repvgg_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_repvgg_fp16_accuracy.sh │ │ │ │ └── infer_repvgg_fp16_performance.sh │ │ ├── res2net50 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_res2net50_fp16_accuracy.sh │ │ │ │ │ └── infer_res2net50_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── export_onnx.py │ │ │ │ └── scripts │ │ │ │ ├── infer_res2net50_fp16_accuracy.sh │ │ │ │ ├── infer_res2net50_fp16_performance.sh │ │ │ │ ├── infer_res2net50_int8_accuracy.sh │ │ │ │ └── infer_res2net50_int8_performance.sh │ │ ├── resnest50 │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_resnest50_fp16_accuracy.sh │ │ │ │ └── infer_resnest50_fp16_performance.sh │ │ ├── resnet101 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_resnet101_fp16_accuracy.sh │ │ │ │ │ ├── infer_resnet101_fp16_performance.sh │ │ │ │ │ ├── infer_resnet101_int8_accuracy.sh │ │ │ │ │ └── infer_resnet101_int8_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_resnet101_fp16_accuracy.sh │ │ │ │ ├── infer_resnet101_fp16_performance.sh │ │ │ │ ├── infer_resnet101_int8_accuracy.sh │ │ │ │ └── infer_resnet101_int8_performance.sh │ │ ├── resnet152 │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_resnet152_fp16_accuracy.sh │ │ │ │ ├── infer_resnet152_fp16_performance.sh │ │ │ │ ├── infer_resnet152_int8_accuracy.sh │ │ │ │ └── infer_resnet152_int8_performance.sh │ │ ├── resnet18 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_resnet18_fp16_accuracy.sh │ │ │ │ │ ├── infer_resnet18_fp16_performance.sh │ │ │ │ │ ├── infer_resnet18_int8_accuracy.sh │ │ │ │ │ └── infer_resnet18_int8_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_resnet18_fp16_accuracy.sh │ │ │ │ ├── infer_resnet18_fp16_performance.sh │ │ │ │ ├── infer_resnet18_int8_accuracy.sh │ │ │ │ └── infer_resnet18_int8_performance.sh │ │ ├── resnet34 │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_resnet34_fp16_accuracy.sh │ │ │ │ ├── infer_resnet34_fp16_performance.sh │ │ │ │ ├── infer_resnet34_int8_accuracy.sh │ │ │ │ └── infer_resnet34_int8_performance.sh │ │ ├── resnet50 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_resnet50_fp16_accuracy.sh │ │ │ │ │ ├── infer_resnet50_fp16_performance.sh │ │ │ │ │ ├── infer_resnet50_int8_accuracy.sh │ │ │ │ │ └── infer_resnet50_int8_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_resnet50_fp16_accuracy.sh │ │ │ │ ├── infer_resnet50_fp16_performance.sh │ │ │ │ ├── infer_resnet50_int8_accuracy.sh │ │ │ │ └── infer_resnet50_int8_performance.sh │ │ ├── resnetv1d50 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_resnetv1d50_fp16_accuracy.sh │ │ │ │ │ └── infer_resnetv1d50_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_resnetv1d50_fp16_accuracy.sh │ │ │ │ ├── infer_resnetv1d50_fp16_performance.sh │ │ │ │ ├── infer_resnetv1d50_int8_accuracy.sh │ │ │ │ └── infer_resnetv1d50_int8_performance.sh │ │ ├── resnext101_32x8d │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_resnext101_32x8d_fp16_accuracy.sh │ │ │ │ │ └── infer_resnext101_32x8d_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_resnext101_32x8d_fp16_accuracy.sh │ │ │ │ └── infer_resnext101_32x8d_fp16_performance.sh │ │ ├── resnext101_64x4d │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_resnext101_64x4d_fp16_accuracy.sh │ │ │ │ │ └── infer_resnext101_64x4d_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_resnext101_64x4d_fp16_accuracy.sh │ │ │ │ └── infer_resnext101_64x4d_fp16_performance.sh │ │ ├── resnext50_32x4d │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_resnext50_32x4d_fp16_accuracy.sh │ │ │ │ │ └── infer_resnext50_32x4d_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_resnext50_32x4d_fp16_accuracy.sh │ │ │ │ └── infer_resnext50_32x4d_fp16_performance.sh │ │ ├── seresnet50 │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_seresnet50_fp16_accuracy.sh │ │ │ │ └── infer_seresnet50_fp16_performance.sh │ │ ├── shufflenet_v1 │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_shufflenet_v1_fp16_accuracy.sh │ │ │ │ └── infer_shufflenet_v1_fp16_performance.sh │ │ ├── shufflenetv2_x0_5 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_shufflenetv2_x0_5_fp16_accuracy.sh │ │ │ │ │ └── infer_shufflenetv2_x0_5_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_shufflenetv2_x0_5_fp16_accuracy.sh │ │ │ │ └── infer_shufflenetv2_x0_5_fp16_performance.sh │ │ ├── shufflenetv2_x1_0 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_shufflenetv2_x1_0_fp16_accuracy.sh │ │ │ │ │ └── infer_shufflenetv2_x1_0_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_shufflenetv2_x1_0_fp16_accuracy.sh │ │ │ │ └── infer_shufflenetv2_x1_0_fp16_performance.sh │ │ ├── shufflenetv2_x1_5 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_shufflenetv2_x1_5_fp16_accuracy.sh │ │ │ │ │ └── infer_shufflenetv2_x1_5_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_shufflenetv2_x1_5_fp16_accuracy.sh │ │ │ │ └── infer_shufflenetv2_x1_5_fp16_performance.sh │ │ ├── shufflenetv2_x2_0 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_shufflenetv2_x2_0_fp16_accuracy.sh │ │ │ │ │ └── infer_shufflenetv2_x2_0_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_shufflenetv2_x2_0_fp16_accuracy.sh │ │ │ │ └── infer_shufflenetv2_x2_0_fp16_performance.sh │ │ ├── squeezenet_v1_0 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_squeezenet_v1_0_fp16_accuracy.sh │ │ │ │ │ └── infer_squeezenet_v1_0_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_squeezenet_v1_0_fp16_accuracy.sh │ │ │ │ ├── infer_squeezenet_v1_0_fp16_performance.sh │ │ │ │ ├── infer_squeezenet_v1_0_int8_accuracy.sh │ │ │ │ └── infer_squeezenet_v1_0_int8_performance.sh │ │ ├── squeezenet_v1_1 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_squeezenet_v1_1_fp16_accuracy.sh │ │ │ │ │ └── infer_squeezenet_v1_1_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_squeezenet_v1_1_fp16_accuracy.sh │ │ │ │ ├── infer_squeezenet_v1_1_fp16_performance.sh │ │ │ │ ├── infer_squeezenet_v1_1_int8_accuracy.sh │ │ │ │ └── infer_squeezenet_v1_1_int8_performance.sh │ │ ├── svt_base │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_svt_base_fp16_accuracy.sh │ │ │ │ └── infer_svt_base_fp16_performance.sh │ │ ├── swin_transformer │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_swin_transformer_fp16_accuracy.sh │ │ │ │ └── infer_swin_transformer_fp16_performance.sh │ │ ├── swin_transformer_large │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ ├── infer_swinl_fp16_performance.sh │ │ │ │ └── prepare_model_and_dataset.sh │ │ │ │ └── torch2onnx.py │ │ ├── twins_pcpvt │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_twins_pcpvt_fp16_accuracy.sh │ │ │ │ └── infer_twins_pcpvt_fp16_performance.sh │ │ ├── van_b0 │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_van_b0_fp16_accuracy.sh │ │ │ │ └── infer_van_b0_fp16_performance.sh │ │ ├── vgg11 │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_vgg11_fp16_accuracy.sh │ │ │ │ └── infer_vgg11_fp16_performance.sh │ │ ├── vgg13 │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_vgg13_fp16_accuracy.sh │ │ │ │ └── infer_vgg13_fp16_performance.sh │ │ ├── vgg13_bn │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_vgg13_bn_fp16_accuracy.sh │ │ │ │ └── infer_vgg13_bn_fp16_performance.sh │ │ ├── vgg16 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ │ ├── infer_vgg16_fp16_accuracy.sh │ │ │ │ │ ├── infer_vgg16_fp16_performance.sh │ │ │ │ │ ├── infer_vgg16_int8_accuracy.sh │ │ │ │ │ └── infer_vgg16_int8_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_vgg16_fp16_accuracy.sh │ │ │ │ ├── infer_vgg16_fp16_performance.sh │ │ │ │ ├── infer_vgg16_int8_accuracy.sh │ │ │ │ └── infer_vgg16_int8_performance.sh │ │ ├── vgg19 │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_vgg19_fp16_accuracy.sh │ │ │ │ └── infer_vgg19_fp16_performance.sh │ │ ├── vgg19_bn │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_vgg19_bn_fp16_accuracy.sh │ │ │ │ └── infer_vgg19_bn_fp16_performance.sh │ │ ├── vit │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ ├── infer_vit_fp16_accuracy.sh │ │ │ │ └── infer_vit_fp16_performance.sh │ │ │ │ └── test.py │ │ ├── wide_resnet101 │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_wide_resnet101_fp16_accuracy.sh │ │ │ │ └── infer_wide_resnet101_fp16_performance.sh │ │ └── wide_resnet50 │ │ │ ├── igie │ │ │ ├── README.md │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ └── scripts │ │ │ │ ├── infer_wide_resnet50_fp16_accuracy.sh │ │ │ │ ├── infer_wide_resnet50_fp16_performance.sh │ │ │ │ ├── infer_wide_resnet50_int8_accuracy.sh │ │ │ │ └── infer_wide_resnet50_int8_performance.sh │ │ │ └── ixrt │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ └── scripts │ │ │ ├── infer_wide_resnet50_fp16_accuracy.sh │ │ │ ├── infer_wide_resnet50_fp16_performance.sh │ │ │ ├── infer_wide_resnet50_int8_accuracy.sh │ │ │ └── infer_wide_resnet50_int8_performance.sh │ ├── face_recognition │ │ └── facenet │ │ │ └── ixrt │ │ │ ├── README.md │ │ │ ├── build_engine.py │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ ├── common.py │ │ │ ├── config │ │ │ └── FACENET_CONFIG │ │ │ ├── deploy.py │ │ │ ├── inference.py │ │ │ ├── load_ixrt_plugin.py │ │ │ ├── modify_batchsize.py │ │ │ ├── quant.py │ │ │ ├── requirements.txt │ │ │ ├── scripts │ │ │ ├── infer_facenet_fp16_accuracy.sh │ │ │ ├── infer_facenet_fp16_performance.sh │ │ │ ├── infer_facenet_int8_accuracy.sh │ │ │ └── infer_facenet_int8_performance.sh │ │ │ └── utils.py │ ├── instance_segmentation │ │ ├── mask_rcnn │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── plugins │ │ │ │ ├── CMakeLists.txt │ │ │ │ ├── LICENSE │ │ │ │ ├── README.md │ │ │ │ ├── cmake │ │ │ │ │ ├── FindCompiler.cmake │ │ │ │ │ ├── FindCuda.cmake │ │ │ │ │ ├── FindIxrt.cmake │ │ │ │ │ └── SetFlags.cmake │ │ │ │ ├── common.hpp │ │ │ │ ├── cuda_utils.h │ │ │ │ ├── logging.h │ │ │ │ ├── macros.h │ │ │ │ └── plugin_src │ │ │ │ │ ├── BatchedNms.cu │ │ │ │ │ ├── BatchedNmsPlugin.h │ │ │ │ │ ├── MaskRcnnInference.cu │ │ │ │ │ ├── MaskRcnnInferencePlugin.h │ │ │ │ │ ├── PredictorDecode.cu │ │ │ │ │ ├── PredictorDecodePlugin.h │ │ │ │ │ ├── RoiAlign.cu │ │ │ │ │ ├── RoiAlignPlugin.h │ │ │ │ │ ├── RpnDecode.cu │ │ │ │ │ ├── RpnDecodePlugin.h │ │ │ │ │ ├── RpnNms.cu │ │ │ │ │ └── RpnNmsPlugin.h │ │ │ │ ├── python │ │ │ │ ├── backbone.py │ │ │ │ ├── dataloader.py │ │ │ │ ├── head.py │ │ │ │ ├── img_postprocess.py │ │ │ │ ├── img_preprocess.py │ │ │ │ ├── maskrcnn.py │ │ │ │ ├── plugins.py │ │ │ │ └── utils.py │ │ │ │ ├── samples │ │ │ │ └── demo.jpg │ │ │ │ └── scripts │ │ │ │ ├── infer_mask_rcnn_fp16_accuracy.sh │ │ │ │ ├── infer_mask_rcnn_fp16_performance.sh │ │ │ │ ├── init.sh │ │ │ │ ├── init_nv.sh │ │ │ │ └── prepare_system_env.sh │ │ └── solov1 │ │ │ └── ixrt │ │ │ ├── README.md │ │ │ ├── build_engine.py │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ ├── coco_instance.py │ │ │ ├── common.py │ │ │ ├── requirements.txt │ │ │ ├── scripts │ │ │ ├── infer_solov1_fp16_accuracy.sh │ │ │ └── infer_solov1_fp16_performance.sh │ │ │ ├── simplify_model.py │ │ │ ├── solo_r50_fpn_3x_coco.py │ │ │ ├── solo_torch2onnx.py │ │ │ └── solov1_inference.py │ ├── multi_object_tracking │ │ ├── deepsort │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── build_engine.py │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── quantize.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ │ ├── infer_deepsort_fp16_accuracy.sh │ │ │ │ │ ├── infer_deepsort_fp16_performance.sh │ │ │ │ │ ├── infer_deepsort_int8_accuracy.sh │ │ │ │ │ └── infer_deepsort_int8_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── build_engine.py │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── common.py │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ ├── infer_deepsort_fp16_accuracy.sh │ │ │ │ └── infer_deepsort_fp16_performance.sh │ │ │ │ └── utils.py │ │ ├── fastreid │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── build_engine.py │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ │ ├── infer_fastreid_fp16_accuracy.sh │ │ │ │ │ └── infer_fastreid_fp16_performance.sh │ │ │ │ └── utils.py │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── build_engine.py │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── common.py │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ ├── infer_fastreid_fp16_accuracy.sh │ │ │ │ └── infer_fastreid_fp16_performance.sh │ │ │ │ └── utils.py │ │ └── repnet │ │ │ └── igie │ │ │ ├── README.md │ │ │ ├── build_engine.py │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ ├── export.py │ │ │ ├── inference.py │ │ │ ├── pair_set_vehicle.txt │ │ │ ├── requirements.txt │ │ │ └── scripts │ │ │ ├── infer_repnet_fp16_accuracy.sh │ │ │ └── infer_repnet_fp16_performance.sh │ ├── object_detection │ │ ├── atss │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── atss_r50_fpn_1x_coco.py │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ │ ├── infer_atss_fp16_accuracy.sh │ │ │ │ │ └── infer_atss_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_atss_fp16_accuracy.sh │ │ │ │ └── infer_atss_fp16_performance.sh │ │ ├── centernet │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── centernet_r18_8xb16-crop512-140e_coco.py │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ │ ├── infer_centernet_fp16_accuracy.sh │ │ │ │ │ └── infer_centernet_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── base │ │ │ │ ├── centernet_r18-dcnv2_8xb16-crop512-140e_coco.py │ │ │ │ ├── coco_detection.py │ │ │ │ ├── default_runtime.py │ │ │ │ └── schedule_1x.py │ │ │ │ ├── build_engine.py │ │ │ │ ├── centernet_r18_8xb16-crop512-140e_coco.py │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── common.py │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_centernet_fp16_accuracy.sh │ │ │ │ └── infer_centernet_fp16_performance.sh │ │ ├── detr │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── build_engine.py │ │ │ │ ├── calibration_dataset.py │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── coco_labels.py │ │ │ │ ├── common.py │ │ │ │ ├── config │ │ │ │ └── DETR_CONFIG │ │ │ │ ├── datasets │ │ │ │ ├── __init__.py │ │ │ │ ├── coco.py │ │ │ │ ├── common.py │ │ │ │ ├── post_process.py │ │ │ │ ├── pre_process.py │ │ │ │ └── vision.py │ │ │ │ ├── deploy.py │ │ │ │ ├── export_model.py │ │ │ │ ├── extract_graph_weight.py │ │ │ │ ├── inference.py │ │ │ │ ├── load_ixrt_plugin.py │ │ │ │ ├── modify_batchsize.py │ │ │ │ ├── refine_model.py │ │ │ │ ├── refine_utils │ │ │ │ ├── __init__.py │ │ │ │ ├── common.py │ │ │ │ ├── linear_pass.py │ │ │ │ └── matmul_to_gemm_pass.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ ├── infer_detr_fp16_accuracy.sh │ │ │ │ └── infer_detr_fp16_performance.sh │ │ │ │ └── simplify_model.py │ │ ├── fcos │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── fcos_r50_caffe_fpn_gn-head_1x_coco.py │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ │ ├── infer_fcos_fp16_accuracy.sh │ │ │ │ │ └── infer_fcos_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_fcos_fp16_accuracy.sh │ │ │ │ └── infer_fcos_fp16_performance.sh │ │ ├── foveabox │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── fovea_r50_fpn_4xb4-1x_coco.py │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ │ ├── infer_foveabox_fp16_accuracy.sh │ │ │ │ │ └── infer_foveabox_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── build_engine.py │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── common.py │ │ │ │ ├── config │ │ │ │ └── FOVEABOX_CONFIG │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── fovea_r50_fpn_4xb4-1x_coco.py │ │ │ │ ├── inference.py │ │ │ │ ├── modify_batchsize.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ ├── infer_foveabox_fp16_accuracy.sh │ │ │ │ └── infer_foveabox_fp16_performance.sh │ │ │ │ └── simplify_model.py │ │ ├── fsaf │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── fsaf_r50_fpn_1x_coco.py │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ │ ├── infer_fsaf_fp16_accuracy.sh │ │ │ │ │ └── infer_fsaf_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── build_engine.py │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── common.py │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── fsaf_r50_fpn_1x_coco.py │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_fsaf_fp16_accuracy.sh │ │ │ │ └── infer_fsaf_fp16_performance.sh │ │ ├── gfl │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── gfl_r50_fpn_1x_coco.py │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_gfl_fp16_accuracy.sh │ │ │ │ └── infer_gfl_fp16_performance.sh │ │ ├── hrnet │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── fcos_hrnetv2p-w18-gn-head_4xb4-1x_coco.py │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ │ ├── infer_hrnet_fp16_accuracy.sh │ │ │ │ │ └── infer_hrnet_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── build_engine.py │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── common.py │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── fcos_hrnetv2p-w18-gn-head_4xb4-1x_coco.py │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_hrnet_fp16_accuracy.sh │ │ │ │ └── infer_hrnet_fp16_performance.sh │ │ ├── igie_common │ │ │ └── build_engine.py │ │ ├── ixrt_common │ │ │ ├── atss_r50_fpn_1x_coco.py │ │ │ ├── build_engine.py │ │ │ ├── build_nms_engine.py │ │ │ ├── calibration_dataset.py │ │ │ ├── coco_labels.py │ │ │ ├── common.py │ │ │ ├── config │ │ │ │ ├── YOLOV5M_CONFIG │ │ │ │ ├── YOLOV5S_CONFIG │ │ │ │ └── YOLOV7_CONFIG │ │ │ ├── cut_model.py │ │ │ ├── datasets │ │ │ │ ├── __init__.py │ │ │ │ ├── coco.py │ │ │ │ ├── common.py │ │ │ │ ├── post_process.py │ │ │ │ ├── pre_process.py │ │ │ │ └── vision.py │ │ │ ├── deploy.py │ │ │ ├── fcos_r50_caffe_fpn_gn-head_1x_coco.py │ │ │ ├── inference.py │ │ │ ├── inference_mmdet.py │ │ │ ├── load_ixrt_plugin.py │ │ │ ├── modify_batchsize.py │ │ │ ├── paa_r50_fpn_1x_coco.py │ │ │ ├── quant.py │ │ │ ├── requirements.txt │ │ │ ├── retinanet_r50_fpn_1x_coco.py │ │ │ ├── simplify_model.py │ │ │ └── yolof_r50-c5_8xb8-1x_coco.py │ │ ├── paa │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── paa_r50_fpn_1x_coco.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ │ ├── infer_paa_fp16_accuracy.sh │ │ │ │ │ └── infer_paa_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_paa_fp16_accuracy.sh │ │ │ │ └── infer_paa_fp16_performance.sh │ │ ├── retinaface │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── models │ │ │ │ │ ├── net.py │ │ │ │ │ └── retinaface.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ │ ├── infer_retinaface_fp16_accuracy.sh │ │ │ │ │ └── infer_retinaface_fp16_performance.sh │ │ │ │ ├── utils │ │ │ │ │ ├── box_utils.py │ │ │ │ │ ├── post_process.py │ │ │ │ │ ├── prior_box.py │ │ │ │ │ └── py_cpu_nms.py │ │ │ │ └── widerface_evaluate │ │ │ │ │ ├── README.md │ │ │ │ │ ├── box_overlaps.c │ │ │ │ │ ├── box_overlaps.pyx │ │ │ │ │ ├── evaluation.py │ │ │ │ │ ├── ground_truth │ │ │ │ │ ├── wider_easy_val.mat │ │ │ │ │ ├── wider_face_val.mat │ │ │ │ │ ├── wider_hard_val.mat │ │ │ │ │ └── wider_medium_val.mat │ │ │ │ │ └── setup.py │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── __init__.py │ │ │ │ ├── box_overlaps.c │ │ │ │ ├── box_overlaps.pyx │ │ │ │ ├── build_engine.py │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── common.py │ │ │ │ ├── config │ │ │ │ └── RETINAFACE_CONFIG │ │ │ │ ├── evaluation.py │ │ │ │ ├── inference.py │ │ │ │ ├── models │ │ │ │ ├── __init__.py │ │ │ │ ├── config.py │ │ │ │ ├── net.py │ │ │ │ ├── prior_box.py │ │ │ │ └── retinaface.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ ├── infer_retinaface_fp16_accuracy.sh │ │ │ │ └── infer_retinaface_fp16_performance.sh │ │ │ │ ├── setup.py │ │ │ │ ├── torch2onnx.py │ │ │ │ ├── utils │ │ │ │ ├── __init__.py │ │ │ │ ├── box_utils.py │ │ │ │ └── nms │ │ │ │ │ └── py_cpu_nms.py │ │ │ │ └── wider_face_dataset.py │ │ ├── retinanet │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── retinanet_r50_fpn_1x_coco.py │ │ │ │ └── scripts │ │ │ │ │ ├── infer_retinanet_fp16_accuracy.sh │ │ │ │ │ └── infer_retinanet_fp16_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_retinanet_fp16_accuracy.sh │ │ │ │ └── infer_retinanet_fp16_performance.sh │ │ ├── rtmdet │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── rtmdet_nano_320-8xb32_coco-person.py │ │ │ │ └── scripts │ │ │ │ ├── infer_rtmdet_fp16_accuracy.sh │ │ │ │ └── infer_rtmdet_fp16_performance.sh │ │ ├── sabl │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── sabl-retinanet_r50_fpn_1x_coco.py │ │ │ │ └── scripts │ │ │ │ ├── infer_sabl_fp16_accuracy.sh │ │ │ │ └── infer_sabl_fp16_performance.sh │ │ ├── ssd │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ ├── infer_ssd_fp16_accuracy.sh │ │ │ │ └── infer_ssd_fp16_performance.sh │ │ │ │ └── ssd300_coco.py │ │ ├── yolof │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ │ ├── infer_yolof_fp16_accuracy.sh │ │ │ │ │ └── infer_yolof_fp16_performance.sh │ │ │ │ └── yolof_r50-c5_8xb8-1x_coco.py │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_yolof_fp16_accuracy.sh │ │ │ │ └── infer_yolof_fp16_performance.sh │ │ ├── yolov10 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── quantize.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ │ ├── infer_yolov10_fp16_accuracy.sh │ │ │ │ │ ├── infer_yolov10_fp16_performance.sh │ │ │ │ │ ├── infer_yolov10_int8_accuracy.sh │ │ │ │ │ └── infer_yolov10_int8_performance.sh │ │ │ │ ├── utils.py │ │ │ │ └── validator.py │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── build_engine.py │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── common.py │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── quant.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_yolov10_fp16_accuracy.sh │ │ │ │ └── infer_yolov10_fp16_performance.sh │ │ ├── yolov11 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── quantize.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ │ ├── infer_yolov11_fp16_accuracy.sh │ │ │ │ │ ├── infer_yolov11_fp16_performance.sh │ │ │ │ │ ├── infer_yolov11_int8_accuracy.sh │ │ │ │ │ └── infer_yolov11_int8_performance.sh │ │ │ │ └── validator.py │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── build_engine.py │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── common.py │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── quant.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_yolov11_fp16_accuracy.sh │ │ │ │ └── infer_yolov11_fp16_performance.sh │ │ ├── yolov12 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── quantize.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ │ ├── infer_yolov12_fp16_accuracy.sh │ │ │ │ │ ├── infer_yolov12_fp16_performance.sh │ │ │ │ │ ├── infer_yolov12_int8_accuracy.sh │ │ │ │ │ └── infer_yolov12_int8_performance.sh │ │ │ │ ├── utils.py │ │ │ │ └── validator.py │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── build_engine.py │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── common.py │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── quant.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_yolov12_fp16_accuracy.sh │ │ │ │ └── infer_yolov12_fp16_performance.sh │ │ ├── yolov13 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── quantize.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ │ ├── infer_yolov13_fp16_accuracy.sh │ │ │ │ │ ├── infer_yolov13_fp16_performance.sh │ │ │ │ │ ├── infer_yolov13_int8_accuracy.sh │ │ │ │ │ └── infer_yolov13_int8_performance.sh │ │ │ │ ├── utils.py │ │ │ │ └── validator.py │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── build_engine.py │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── common.py │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── quant.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_yolov13_fp16_accuracy.sh │ │ │ │ └── infer_yolov13_fp16_performance.sh │ │ ├── yolov3 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── quantize.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ │ ├── infer_yolov3_fp16_accuracy.sh │ │ │ │ │ ├── infer_yolov3_fp16_performance.sh │ │ │ │ │ ├── infer_yolov3_int8_accuracy.sh │ │ │ │ │ └── infer_yolov3_int8_performance.sh │ │ │ │ └── utils.py │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── config │ │ │ │ └── YOLOV3_CONFIG │ │ │ │ └── scripts │ │ │ │ ├── infer_yolov3_fp16_accuracy.sh │ │ │ │ ├── infer_yolov3_fp16_performance.sh │ │ │ │ ├── infer_yolov3_int8_accuracy.sh │ │ │ │ └── infer_yolov3_int8_performance.sh │ │ ├── yolov4 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── quantize.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ │ ├── infer_yolov4_fp16_accuracy.sh │ │ │ │ │ ├── infer_yolov4_fp16_performance.sh │ │ │ │ │ ├── infer_yolov4_int8_accuracy.sh │ │ │ │ │ └── infer_yolov4_int8_performance.sh │ │ │ │ └── utils.py │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── build_engine.py │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── coco_labels.py │ │ │ │ ├── common.py │ │ │ │ ├── cut_model.py │ │ │ │ ├── deploy.py │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── load_ixrt_plugin.py │ │ │ │ ├── quant.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_yolov4_fp16_accuracy.sh │ │ │ │ ├── infer_yolov4_fp16_performance.sh │ │ │ │ ├── infer_yolov4_int8_accuracy.sh │ │ │ │ └── infer_yolov4_int8_performance.sh │ │ ├── yolov5 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── quantize.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ │ ├── infer_yolov5_fp16_accuracy.sh │ │ │ │ │ ├── infer_yolov5_fp16_performance.sh │ │ │ │ │ ├── infer_yolov5_int8_accuracy.sh │ │ │ │ │ └── infer_yolov5_int8_performance.sh │ │ │ │ └── utils.py │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_yolov5_fp16_accuracy.sh │ │ │ │ ├── infer_yolov5_fp16_performance.sh │ │ │ │ ├── infer_yolov5_int8_accuracy.sh │ │ │ │ └── infer_yolov5_int8_performance.sh │ │ ├── yolov5s │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── build_engine.py │ │ │ │ ├── build_nms_engine.py │ │ │ │ ├── calibration_dataset.py │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── coco_labels.py │ │ │ │ ├── common.py │ │ │ │ ├── config │ │ │ │ │ └── YOLOV5S_CONFIG │ │ │ │ ├── cut_model.py │ │ │ │ ├── datasets │ │ │ │ │ ├── __init__.py │ │ │ │ │ ├── coco.py │ │ │ │ │ ├── common.py │ │ │ │ │ ├── post_process.py │ │ │ │ │ ├── pre_process.py │ │ │ │ │ └── vision.py │ │ │ │ ├── deploy.py │ │ │ │ ├── inference.py │ │ │ │ ├── load_ixrt_plugin.py │ │ │ │ ├── modify_batchsize.py │ │ │ │ ├── quant.py │ │ │ │ ├── scripts │ │ │ │ │ ├── infer_yolov5s_fp16_accuracy.sh │ │ │ │ │ └── infer_yolov5s_fp16_performance.sh │ │ │ │ └── simplify_model.py │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_yolov5s_fp16_accuracy.sh │ │ │ │ ├── infer_yolov5s_fp16_performance.sh │ │ │ │ ├── infer_yolov5s_int8_accuracy.sh │ │ │ │ └── infer_yolov5s_int8_performance.sh │ │ ├── yolov6 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ │ ├── infer_yolov6_fp16_accuracy.sh │ │ │ │ │ └── infer_yolov6_fp16_performance.sh │ │ │ │ └── utils.py │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── build_engine.py │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── common.py │ │ │ │ ├── deploy.py │ │ │ │ ├── inference.py │ │ │ │ ├── quant.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_yolov6_fp16_accuracy.sh │ │ │ │ ├── infer_yolov6_fp16_performance.sh │ │ │ │ ├── infer_yolov6_int8_accuracy.sh │ │ │ │ └── infer_yolov6_int8_performance.sh │ │ ├── yolov7 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── inference.py │ │ │ │ ├── quantize.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ │ ├── infer_yolov7_fp16_accuracy.sh │ │ │ │ │ ├── infer_yolov7_fp16_performance.sh │ │ │ │ │ ├── infer_yolov7_int8_accuracy.sh │ │ │ │ │ └── infer_yolov7_int8_performance.sh │ │ │ │ └── utils.py │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── scripts │ │ │ │ ├── infer_yolov7_fp16_accuracy.sh │ │ │ │ ├── infer_yolov7_fp16_performance.sh │ │ │ │ ├── infer_yolov7_int8_accuracy.sh │ │ │ │ └── infer_yolov7_int8_performance.sh │ │ ├── yolov8 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── quantize.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ │ ├── infer_yolov8_fp16_accuracy.sh │ │ │ │ │ ├── infer_yolov8_fp16_performance.sh │ │ │ │ │ ├── infer_yolov8_int8_accuracy.sh │ │ │ │ │ └── infer_yolov8_int8_performance.sh │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── build_engine.py │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── common.py │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── quant.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_yolov8_fp16_accuracy.sh │ │ │ │ ├── infer_yolov8_fp16_performance.sh │ │ │ │ ├── infer_yolov8_int8_accuracy.sh │ │ │ │ └── infer_yolov8_int8_performance.sh │ │ ├── yolov8n │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── quantize.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_yolov8n_fp16_accuracy.sh │ │ │ │ ├── infer_yolov8n_fp16_performance.sh │ │ │ │ ├── infer_yolov8n_int8_accuracy.sh │ │ │ │ └── infer_yolov8n_int8_performance.sh │ │ ├── yolov9 │ │ │ ├── igie │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── quantize.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ │ ├── infer_yolov9_fp16_accuracy.sh │ │ │ │ │ ├── infer_yolov9_fp16_performance.sh │ │ │ │ │ ├── infer_yolov9_int8_accuracy.sh │ │ │ │ │ └── infer_yolov9_int8_performance.sh │ │ │ │ ├── utils.py │ │ │ │ └── validator.py │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── build_engine.py │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── common.py │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── quant.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_yolov9_fp16_accuracy.sh │ │ │ │ └── infer_yolov9_fp16_performance.sh │ │ └── yolox │ │ │ ├── igie │ │ │ ├── README.md │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ ├── inference.py │ │ │ ├── quantize.py │ │ │ ├── requirements.txt │ │ │ ├── scripts │ │ │ │ ├── infer_yolox_fp16_accuracy.sh │ │ │ │ ├── infer_yolox_fp16_performance.sh │ │ │ │ ├── infer_yolox_int8_accuracy.sh │ │ │ │ └── infer_yolox_int8_performance.sh │ │ │ └── utils.py │ │ │ └── ixrt │ │ │ ├── README.md │ │ │ ├── build_engine.py │ │ │ ├── build_engine_by_write_qparams.py │ │ │ ├── calibration_dataset.py │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ ├── cut_model.py │ │ │ ├── datasets │ │ │ ├── __init__.py │ │ │ ├── coco.py │ │ │ ├── common.py │ │ │ ├── post_process.py │ │ │ ├── pre_process.py │ │ │ └── vision.py │ │ │ ├── deploy.py │ │ │ ├── inference.py │ │ │ ├── load_ixrt_plugin.py │ │ │ ├── quant.py │ │ │ ├── requirements.txt │ │ │ ├── scripts │ │ │ ├── infer_yolox_fp16_accuracy.sh │ │ │ ├── infer_yolox_fp16_performance.sh │ │ │ ├── infer_yolox_int8_accuracy.sh │ │ │ └── infer_yolox_int8_performance.sh │ │ │ └── utils.py │ ├── ocr │ │ ├── kie_layoutxlm │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── build_engine.py │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ ├── infer_kie_layoutxlm_fp16_accuracy.sh │ │ │ │ └── infer_kie_layoutxlm_fp16_performance.sh │ │ │ │ └── ser_vi_layoutxlm_xfund_zh.yml │ │ └── svtr │ │ │ └── igie │ │ │ ├── README.md │ │ │ ├── build_engine.py │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ ├── inference.py │ │ │ ├── rec_svtr_tiny_6local_6global_stn_en.yml │ │ │ ├── requirements.txt │ │ │ └── scripts │ │ │ ├── infer_svtr_fp16_accuracy.sh │ │ │ └── infer_svtr_fp16_performance.sh │ ├── pose_estimation │ │ ├── hrnetpose │ │ │ └── igie │ │ │ │ ├── README.md │ │ │ │ ├── build_engine.py │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── deploy_default.py │ │ │ │ ├── export.py │ │ │ │ ├── inference.py │ │ │ │ ├── requirements.txt │ │ │ │ ├── scripts │ │ │ │ ├── infer_hrnetpose_fp16_accuracy.sh │ │ │ │ └── infer_hrnetpose_fp16_performance.sh │ │ │ │ └── td-hm_hrnet-w32_8xb64-210e_coco-256x192.py │ │ ├── lightweight_openpose │ │ │ └── ixrt │ │ │ │ ├── README.md │ │ │ │ ├── build_engine.py │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── common.py │ │ │ │ ├── datasets │ │ │ │ ├── __init__.py │ │ │ │ ├── coco.py │ │ │ │ └── transformations.py │ │ │ │ ├── inference_accuracy.py │ │ │ │ ├── inference_performance.py │ │ │ │ ├── modules │ │ │ │ ├── __init__.py │ │ │ │ ├── keypoints.py │ │ │ │ └── pose.py │ │ │ │ ├── requirements.txt │ │ │ │ └── scripts │ │ │ │ ├── infer_lightweight_openpose_fp16_accuracy.sh │ │ │ │ └── infer_lightweight_openpose_fp16_performance.sh │ │ └── rtmpose │ │ │ ├── igie │ │ │ ├── README.md │ │ │ ├── build_engine.py │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ ├── deploy_default.py │ │ │ ├── export.py │ │ │ ├── inference.py │ │ │ ├── requirements.txt │ │ │ ├── rtmpose-m_8xb256-420e_coco-256x192.py │ │ │ └── scripts │ │ │ │ ├── infer_rtmpose_fp16_accuracy.sh │ │ │ │ └── infer_rtmpose_fp16_performance.sh │ │ │ └── ixrt │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ ├── demo │ │ │ └── demo.jpg │ │ │ ├── deploy_default.py │ │ │ ├── export.py │ │ │ ├── predict.py │ │ │ ├── requirements.txt │ │ │ ├── result.jpg │ │ │ ├── rtmpose-m_8xb256-420e_coco-256x192.py │ │ │ └── tensorrt_common.py │ └── semantic_segmentation │ │ └── unet │ │ └── igie │ │ ├── README.md │ │ ├── build_engine.py │ │ ├── ci │ │ └── prepare.sh │ │ ├── deploy_default.py │ │ ├── export.py │ │ ├── fcn_unet_s5-d16_4x4_512x1024_160k_cityscapes.py │ │ ├── inference.py │ │ ├── requirements.txt │ │ └── scripts │ │ ├── infer_unet_fp16_accuracy.sh │ │ └── infer_unet_fp16_performance.sh ├── multimodal │ ├── diffusion_model │ │ ├── stable-diffusion-1.5 │ │ │ └── diffusers │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── demo.py │ │ │ │ └── requirements.txt │ │ └── stable-diffusion-3 │ │ │ └── diffusers │ │ │ ├── README.md │ │ │ └── demo_sd3.py │ └── vision_language_model │ │ ├── aria │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ └── offline_inference_vision_language.py │ │ ├── chameleon_7b │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ └── offline_inference_vision_language.py │ │ ├── clip │ │ └── ixformer │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ └── inference.py │ │ ├── deepseek-ocr │ │ └── transformers │ │ │ └── README.md │ │ ├── e5-v │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ └── offline_inference_vision_language_embedding.py │ │ ├── fuyu_8b │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ └── offline_inference_vision_language.py │ │ ├── glm-4v │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ └── offline_inference_vision_language.py │ │ ├── h2vol │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ └── offline_inference_vision_language.py │ │ ├── idefics3 │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ └── offline_inference_vision_language.py │ │ ├── intern_vl │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ └── offline_inference_vision_language.py │ │ ├── llama-3.2 │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ └── offline_inference_vision_language.py │ │ ├── llava │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ └── offline_inference_vision_language.py │ │ ├── llava_next_video_7b │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ ├── offline_inference_vision_language.py │ │ │ ├── utils.py │ │ │ └── video-example-data │ │ │ └── sample_demo_1.mp4 │ │ ├── minicpm_o │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ └── offline_inference_vision_language.py │ │ ├── minicpm_v │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ └── offline_inference_vision_language.py │ │ ├── nvlm │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ └── offline_inference_vision_language.py │ │ ├── paligemma │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ └── offline_inference_vision_language.py │ │ ├── phi3_v │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ └── offline_inference_vision_language.py │ │ ├── pixtral │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ ├── offline_inference_2411_w4a16.py │ │ │ └── offline_inference_vision_language.py │ │ ├── qwen2_5_vl │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ └── offline_inference_vision_language.py │ │ ├── qwen2_vl │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ └── offline_inference_vision_language.py │ │ ├── qwen_vl │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ └── offline_inference_vision_language.py │ │ ├── step3 │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── bf162int8.py │ │ │ ├── bf16Toint4.py │ │ │ ├── eval_dataset.py │ │ │ ├── eval_dataset_w8a8.py │ │ │ └── requirements.txt │ │ ├── utils.py │ │ ├── vllm_public_assets │ │ ├── American_Eskimo_Dog.jpg │ │ └── cherry_blossom.jpg │ │ └── xlmroberta │ │ └── vllm │ │ ├── README.md │ │ ├── ci │ │ └── prepare.sh │ │ ├── offline_inference_embedding.py │ │ └── offline_inference_scoring.py ├── nlp │ ├── llm │ │ ├── baichuan2-7b │ │ │ └── vllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── convert2int8.py │ │ │ │ ├── offline_inference.py │ │ │ │ └── template_baichuan.jinja │ │ ├── chatglm3-6b-32k │ │ │ └── vllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── offline_inference.py │ │ │ │ └── server_inference.py │ │ ├── chatglm3-6b │ │ │ └── vllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── offline_inference.py │ │ │ │ └── server_inference.py │ │ ├── deepseek-r1-distill-llama-70b │ │ │ └── vllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── offline_inference.py │ │ ├── deepseek-r1-distill-llama-8b │ │ │ └── vllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── offline_inference.py │ │ ├── deepseek-r1-distill-qwen-1.5b │ │ │ └── vllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── offline_inference.py │ │ ├── deepseek-r1-distill-qwen-14b │ │ │ └── vllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── offline_inference.py │ │ ├── deepseek-r1-distill-qwen-32b │ │ │ └── vllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── offline_inference.py │ │ ├── deepseek-r1-distill-qwen-7b │ │ │ └── vllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── offline_inference.py │ │ ├── ernie-4.5-21b-a3b │ │ │ └── fastdeploy │ │ │ │ ├── README.md │ │ │ │ ├── run_demo.py │ │ │ │ └── run_demo.sh │ │ ├── ernie-4.5-300b-a47b │ │ │ └── fastdeploy │ │ │ │ ├── README.md │ │ │ │ ├── run_demo.py │ │ │ │ └── run_demo.sh │ │ ├── internlm3 │ │ │ └── lmdeploy │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── offline_inference.py │ │ ├── llama2-13b │ │ │ └── trtllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── convert_checkpoint.py │ │ │ │ ├── run.py │ │ │ │ ├── scripts │ │ │ │ ├── requirements.txt │ │ │ │ ├── set_environment.sh │ │ │ │ ├── test_trtllm_llama2_13b_gpu2.sh │ │ │ │ └── test_trtllm_llama2_13b_gpu2_build.sh │ │ │ │ ├── summarize.py │ │ │ │ └── utils.py │ │ ├── llama2-70b │ │ │ └── trtllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── convert_checkpoint.py │ │ │ │ ├── run.py │ │ │ │ ├── scripts │ │ │ │ ├── requirements.txt │ │ │ │ ├── set_environment.sh │ │ │ │ ├── test_trtllm_llama2_70b_gpu8.sh │ │ │ │ └── test_trtllm_llama2_70b_gpu8_build.sh │ │ │ │ ├── summarize.py │ │ │ │ └── utils.py │ │ ├── llama2-7b │ │ │ ├── trtllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ ├── convert_checkpoint.py │ │ │ │ ├── run.py │ │ │ │ ├── scripts │ │ │ │ │ ├── requirements.txt │ │ │ │ │ ├── set_environment.sh │ │ │ │ │ ├── test_trtllm_llama2_7b_gpu1.sh │ │ │ │ │ └── test_trtllm_llama2_7b_gpu1_build.sh │ │ │ │ ├── summarize.py │ │ │ │ └── utils.py │ │ │ └── vllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ ├── offline_inference.py │ │ │ │ └── template_llama.jinja │ │ ├── llama3-70b │ │ │ └── vllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── offline_inference.py │ │ ├── qwen-7b │ │ │ └── vllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── offline_inference.py │ │ ├── qwen1.5-14b │ │ │ └── vllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── offline_inference.py │ │ ├── qwen1.5-32b │ │ │ └── vllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── offline_inference.py │ │ ├── qwen1.5-72b │ │ │ └── vllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── offline_inference.py │ │ ├── qwen1.5-7b │ │ │ ├── tgi │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ │ └── prepare.sh │ │ │ │ └── offline_inference.py │ │ │ └── vllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── offline_inference.py │ │ ├── qwen2-72b │ │ │ └── vllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── offline_inference.py │ │ ├── qwen2-7b │ │ │ └── vllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── offline_inference.py │ │ ├── qwen3-235b │ │ │ └── vllm │ │ │ │ ├── README.md │ │ │ │ ├── bf16ToInt4.py │ │ │ │ └── ci │ │ │ │ └── prepare.sh │ │ ├── stablelm │ │ │ └── vllm │ │ │ │ ├── README.md │ │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ │ └── offline_inference.py │ │ └── utils.py │ └── plm │ │ ├── albert │ │ └── ixrt │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ ├── requirements.txt │ │ │ ├── scripts │ │ │ ├── infer_albert_fp16_performance.sh │ │ │ └── prepare_model_and_dataset.sh │ │ │ └── torch2onnx.py │ │ ├── bert_base_ner │ │ └── igie │ │ │ ├── Int8QAT │ │ │ ├── bert4torch │ │ │ │ ├── __init__.py │ │ │ │ ├── activations.py │ │ │ │ ├── layers.py │ │ │ │ ├── losses.py │ │ │ │ ├── models.py │ │ │ │ ├── optimizers.py │ │ │ │ ├── snippets.py │ │ │ │ └── tokenizers.py │ │ │ ├── export_hdf5.py │ │ │ ├── ls_hf_transformer_layer.py │ │ │ ├── pytorch_quantization │ │ │ │ ├── __init__.py │ │ │ │ ├── calib │ │ │ │ │ ├── __init__.py │ │ │ │ │ ├── calibrator.py │ │ │ │ │ ├── histogram.py │ │ │ │ │ └── max.py │ │ │ │ ├── nn │ │ │ │ │ ├── __init__.py │ │ │ │ │ ├── _functions │ │ │ │ │ │ ├── __init__.py │ │ │ │ │ │ └── quant_rnn.py │ │ │ │ │ ├── functional.py │ │ │ │ │ └── modules │ │ │ │ │ │ ├── __init__.py │ │ │ │ │ │ ├── _utils.py │ │ │ │ │ │ ├── clip.py │ │ │ │ │ │ ├── quant_bert.py │ │ │ │ │ │ ├── quant_conv.py │ │ │ │ │ │ ├── quant_instancenorm.py │ │ │ │ │ │ ├── quant_linear.py │ │ │ │ │ │ ├── quant_pooling.py │ │ │ │ │ │ ├── quant_rnn.py │ │ │ │ │ │ └── tensor_quantizer.py │ │ │ │ ├── optim │ │ │ │ │ ├── __init__.py │ │ │ │ │ └── helper.py │ │ │ │ ├── quant_modules.py │ │ │ │ ├── tensor_quant.py │ │ │ │ ├── utils │ │ │ │ │ ├── __init__.py │ │ │ │ │ ├── quant_logging.py │ │ │ │ │ └── reduce_amax.py │ │ │ │ └── version.py │ │ │ ├── quantization.py │ │ │ ├── run_qat.py │ │ │ └── training_ops │ │ │ │ ├── layer_base.py │ │ │ │ ├── pytorch │ │ │ │ ├── export.py │ │ │ │ └── export_quant.py │ │ │ │ ├── torch_transformer_layers.py │ │ │ │ └── util.py │ │ │ ├── README.md │ │ │ ├── build_engine.py │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ ├── get_weights.py │ │ │ ├── inference.py │ │ │ ├── requirements.txt │ │ │ └── scripts │ │ │ ├── infer_bert_base_ner_int8_accuracy.sh │ │ │ └── infer_bert_base_ner_int8_performance.sh │ │ ├── bert_base_squad │ │ ├── igie │ │ │ ├── README.md │ │ │ ├── build_engine.py │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ ├── export.py │ │ │ ├── inference.py │ │ │ ├── requirements.txt │ │ │ └── scripts │ │ │ │ ├── infer_bert_base_squad_fp16_accuracy.sh │ │ │ │ └── infer_bert_base_squad_fp16_performance.sh │ │ └── ixrt │ │ │ ├── README.md │ │ │ ├── builder.py │ │ │ ├── builder_utils.py │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ ├── evaluate-v1.1.py │ │ │ ├── helpers │ │ │ ├── __init__.py │ │ │ ├── calibrator.py │ │ │ ├── data_processing.py │ │ │ └── tokenization.py │ │ │ ├── inference.py │ │ │ ├── load_ixrt_plugin.py │ │ │ ├── perf.py │ │ │ ├── requirements.txt │ │ │ └── scripts │ │ │ ├── infer_bert_base_squad_fp16_accuracy.sh │ │ │ └── infer_bert_base_squad_fp16_performance.sh │ │ ├── bert_large_squad │ │ ├── igie │ │ │ ├── Int8QAT │ │ │ │ ├── evaluate.py │ │ │ │ ├── export_hdf5.py │ │ │ │ ├── ls_hf_transformer_layer.py │ │ │ │ ├── pytorch_quantization │ │ │ │ │ ├── __init__.py │ │ │ │ │ ├── calib │ │ │ │ │ │ ├── __init__.py │ │ │ │ │ │ ├── calibrator.py │ │ │ │ │ │ ├── histogram.py │ │ │ │ │ │ └── max.py │ │ │ │ │ ├── nn │ │ │ │ │ │ ├── __init__.py │ │ │ │ │ │ ├── _functions │ │ │ │ │ │ │ ├── __init__.py │ │ │ │ │ │ │ └── quant_rnn.py │ │ │ │ │ │ ├── functional.py │ │ │ │ │ │ └── modules │ │ │ │ │ │ │ ├── __init__.py │ │ │ │ │ │ │ ├── _utils.py │ │ │ │ │ │ │ ├── clip.py │ │ │ │ │ │ │ ├── quant_bert.py │ │ │ │ │ │ │ ├── quant_conv.py │ │ │ │ │ │ │ ├── quant_instancenorm.py │ │ │ │ │ │ │ ├── quant_linear.py │ │ │ │ │ │ │ ├── quant_pooling.py │ │ │ │ │ │ │ ├── quant_rnn.py │ │ │ │ │ │ │ └── tensor_quantizer.py │ │ │ │ │ ├── optim │ │ │ │ │ │ ├── __init__.py │ │ │ │ │ │ └── helper.py │ │ │ │ │ ├── quant_modules.py │ │ │ │ │ ├── tensor_quant.py │ │ │ │ │ ├── utils │ │ │ │ │ │ ├── __init__.py │ │ │ │ │ │ ├── quant_logging.py │ │ │ │ │ │ └── reduce_amax.py │ │ │ │ │ └── version.py │ │ │ │ ├── quantization.py │ │ │ │ ├── run_qat.py │ │ │ │ ├── run_qat.sh │ │ │ │ ├── squad_data.py │ │ │ │ ├── squad_download.py │ │ │ │ ├── trainer_qa.py │ │ │ │ ├── training_ops │ │ │ │ │ ├── layer_base.py │ │ │ │ │ ├── pytorch │ │ │ │ │ │ ├── export.py │ │ │ │ │ │ └── export_quant.py │ │ │ │ │ ├── torch_transformer_layers.py │ │ │ │ │ └── util.py │ │ │ │ └── utils_qa.py │ │ │ ├── README.md │ │ │ ├── build_engine.py │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ ├── export.py │ │ │ ├── inference.py │ │ │ ├── requirements.txt │ │ │ └── scripts │ │ │ │ ├── infer_bert_large_squad_fp16_accuracy.sh │ │ │ │ ├── infer_bert_large_squad_fp16_performance.sh │ │ │ │ ├── infer_bert_large_squad_int8_accuracy.sh │ │ │ │ └── infer_bert_large_squad_int8_performance.sh │ │ └── ixrt │ │ │ ├── README.md │ │ │ ├── builder.py │ │ │ ├── builder_int8.py │ │ │ ├── builder_utils.py │ │ │ ├── builder_utils_int8.py │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ ├── evaluate-v1.1.py │ │ │ ├── helpers │ │ │ ├── __init__.py │ │ │ ├── calibrator.py │ │ │ ├── data_processing.py │ │ │ └── tokenization.py │ │ │ ├── inference.py │ │ │ ├── load_ixrt_plugin.py │ │ │ ├── perf.py │ │ │ └── scripts │ │ │ ├── infer_bert_large_squad_fp16_accuracy.sh │ │ │ ├── infer_bert_large_squad_fp16_performance.sh │ │ │ ├── infer_bert_large_squad_int8_accuracy.sh │ │ │ └── infer_bert_large_squad_int8_performance.sh │ │ ├── deberta │ │ └── ixrt │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ ├── remove_clip_and_cast.py │ │ │ ├── requirements.txt │ │ │ ├── scripts │ │ │ ├── infer_deberta_fp16_performance.sh │ │ │ └── prepare_model_and_dataset.sh │ │ │ └── torch2onnx.py │ │ ├── roberta │ │ └── ixrt │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ ├── export_onnx.py │ │ │ ├── gen_data.py │ │ │ ├── requirements.txt │ │ │ └── scripts │ │ │ └── infer_roberta_fp16_performance.sh │ │ ├── roformer │ │ └── ixrt │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ ├── deploy.py │ │ │ ├── export_onnx.py │ │ │ ├── requirements.txt │ │ │ └── scripts │ │ │ └── infer_roformer_fp16_performance.sh │ │ ├── transformer │ │ ├── build_helpers │ │ │ ├── __init__.py │ │ │ ├── build_helpers.py │ │ │ ├── get_vendored.py │ │ │ └── test_helpers.py │ │ ├── igie │ │ │ ├── __init__.py │ │ │ ├── build_engine.py │ │ │ ├── builder_utils.py │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ ├── common.py │ │ │ ├── inference_wmt14_en_fr_fp16_accuracy.py │ │ │ ├── inference_wmt14_en_fr_fp16_performance.py │ │ │ ├── load_ixrt_plugin.py │ │ │ ├── plugin_utils.py │ │ │ ├── requirements.txt │ │ │ ├── scripts │ │ │ │ ├── infer_transformer_fp16_accuracy.sh │ │ │ │ └── infer_transformer_fp16_performance.sh │ │ │ └── transformer_cfg.py │ │ ├── ixrt │ │ │ ├── build_engine.py │ │ │ ├── ci │ │ │ │ └── prepare.sh │ │ │ ├── common.py │ │ │ ├── inference_wmt14_en_fr_fp16_accuracy.py │ │ │ ├── inference_wmt14_en_fr_fp16_accuracy_plugin.py │ │ │ ├── inference_wmt14_en_fr_fp16_performance.py │ │ │ ├── inference_wmt14_en_fr_fp16_performance_plugin.py │ │ │ ├── requirements.txt │ │ │ └── scripts │ │ │ │ ├── infer_transformer_fp16_accuracy.sh │ │ │ │ └── infer_transformer_fp16_performance.sh │ │ ├── omegaconf.py │ │ ├── plugin │ │ │ ├── __init__.py │ │ │ ├── build_engine.py │ │ │ ├── builder_utils.py │ │ │ ├── load_ixrt_plugin.py │ │ │ ├── plugin_utils.py │ │ │ ├── transformer_cfg.py │ │ │ └── trt.py │ │ └── setup.py │ │ └── videobert │ │ └── ixrt │ │ ├── README.md │ │ ├── ci │ │ └── prepare.sh │ │ ├── requirements.txt │ │ └── scripts │ │ ├── infer_videobert_fp16_performance.sh │ │ └── prepare_model_and_dataset.sh ├── others │ └── recommendation │ │ └── wide_and_deep │ │ └── ixrt │ │ ├── README.md │ │ ├── change2dynamic.py │ │ ├── ci │ │ └── prepare.sh │ │ ├── deploy.py │ │ ├── export_onnx.py │ │ ├── requirements.txt │ │ └── scripts │ │ └── infer_wide_and_deep_fp16_performance.sh └── speech │ ├── asr │ ├── ultravox │ │ └── vllm │ │ │ ├── README.md │ │ │ ├── ci │ │ │ └── prepare.sh │ │ │ └── offline_inference_audio_language.py │ ├── utils.py │ └── whisper │ │ └── vllm │ │ ├── README.md │ │ ├── ci │ │ └── prepare.sh │ │ └── offline_inference_audio_language.py │ └── speech_synthesis │ └── cosyvoice │ └── pytorch │ ├── README.md │ ├── ci │ └── prepare.sh │ ├── inference_test.py │ └── requirements.txt ├── tests ├── README.md ├── model_info.json ├── run_igie.py ├── run_ixrt.py ├── run_trtllm.py ├── run_vllm.py └── utils.py └── toolbox └── ByteMLPerf ├── .gitignore ├── CONTRIBUTING.md ├── LICENSE ├── NOTICE ├── README.md ├── README.zh_CN.md ├── byte_infer_perf └── general_perf │ ├── README.md │ ├── README.zh_CN.md │ ├── __init__.py │ ├── backends │ ├── CPU │ │ ├── calculate_cpu_diff.py │ │ ├── calculate_cpu_diff.sh │ │ ├── compile_backend_cpu.py │ │ ├── requirements.txt │ │ └── runtime_backend_cpu.py │ ├── ILUVATAR │ │ ├── README.zh_CN.md │ │ ├── common.py │ │ ├── compile_backend_iluvatar.py │ │ ├── requirements.txt │ │ ├── runtime_backend_iluvatar.py │ │ └── utils │ │ │ ├── __init__.py │ │ │ ├── argument.py │ │ │ ├── coco_metric.py │ │ │ ├── compile_engine.py │ │ │ ├── dataloader.py │ │ │ ├── fastCoCoeval │ │ │ ├── __init__.py │ │ │ ├── cocoeval │ │ │ │ ├── cocoeval.cpp │ │ │ │ └── cocoeval.h │ │ │ ├── fast_coco_eval_api.py │ │ │ └── jit_ops.py │ │ │ ├── file.py │ │ │ ├── imagenet_metric.py │ │ │ ├── import_model.py │ │ │ ├── mod_rewriter.py │ │ │ ├── onnx_rewrite_batch_size.py │ │ │ ├── onnx_util.py │ │ │ ├── quantization.py │ │ │ ├── stauts_checker.py │ │ │ ├── target.py │ │ │ └── timer.py │ ├── compile_backend.py │ └── runtime_backend.py │ ├── core │ ├── __init__.py │ ├── configs │ │ ├── __init__.py │ │ ├── backend_store.py │ │ ├── dataset_store.py │ │ └── workload_store.py │ └── perf_engine.py │ ├── datasets │ ├── data_loader.py │ ├── fake_dataset │ │ ├── data_loader.py │ │ └── test_accuracy.py │ ├── open_cail2019 │ │ ├── data_loader.py │ │ ├── pre_process_data.py │ │ └── test_accuracy.py │ ├── open_cifar │ │ ├── data_loader.py │ │ └── test_accuracy.py │ ├── open_criteo_kaggle │ │ ├── data_loader.py │ │ ├── preprocess_dataset.py │ │ └── test_accuracy.py │ ├── open_imagenet │ │ ├── data_loader.py │ │ └── test_accuracy.py │ ├── open_squad │ │ ├── bert │ │ │ ├── accuracy_squad.py │ │ │ └── evaluate.py │ │ ├── create_squad_data.py │ │ ├── data_loader.py │ │ └── test_accuracy.py │ └── test_accuracy.py │ ├── launch.py │ ├── model_zoo │ ├── albert-torch-fp32.json │ ├── bert-tf-fp32.json │ ├── bert-torch-fp32.json │ ├── chatglm2-6b-torch-fp16.json │ ├── clip-onnx-fp32.json │ ├── conformer-encoder-onnx-fp32.json │ ├── deberta-torch-fp32.json │ ├── gpt2-torch-fp32.json │ ├── llama2-7b-torch-fp16.json │ ├── resnet50-tf-fp32.json │ ├── resnet50-torch-fp32.json │ ├── roberta-torch-fp32.json │ ├── roformer-tf-fp32.json │ ├── swin-large-torch-fp32.json │ ├── unet-onnx-fp32.json │ ├── vae-decoder-onnx-fp32.json │ ├── vae-encoder-onnx-fp32.json │ ├── videobert-onnx-fp32.json │ ├── widedeep-tf-fp32.json │ └── yolov5-onnx-fp32.json │ ├── prepare_model_and_dataset.sh │ ├── requirements.txt │ ├── tools │ ├── build_pdf.py │ ├── convert.sh │ ├── frozen_to_saved.py │ ├── h5_to_frozen.py │ ├── model_trt_convert.py │ ├── mxnet_to_onnx.py │ ├── onnx_utils.py │ ├── requirements.txt │ ├── saved_to_frozen.py │ ├── saved_to_onnx.py │ ├── tf_fp32_to_fp16.py │ ├── tf_utils.py │ └── torch_to_onnx.py │ └── version.py ├── byte_micro_perf ├── README.md ├── backends │ ├── GPU │ │ ├── backend_gpu.py │ │ ├── custom_ops.py │ │ └── requirements.txt │ ├── ILUVATAR │ │ ├── backend_iluvatar.py │ │ ├── custom_ops.py │ │ └── requirements.txt │ ├── __init__.py │ ├── backend.py │ ├── module_store.py │ └── utils.py ├── compiled_cache.db ├── core │ └── perf_engine.py ├── launch.py └── requirements.txt ├── docs └── images │ ├── flowchat.png │ └── icon.png ├── pylint.conf └── vendor_zoo ├── AWS ├── mem_16.png └── mem_17.png ├── Graphcore ├── image_12.png └── mem_12.png ├── Habana ├── image_14.png ├── mem_14.png └── pe_14.png ├── Moffett ├── image_7.png ├── image_8.png ├── image_9.png ├── mem_7.png ├── mem_8.png ├── mem_9.png ├── pe_7.png └── pe_9.png ├── NVIDIA ├── image_0.png ├── image_1.png ├── image_2.png ├── image_3.png ├── image_4.png ├── image_5.png └── image_6.png ├── QUALCOMM ├── image_10.png ├── mem_10.png └── pe_10.png └── Stream └── image_13.png /.gitee/ISSUE_TEMPLATE.zh-CN.md: -------------------------------------------------------------------------------- 1 | ### 该问题是怎么引起的？ 2 | 3 | 4 | 5 | ### 重现步骤 6 | 7 | 8 | 9 | ### 报错信息 10 | 11 | 12 | 13 | 14 | -------------------------------------------------------------------------------- /.gitee/PULL_REQUEST_TEMPLATE.zh-CN.md: -------------------------------------------------------------------------------- 1 | ### 内容说明（相关的Issue） 2 | 3 | 4 | 5 | -------------------------------------------------------------------------------- /data/datasets/README.md: -------------------------------------------------------------------------------- 1 | # This is the default datasets location required by inference models 2 | -------------------------------------------------------------------------------- /models/audio/speech_recognition/conformer/igie/load_ixrt_plugin.py: -------------------------------------------------------------------------------- 1 | import ctypes 2 | import tensorrt 3 | from os.path import join, dirname, exists 4 | def load_ixrt_plugin(logger=tensorrt.Logger(tensorrt.Logger.INFO), namespace="", dynamic_path=""): 5 | if not dynamic_path: 6 | dynamic_path = join(dirname(tensorrt.__file__), "lib", "libixrt_plugin.so") 7 | if not exists(dynamic_path): 8 | raise FileNotFoundError( 9 | f"The ixrt_plugin lib {dynamic_path} is not existed, please provided effective plugin path!") 10 | ctypes.CDLL(dynamic_path) 11 | tensorrt.init_libnvinfer_plugins(logger, namespace) 12 | print(f"Loaded plugin from {dynamic_path}") 13 | -------------------------------------------------------------------------------- /models/audio/speech_recognition/conformer/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx==1.18.0 3 | onnxsim 4 | librosa 5 | soundfile 6 | pycuda -------------------------------------------------------------------------------- /models/audio/speech_recognition/conformer/igie/tools/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /models/audio/speech_recognition/conformer/ixrt/load_ixrt_plugin.py: -------------------------------------------------------------------------------- 1 | import ctypes 2 | import tensorrt 3 | from os.path import join, dirname, exists 4 | def load_ixrt_plugin(logger=tensorrt.Logger(tensorrt.Logger.INFO), namespace="", dynamic_path=""): 5 | if not dynamic_path: 6 | dynamic_path = join(dirname(tensorrt.__file__), "lib", "libixrt_plugin.so") 7 | if not exists(dynamic_path): 8 | raise FileNotFoundError( 9 | f"The ixrt_plugin lib {dynamic_path} is not existed, please provided effective plugin path!") 10 | ctypes.CDLL(dynamic_path) 11 | tensorrt.init_libnvinfer_plugins(logger, namespace) 12 | print(f"Loaded plugin from {dynamic_path}") -------------------------------------------------------------------------------- /models/audio/speech_recognition/conformer/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx==1.18.0 3 | onnxsim 4 | librosa 5 | soundfile -------------------------------------------------------------------------------- /models/audio/speech_recognition/conformer/ixrt/tools/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /models/audio/speech_recognition/transformer_asr/ixrt/build.sh: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env bash 2 | 3 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 4 | # All Rights Reserved. 5 | # 6 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 7 | # not use this file except in compliance with the License. You may obtain 8 | # a copy of the License at 9 | # 10 | # http://www.apache.org/licenses/LICENSE-2.0 11 | # 12 | # Unless required by applicable law or agreed to in writing, software 13 | # distributed under the License is distributed on an "AS IS" BASIS, WITHOUT 14 | # WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the 15 | # License for the specific language governing permissions and limitations 16 | # under the License. 17 | 18 | for i in fast* 19 | do 20 | cd $i 21 | bash build.sh 22 | cd .. 23 | done 24 | -------------------------------------------------------------------------------- /models/audio/speech_recognition/transformer_asr/ixrt/faster_cat/__init__.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from faster_cat import sp_opt 3 | 4 | def FastCat(inputs,dim=0): 5 | if len(inputs) == 2 and dim==0: 6 | a,b = inputs 7 | in_shape = a.shape 8 | if len(in_shape)>1: 9 | res, = sp_opt.test_opt_2(a.view(a.shape[0],-1),b.view(b.shape[0],-1)) 10 | new_shape = (a.shape[0]+b.shape[0],) + in_shape[1:] 11 | res = res.view(*new_shape) 12 | return res 13 | return torch.cat(inputs,dim=dim) -------------------------------------------------------------------------------- /models/audio/speech_recognition/transformer_asr/ixrt/faster_cat/build.sh: -------------------------------------------------------------------------------- 1 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 2 | # All Rights Reserved. 3 | # 4 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 5 | # not use this file except in compliance with the License. You may obtain 6 | # a copy of the License at 7 | # 8 | # http://www.apache.org/licenses/LICENSE-2.0 9 | # 10 | # Unless required by applicable law or agreed to in writing, software 11 | # distributed under the License is distributed on an "AS IS" BASIS, WITHOUT 12 | # WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the 13 | # License for the specific language governing permissions and limitations 14 | # under the License. 15 | set -euox pipefail 16 | 17 | rm -rf build 18 | rm -rf *.so 19 | 20 | python3 setup.py build 21 | 22 | cp build/lib*/*.so . -------------------------------------------------------------------------------- /models/audio/speech_recognition/transformer_asr/ixrt/faster_cat/test.cpp: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | #include 6 | #include 7 | 8 | #include 9 | #include 10 | #include 11 | 12 | 13 | std::vector one_test_opt_2(at::Tensor a, at::Tensor b); 14 | 15 | std::vector test_opt_2(at::Tensor a, at::Tensor b) { 16 | return one_test_opt_2(a, b); 17 | } 18 | 19 | PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) { 20 | m.def("test_opt_2", &test_opt_2, ""); 21 | } 22 | -------------------------------------------------------------------------------- /models/audio/speech_recognition/transformer_asr/ixrt/faster_layer_norm/__init__.py: -------------------------------------------------------------------------------- 1 | import torch 2 | from faster_layer_norm import sp_opt 3 | 4 | class FasterLayerNorm(torch.nn.Module): 5 | def __init__(self, weight, bias): 6 | super(FasterLayerNorm, self).__init__() 7 | self.weight = weight 8 | self.bias = bias 9 | 10 | def forward(self, inputs, *args, **kwargs): 11 | hidden_size = self.weight.size(0) 12 | in_shape = inputs.shape 13 | inputs = inputs.view(-1,hidden_size) 14 | output, = sp_opt.test_opt(inputs,self.weight,self.bias) 15 | output = output.view(*in_shape) 16 | return output 17 | -------------------------------------------------------------------------------- /models/audio/speech_recognition/transformer_asr/ixrt/faster_layer_norm/build.sh: -------------------------------------------------------------------------------- 1 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 2 | # All Rights Reserved. 3 | # 4 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 5 | # not use this file except in compliance with the License. You may obtain 6 | # a copy of the License at 7 | # 8 | # http://www.apache.org/licenses/LICENSE-2.0 9 | # 10 | # Unless required by applicable law or agreed to in writing, software 11 | # distributed under the License is distributed on an "AS IS" BASIS, WITHOUT 12 | # WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the 13 | # License for the specific language governing permissions and limitations 14 | # under the License. 15 | set -euox pipefail 16 | 17 | rm -rf build 18 | rm -rf *.so 19 | 20 | python3 setup.py build 21 | 22 | cp build/lib*/*.so . -------------------------------------------------------------------------------- /models/audio/speech_recognition/transformer_asr/ixrt/faster_layer_norm/test.cpp: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | #include 6 | #include 7 | 8 | #include 9 | #include 10 | #include 11 | 12 | std::vector one_test_opt(at::Tensor input, at::Tensor ln_weight, 13 | at::Tensor ln_bias); 14 | 15 | std::vector test_opt(at::Tensor input, at::Tensor ln_weight, 16 | at::Tensor ln_bias) { 17 | return one_test_opt(input, ln_weight, ln_bias); 18 | } 19 | 20 | PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) { 21 | m.def("test_opt", &test_opt, "fast depthwise conv1d forward"); 22 | } 23 | -------------------------------------------------------------------------------- /models/audio/speech_recognition/transformer_asr/ixrt/faster_logsumexp/build.sh: -------------------------------------------------------------------------------- 1 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 2 | # All Rights Reserved. 3 | # 4 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 5 | # not use this file except in compliance with the License. You may obtain 6 | # a copy of the License at 7 | # 8 | # http://www.apache.org/licenses/LICENSE-2.0 9 | # 10 | # Unless required by applicable law or agreed to in writing, software 11 | # distributed under the License is distributed on an "AS IS" BASIS, WITHOUT 12 | # WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the 13 | # License for the specific language governing permissions and limitations 14 | # under the License. 15 | set -euox pipefail 16 | 17 | rm -rf build 18 | rm -rf *.so 19 | 20 | python3 setup.py build 21 | 22 | cp build/lib*/*.so . -------------------------------------------------------------------------------- /models/audio/speech_recognition/transformer_asr/ixrt/faster_logsumexp/test.cpp: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | #include 6 | #include 7 | 8 | #include 9 | #include 10 | #include 11 | 12 | std::vector one_test_opt(at::Tensor input); 13 | 14 | std::vector test_opt(at::Tensor input) { 15 | return one_test_opt(input); 16 | } 17 | 18 | std::vector one_test_dim0(at::Tensor input); 19 | 20 | std::vector test_opt_dim0(at::Tensor input) { 21 | return one_test_dim0(input); 22 | } 23 | 24 | PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) { 25 | m.def("test_opt", &test_opt, ""); 26 | m.def("test_opt_dim0", &test_opt_dim0, ""); 27 | } 28 | -------------------------------------------------------------------------------- /models/audio/speech_recognition/transformer_asr/ixrt/faster_stack/build.sh: -------------------------------------------------------------------------------- 1 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 2 | # All Rights Reserved. 3 | # 4 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 5 | # not use this file except in compliance with the License. You may obtain 6 | # a copy of the License at 7 | # 8 | # http://www.apache.org/licenses/LICENSE-2.0 9 | # 10 | # Unless required by applicable law or agreed to in writing, software 11 | # distributed under the License is distributed on an "AS IS" BASIS, WITHOUT 12 | # WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the 13 | # License for the specific language governing permissions and limitations 14 | # under the License. 15 | set -euox pipefail 16 | 17 | rm -rf build 18 | rm -rf *.so 19 | 20 | python3 setup.py build 21 | 22 | cp build/lib*/*.so . -------------------------------------------------------------------------------- /models/audio/speech_recognition/transformer_asr/ixrt/faster_stack/test.cpp: -------------------------------------------------------------------------------- 1 | #include 2 | #include 3 | #include 4 | #include 5 | #include 6 | #include 7 | 8 | #include 9 | #include 10 | #include 11 | 12 | std::vector one_test_opt(at::Tensor a, at::Tensor b, at::Tensor c, 13 | at::Tensor d); 14 | 15 | std::vector test_opt(at::Tensor a, at::Tensor b, at::Tensor c, 16 | at::Tensor d) { 17 | return one_test_opt(a, b, c, d); 18 | } 19 | 20 | std::vector one_test_opt_2(at::Tensor a, at::Tensor b); 21 | 22 | std::vector test_opt_2(at::Tensor a, at::Tensor b) { 23 | return one_test_opt_2(a, b); 24 | } 25 | 26 | PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) { 27 | m.def("test_opt", &test_opt, ""); 28 | m.def("test_opt_2", &test_opt_2, ""); 29 | } 30 | -------------------------------------------------------------------------------- /models/audio/speech_recognition/transformer_asr/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | speechbrain==0.5.13 2 | requests -------------------------------------------------------------------------------- /models/benchmark/internvl2/README.md: -------------------------------------------------------------------------------- 1 | # InternVL2-26B 2 | 3 | * 下载数据 4 | 5 | 数据链接: [InternVL2-26B](https://huggingface.co/OpenGVLab/InternVL2-26B) 6 | 7 | * 测试 InternVL2-26B 8 | 9 | ```bash 10 | cd .. 11 | bash test_performance_server_multimodal.sh --model /path/to/model -tp 4 --max-num-seqs 32 --max-num-batched-tokens 8192 --max-model-len 8192 --host 127.0.0.1 --port 12345 --trust-remote-code,--model /path/to/model --host 127.0.0.1 --port 12345 --num-prompts 1 --output-tokens 128 --image-path test.jpg --trust-remote-code true --image-size "512,512" 12 | ``` 13 | -------------------------------------------------------------------------------- /models/benchmark/llava.jinja: -------------------------------------------------------------------------------- 1 | {%- if messages[0]['role'] == 'system' -%} 2 | {%- set system_message = messages[0]['content'] -%} 3 | {%- set messages = messages[1:] -%} 4 | {%- else -%} 5 | {% set system_message = '' -%} 6 | {%- endif -%} 7 | 8 | {{ bos_token + system_message }} 9 | {%- for message in messages -%} 10 | {%- if (message['role'] == 'user') != (loop.index0 % 2 == 0) -%} 11 | {{ raise_exception('Conversation roles must alternate user/assistant/user/assistant/...') }} 12 | {%- endif -%} 13 | 14 | {%- if message['role'] == 'user' -%} 15 | {{ 'USER: ' + message['content'] + '\n' }} 16 | {%- elif message['role'] == 'assistant' -%} 17 | {{ 'ASSISTANT: ' + message['content'] + eos_token + '\n' }} 18 | {%- endif -%} 19 | {%- endfor -%} 20 | 21 | {%- if add_generation_prompt -%} 22 | {{ 'ASSISTANT:' }} 23 | {% endif %} -------------------------------------------------------------------------------- /models/benchmark/longbench/config/dataset2maxlen.json: -------------------------------------------------------------------------------- 1 | { 2 | "narrativeqa": 128, 3 | "qasper": 128, 4 | "multifieldqa_en": 64, 5 | "multifieldqa_zh": 64, 6 | "hotpotqa": 32, 7 | "2wikimqa": 32, 8 | "musique": 32, 9 | "dureader": 128, 10 | "gov_report": 512, 11 | "qmsum": 512, 12 | "multi_news": 512, 13 | "vcsum": 512, 14 | "trec": 64, 15 | "triviaqa": 32, 16 | "samsum": 128, 17 | "lsht": 64, 18 | "passage_count": 32, 19 | "passage_retrieval_en": 32, 20 | "passage_retrieval_zh": 32, 21 | "lcc": 64, 22 | "repobench-p": 64 23 | } 24 | -------------------------------------------------------------------------------- /models/benchmark/longbench/config/model2maxlen.json: -------------------------------------------------------------------------------- 1 | { 2 | "llama2-7b-chat": 3584, 3 | "llama2-13b-chat": 3584, 4 | "llama3-8b-chat": 7680, 5 | "llama3-70b-chat": 7680, 6 | "baichuan2-13b-chat": 3584, 7 | "qwen2-72b-chat-awq": 32256, 8 | "llama3.1-8b-chat": 32256, 9 | "qwen2.5-14b-chat": 32256, 10 | "qwen2-vl-chat": 32256 11 | } 12 | -------------------------------------------------------------------------------- /models/benchmark/longbench/config/model2path.json: -------------------------------------------------------------------------------- 1 | { 2 | "llama2-7b-chat-4k": "meta-llama/Llama-2-7b-chat-hf", 3 | "chatglm2-6b": "THUDM/chatglm2-6b", 4 | "chatglm2-6b-32k": "THUDM/chatglm2-6b-32k", 5 | "vicuna-v1.5-7b-16k": "lmsys/vicuna-7b-v1.5-16k", 6 | "llama-2-7B-32K-Instruct": "togethercomputer/Llama-2-7B-32K-Instruct" 7 | } 8 | -------------------------------------------------------------------------------- /models/benchmark/longbench/result_record/llama3.1-8b-chat.jsonl: -------------------------------------------------------------------------------- 1 | { 2 | "narrativeqa": 27.7, 3 | "qasper": 45.17, 4 | "multifieldqa_en": 55.0, 5 | "multifieldqa_zh": 62.36, 6 | "hotpotqa": 55.3, 7 | "2wikimqa": 43.68, 8 | "musique": 30.76, 9 | "dureader": 33.84, 10 | "gov_report": 35.17, 11 | "qmsum": 25.3, 12 | "multi_news": 27.41, 13 | "vcsum": 17.06, 14 | "trec": 72.5, 15 | "triviaqa": 91.65, 16 | "samsum": 43.7, 17 | "lsht": 46.5, 18 | "passage_count": 7.12, 19 | "passage_retrieval_en": 99.5, 20 | "passage_retrieval_zh": 96.96, 21 | "lcc": 63.08, 22 | "repobench-p": 57.02 23 | } -------------------------------------------------------------------------------- /models/benchmark/longbench/result_record/qwen2-vl-chat.jsonl: -------------------------------------------------------------------------------- 1 | { 2 | "narrativeqa": 25.02, 3 | "qasper": 31.85, 4 | "multifieldqa_en": 51.41, 5 | "multifieldqa_zh": 59.48, 6 | "hotpotqa": 55.05, 7 | "2wikimqa": 48.26, 8 | "musique": 31.67, 9 | "dureader": 33.51, 10 | "gov_report": 31.03, 11 | "qmsum": 25.22, 12 | "multi_news": 26.47, 13 | "vcsum": 18.46, 14 | "trec": 80.0, 15 | "triviaqa": 81.89, 16 | "samsum": 43.91, 17 | "lsht": 39.0, 18 | "passage_count": 4.0, 19 | "passage_retrieval_en": 74.5, 20 | "passage_retrieval_zh": 74.0, 21 | "lcc": 53.89, 22 | "repobench-p": 39.81 23 | } -------------------------------------------------------------------------------- /models/benchmark/longbench/result_record/qwen2.5-14b-chat.jsonl: -------------------------------------------------------------------------------- 1 | { 2 | "narrativeqa": 28.33, 3 | "qasper": 45.5, 4 | "multifieldqa_en": 54.52, 5 | "multifieldqa_zh": 62.73, 6 | "hotpotqa": 62.55, 7 | "2wikimqa": 58.52, 8 | "musique": 37.71, 9 | "dureader": 32.12, 10 | "gov_report": 32.54, 11 | "qmsum": 24.95, 12 | "multi_news": 24.62, 13 | "vcsum": 16.11, 14 | "trec": 77.5, 15 | "triviaqa": 90.06, 16 | "samsum": 47.39, 17 | "lsht": 49.17, 18 | "passage_count": 12.88, 19 | "passage_retrieval_en": 98.75, 20 | "passage_retrieval_zh": 98.58, 21 | "lcc": 65.23, 22 | "repobench-p": 52.61 23 | } -------------------------------------------------------------------------------- /models/benchmark/mixtral/README.md: -------------------------------------------------------------------------------- 1 | # Mixtral 测试 2 | 3 | * 下载数据 4 | 5 | 联系您的应用工程师获取 6 | 7 | * 测试 Mixtral-8x22B-W8A8 8 | 9 | ```bash 10 | cd .. 11 | bash test_performance_server.sh --model /path/to/model -tp 8 --host 127.0.0.1 --port 12345 --enable-chunked-prefill=False --max-num-seqs 10 --max-num-batched-tokens 20480 --max-model-len 20480,--model /path/to/model --host 127.0.0.1 --port 12345 --num-prompts 10 --input-tokens 2048 --output-tokens 1024 12 | ``` 13 | -------------------------------------------------------------------------------- /models/benchmark/set_environment.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | set -e 17 | 18 | pip3 install -r "requirements.txt" -i https://pypi.tuna.tsinghua.edu.cn/simple 19 | pip3 install -U mpi4py -i https://pypi.tuna.tsinghua.edu.cn/simple 20 | 21 | exit $? -------------------------------------------------------------------------------- /models/benchmark/test.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/models/benchmark/test.jpg -------------------------------------------------------------------------------- /models/benchmark/test_vllm_longbench.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | set -e 17 | 18 | bash set_environment.sh 19 | 20 | python3 benchmark_longbench.py $@ 21 | 22 | exit $? -------------------------------------------------------------------------------- /models/benchmark/yi1.5/README.md: -------------------------------------------------------------------------------- 1 | # Yi1.5 测试 2 | 3 | * 下载数据 4 | 5 | 数据链接: [Yi](https://huggingface.co/01-ai/Yi-1.5-34B-Chat) 6 | 7 | * 测试 Yi-1.5-34B-chat 8 | 9 | ```bash 10 | cd .. 11 | bash test_performance_server.sh --model /path/to/model -tp 4 --host 127.0.0.1 --port 12345 --max-num-batched-tokens 20480 --max-num-seqs 10,--model /path/to/model --host 127.0.0.1 --port 12345 --num-prompts 10 --input-tokens 2048 --output-tokens 1024 12 | ``` 13 | -------------------------------------------------------------------------------- /models/cv/classification/alexnet/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name alexnet --weight alexnet-owt-7be5be79.pth --output alexnet.onnx -------------------------------------------------------------------------------- /models/cv/classification/clip/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r requirements.txt 20 | python3 export.py --output clip.onnx 21 | onnxsim clip.onnx clip_opt.onnx -------------------------------------------------------------------------------- /models/cv/classification/clip/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnxsim 3 | transformers==4.33.2 4 | -------------------------------------------------------------------------------- /models/cv/classification/clip/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnxsim 3 | transformers==4.33.2 -------------------------------------------------------------------------------- /models/cv/classification/conformer_base/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r requirements.txt 20 | python3 export.py --weight conformer_base_patch16.pth --output conformer_base.onnx 21 | onnxsim conformer_base.onnx conformer_base_opt.onnx -------------------------------------------------------------------------------- /models/cv/classification/conformer_base/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | timm 4 | onnxsim 5 | -------------------------------------------------------------------------------- /models/cv/classification/convnext_base/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name convnext_base --weight convnext_base-6075fbad.pth --output convnext_base.onnx -------------------------------------------------------------------------------- /models/cv/classification/convnext_small/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name convnext_small --weight convnext_small-0c510722.pth --output convnext_small.onnx -------------------------------------------------------------------------------- /models/cv/classification/convnext_tiny/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name convnext_tiny --weight convnext_tiny-983f1562.pth --output convnext_tiny.onnx -------------------------------------------------------------------------------- /models/cv/classification/densenet121/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name densenet121 --weight densenet121-a639ec97.pth --output densenet121.onnx -------------------------------------------------------------------------------- /models/cv/classification/densenet161/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name densenet161 --weight densenet161-8d451a50.pth --output densenet161.onnx -------------------------------------------------------------------------------- /models/cv/classification/densenet169/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name densenet169 --weight densenet169-b2777c0a.pth --output densenet169.onnx -------------------------------------------------------------------------------- /models/cv/classification/densenet201/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name densenet201 --weight densenet201-c1103571.pth --output densenet201.onnx -------------------------------------------------------------------------------- /models/cv/classification/efficientnet_b0/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name efficientnet_b0 --weight efficientnet_b0_rwightman-7f5810bc.pth --output efficientnet_b0.onnx -------------------------------------------------------------------------------- /models/cv/classification/efficientnet_b1/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name efficientnet_b1 --weight efficientnet_b1-c27df63c.pth --output efficientnet_b1.onnx -------------------------------------------------------------------------------- /models/cv/classification/efficientnet_b2/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name efficientnet_b2 --weight efficientnet_b2_rwightman-c35c1473.pth --output efficientnet_b2.onnx -------------------------------------------------------------------------------- /models/cv/classification/efficientnet_b3/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name efficientnet_b3 --weight efficientnet_b3_rwightman-b3899882.pth --output efficientnet_b3.onnx -------------------------------------------------------------------------------- /models/cv/classification/efficientnet_b4/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name efficientnet_b4 --weight efficientnet_b4_rwightman-23ab8bcd.pth --output efficientnet_b4.onnx -------------------------------------------------------------------------------- /models/cv/classification/efficientnet_b5/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name efficientnet_b5 --weight efficientnet_b5_lukemelas-1a07897c.pth --output efficientnet_b5.onnx 21 | -------------------------------------------------------------------------------- /models/cv/classification/efficientnet_b6/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | -------------------------------------------------------------------------------- /models/cv/classification/efficientnet_b7/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | -------------------------------------------------------------------------------- /models/cv/classification/efficientnet_v2/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name efficientnet_v2_m --weight efficientnet_v2_m-dc08266a.pth --output efficientnet_v2_m.onnx -------------------------------------------------------------------------------- /models/cv/classification/efficientnet_v2_s/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name efficientnet_v2_s --weight efficientnet_v2_s-dd5fe13b.pth --output efficientnet_v2_s.onnx -------------------------------------------------------------------------------- /models/cv/classification/googlenet/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name googlenet --weight googlenet-1378be20.pth --output googlenet.onnx -------------------------------------------------------------------------------- /models/cv/classification/igie_common/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim -------------------------------------------------------------------------------- /models/cv/classification/inception_v3/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name inception_v3 --weight inception_v3_google-0cc3c7bd.pth --output inception_v3.onnx -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/README.md: -------------------------------------------------------------------------------- 1 | # 此代码是分类网络基于imagenet数据集的通用实现 -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/config/ALEXNET_CONFIG: -------------------------------------------------------------------------------- 1 | # IMGSIZE : 模型输入hw大小 2 | # MODEL_NAME : 生成onnx/engine的basename 3 | # ORIGINE_MODEL : 原始onnx文件名称 4 | IMGSIZE=224 5 | MODEL_NAME=AlexNet 6 | ORIGINE_MODEL=alexnet.onnx 7 | 8 | # QUANT CONFIG (仅PRECISION为int8时生效) 9 | # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] 10 | # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) 11 | # QUANT_STEP : 量化步数 12 | # QUANT_SEED : 随机种子 保证量化结果可复现 13 | # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 14 | QUANT_OBSERVER=hist_percentile 15 | QUANT_BATCHSIZE=32 16 | QUANT_STEP=32 17 | QUANT_SEED=42 18 | DISABLE_QUANT_LIST= 19 | QUANT_EXIST_ONNX= 20 | -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/config/CLIP_CONFIG: -------------------------------------------------------------------------------- 1 | # IMGSIZE : 模型输入hw大小 2 | # MODEL_NAME : 生成onnx/engine的basename 3 | # ORIGINE_MODEL : 原始onnx文件名称 4 | IMGSIZE=224 5 | MODEL_NAME=clip 6 | ORIGINE_MODEL=clip.onnx -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/config/CSPRESNET50_CONFIG: -------------------------------------------------------------------------------- 1 | # IMGSIZE : 模型输入hw大小 2 | # MODEL_NAME : 生成onnx/engine的basename 3 | # ORIGINE_MODEL : 原始onnx文件名称 4 | IMGSIZE=224 5 | MODEL_NAME=CSPResNet50 6 | ORIGINE_MODEL=cspresnet50.onnx 7 | 8 | # QUANT CONFIG (仅PRECISION为int8时生效) 9 | # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] 10 | # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) 11 | # QUANT_STEP : 量化步数 12 | # QUANT_SEED : 随机种子 保证量化结果可复现 13 | # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 14 | QUANT_OBSERVER=hist_percentile 15 | QUANT_BATCHSIZE=32 16 | QUANT_STEP=32 17 | QUANT_SEED=42 18 | DISABLE_QUANT_LIST= 19 | QUANT_EXIST_ONNX= 20 | -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/config/HRNET_W18_CONFIG: -------------------------------------------------------------------------------- 1 | # IMGSIZE : 模型输入hw大小 2 | # MODEL_NAME : 生成onnx/engine的basename 3 | # ORIGINE_MODEL : 原始onnx文件名称 4 | IMGSIZE=224 5 | MODEL_NAME=hrnet_w18 6 | ORIGINE_MODEL=hrnet_w18.onnx 7 | 8 | # QUANT CONFIG (仅PRECISION为int8时生效) 9 | # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] 10 | # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) 11 | # QUANT_STEP : 量化步数 12 | # QUANT_SEED : 随机种子 保证量化结果可复现 13 | # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 14 | QUANT_OBSERVER=hist_percentile 15 | QUANT_BATCHSIZE=32 16 | QUANT_STEP=32 17 | QUANT_SEED=42 18 | DISABLE_QUANT_LIST= 19 | QUANT_EXIST_ONNX= 20 | -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/config/INCEPTION_RESNET_V2_CONFIG: -------------------------------------------------------------------------------- 1 | # IMGSIZE : 模型输入hw大小 2 | # MODEL_NAME : 生成onnx/engine的basename 3 | # ORIGINE_MODEL : 原始onnx文件名称 4 | IMGSIZE=299 5 | MODEL_NAME=inceptionresnetv2 6 | ORIGINE_MODEL=inceptionresnetv2.onnx 7 | 8 | # QUANT CONFIG (仅PRECISION为int8时生效) 9 | # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] 10 | # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) 11 | # QUANT_STEP : 量化步数 12 | # QUANT_SEED : 随机种子 保证量化结果可复现 13 | # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 14 | QUANT_OBSERVER=hist_percentile 15 | QUANT_BATCHSIZE=64 16 | QUANT_STEP=32 17 | QUANT_SEED=42 18 | DISABLE_QUANT_LIST= 19 | QUANT_EXIST_ONNX= 20 | -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/config/INCEPTION_V3_CONFIG: -------------------------------------------------------------------------------- 1 | # IMGSIZE : 模型输入hw大小 2 | # MODEL_NAME : 生成onnx/engine的basename 3 | # ORIGINE_MODEL : 原始onnx文件名称 4 | IMGSIZE=224 5 | MODEL_NAME=inception_v3 6 | ORIGINE_MODEL=inception_v3.onnx 7 | 8 | # QUANT CONFIG (仅PRECISION为int8时生效) 9 | # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] 10 | # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) 11 | # QUANT_STEP : 量化步数 12 | # QUANT_SEED : 随机种子 保证量化结果可复现 13 | # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 14 | QUANT_OBSERVER=hist_percentile 15 | QUANT_BATCHSIZE=1 16 | QUANT_STEP=32 17 | QUANT_SEED=42 18 | DISABLE_QUANT_LIST= 19 | QUANT_EXIST_ONNX= 20 | 21 | -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/config/MOBILENET_V2_CONFIG: -------------------------------------------------------------------------------- 1 | # IMGSIZE : 模型输入hw大小 2 | # MODEL_NAME : 生成onnx/engine的basename 3 | # ORIGINE_MODEL : 原始onnx文件名称 4 | IMGSIZE=224 5 | MODEL_NAME=mobilenet_v2 6 | ORIGINE_MODEL=mobilenet_v2.onnx 7 | 8 | # QUANT CONFIG (仅PRECISION为int8时生效) 9 | # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] 10 | # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) 11 | # QUANT_STEP : 量化步数 12 | # QUANT_SEED : 随机种子 保证量化结果可复现 13 | # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 14 | QUANT_OBSERVER=hist_percentile 15 | QUANT_BATCHSIZE=32 16 | QUANT_STEP=32 17 | QUANT_SEED=42 18 | DISABLE_QUANT_LIST= 19 | QUANT_EXIST_ONNX= -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/config/REPVGG_CONFIG: -------------------------------------------------------------------------------- 1 | # IMGSIZE : 模型输入hw大小 2 | # MODEL_NAME : 生成onnx/engine的basename 3 | # ORIGINE_MODEL : 原始onnx文件名称 4 | IMGSIZE=224 5 | MODEL_NAME=RepVGG 6 | ORIGINE_MODEL=repvgg_A0.onnx 7 | 8 | # QUANT CONFIG (仅PRECISION为int8时生效) 9 | # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] 10 | # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) 11 | # QUANT_STEP : 量化步数 12 | # QUANT_SEED : 随机种子 保证量化结果可复现 13 | # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 14 | QUANT_OBSERVER=hist_percentile 15 | QUANT_BATCHSIZE=32 16 | QUANT_STEP=32 17 | QUANT_SEED=42 18 | DISABLE_QUANT_LIST= 19 | QUANT_EXIST_ONNX= 20 | -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/config/REPVIT_CONFIG: -------------------------------------------------------------------------------- 1 | # IMGSIZE : 模型输入hw大小 2 | # MODEL_NAME : 生成onnx/engine的basename 3 | # ORIGINE_MODEL : 原始onnx文件名称 4 | IMGSIZE=224 5 | MODEL_NAME=RepViT 6 | ORIGINE_MODEL=repvit.onnx 7 | 8 | # QUANT CONFIG (仅PRECISION为int8时生效) 9 | # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] 10 | # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) 11 | # QUANT_STEP : 量化步数 12 | # QUANT_SEED : 随机种子 保证量化结果可复现 13 | # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 14 | QUANT_OBSERVER=hist_percentile 15 | QUANT_BATCHSIZE=1 16 | QUANT_STEP=32 17 | QUANT_SEED=42 18 | DISABLE_QUANT_LIST= 19 | QUANT_EXIST_ONNX= 20 | -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/config/SHUFFLENETV2_X0_5_CONFIG: -------------------------------------------------------------------------------- 1 | # IMGSIZE : 模型输入hw大小 2 | # MODEL_NAME : 生成onnx/engine的basename 3 | # ORIGINE_MODEL : 原始onnx文件名称 4 | IMGSIZE=224 5 | MODEL_NAME=shufflenetv2_x0.5 6 | ORIGINE_MODEL=shufflenetv2_x0_5.onnx 7 | 8 | # QUANT CONFIG (仅PRECISION为int8时生效) 9 | # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] 10 | # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) 11 | # QUANT_STEP : 量化步数 12 | # QUANT_SEED : 随机种子 保证量化结果可复现 13 | # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 14 | QUANT_OBSERVER=hist_percentile 15 | QUANT_BATCHSIZE=1 16 | QUANT_STEP=32 17 | QUANT_SEED=42 18 | DISABLE_QUANT_LIST= 19 | QUANT_EXIST_ONNX= 20 | -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/config/SHUFFLENETV2_X1_0_CONFIG: -------------------------------------------------------------------------------- 1 | # IMGSIZE : 模型输入hw大小 2 | # MODEL_NAME : 生成onnx/engine的basename 3 | # ORIGINE_MODEL : 原始onnx文件名称 4 | IMGSIZE=224 5 | MODEL_NAME=shufflenetv2_x1.0 6 | ORIGINE_MODEL=shufflenetv2_x1_0.onnx 7 | 8 | # QUANT CONFIG (仅PRECISION为int8时生效) 9 | # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] 10 | # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) 11 | # QUANT_STEP : 量化步数 12 | # QUANT_SEED : 随机种子 保证量化结果可复现 13 | # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 14 | QUANT_OBSERVER=hist_percentile 15 | QUANT_BATCHSIZE=1 16 | QUANT_STEP=32 17 | QUANT_SEED=42 18 | DISABLE_QUANT_LIST= 19 | QUANT_EXIST_ONNX= 20 | -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/config/SHUFFLENETV2_X1_5_CONFIG: -------------------------------------------------------------------------------- 1 | # IMGSIZE : 模型输入hw大小 2 | # MODEL_NAME : 生成onnx/engine的basename 3 | # ORIGINE_MODEL : 原始onnx文件名称 4 | IMGSIZE=224 5 | MODEL_NAME=shufflenetv2_x1.5 6 | ORIGINE_MODEL=shufflenetv2_x1_5.onnx 7 | 8 | # QUANT CONFIG (仅PRECISION为int8时生效) 9 | # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] 10 | # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) 11 | # QUANT_STEP : 量化步数 12 | # QUANT_SEED : 随机种子 保证量化结果可复现 13 | # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 14 | QUANT_OBSERVER=hist_percentile 15 | QUANT_BATCHSIZE=1 16 | QUANT_STEP=32 17 | QUANT_SEED=42 18 | DISABLE_QUANT_LIST= 19 | QUANT_EXIST_ONNX= 20 | -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/config/SHUFFLENETV2_X2_0_CONFIG: -------------------------------------------------------------------------------- 1 | # IMGSIZE : 模型输入hw大小 2 | # MODEL_NAME : 生成onnx/engine的basename 3 | # ORIGINE_MODEL : 原始onnx文件名称 4 | IMGSIZE=224 5 | MODEL_NAME=shufflenetv2_x2.0 6 | ORIGINE_MODEL=shufflenetv2_x2_0.onnx 7 | 8 | # QUANT CONFIG (仅PRECISION为int8时生效) 9 | # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] 10 | # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) 11 | # QUANT_STEP : 量化步数 12 | # QUANT_SEED : 随机种子 保证量化结果可复现 13 | # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 14 | QUANT_OBSERVER=hist_percentile 15 | QUANT_BATCHSIZE=1 16 | QUANT_STEP=32 17 | QUANT_SEED=42 18 | DISABLE_QUANT_LIST= 19 | QUANT_EXIST_ONNX= 20 | -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/config/SHUFFLENET_V1_CONFIG: -------------------------------------------------------------------------------- 1 | # IMGSIZE : 模型输入hw大小 2 | # MODEL_NAME : 生成onnx/engine的basename 3 | # ORIGINE_MODEL : 原始onnx文件名称 4 | IMGSIZE=224 5 | MODEL_NAME=ShuffleNet_v1 6 | ORIGINE_MODEL=shufflenetv1.onnx 7 | 8 | # QUANT CONFIG (仅PRECISION为int8时生效) 9 | # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] 10 | # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) 11 | # QUANT_STEP : 量化步数 12 | # QUANT_SEED : 随机种子 保证量化结果可复现 13 | # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 14 | QUANT_OBSERVER=hist_percentile 15 | QUANT_BATCHSIZE=1 16 | QUANT_STEP=32 17 | QUANT_SEED=42 18 | DISABLE_QUANT_LIST= 19 | QUANT_EXIST_ONNX= 20 | -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/config/SQUEEZENET_V1_0_CONFIG: -------------------------------------------------------------------------------- 1 | # IMGSIZE : 模型输入hw大小 2 | # MODEL_NAME : 生成onnx/engine的basename 3 | # ORIGINE_MODEL : 原始onnx文件名称 4 | IMGSIZE=224 5 | MODEL_NAME=SqueezeNet_v10 6 | ORIGINE_MODEL=squeezenetv10.onnx 7 | 8 | # QUANT CONFIG (仅PRECISION为int8时生效) 9 | # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] 10 | # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) 11 | # QUANT_STEP : 量化步数 12 | # QUANT_SEED : 随机种子 保证量化结果可复现 13 | # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 14 | QUANT_OBSERVER=hist_percentile 15 | QUANT_BATCHSIZE=1 16 | QUANT_STEP=32 17 | QUANT_SEED=42 18 | DISABLE_QUANT_LIST= 19 | QUANT_EXIST_ONNX= 20 | -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/config/SQUEEZENET_V1_1_CONFIG: -------------------------------------------------------------------------------- 1 | # IMGSIZE : 模型输入hw大小 2 | # MODEL_NAME : 生成onnx/engine的basename 3 | # ORIGINE_MODEL : 原始onnx文件名称 4 | IMGSIZE=224 5 | MODEL_NAME=SqueezeNet_v1_1 6 | ORIGINE_MODEL=squeezenet_v1_1.onnx 7 | 8 | # QUANT CONFIG (仅PRECISION为int8时生效) 9 | # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] 10 | # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) 11 | # QUANT_STEP : 量化步数 12 | # QUANT_SEED : 随机种子 保证量化结果可复现 13 | # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 14 | QUANT_OBSERVER=hist_percentile 15 | QUANT_BATCHSIZE=1 16 | QUANT_STEP=32 17 | QUANT_SEED=42 18 | DISABLE_QUANT_LIST= 19 | QUANT_EXIST_ONNX= 20 | -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/config/VGG16_CONFIG: -------------------------------------------------------------------------------- 1 | # IMGSIZE : 模型输入hw大小 2 | # MODEL_NAME : 生成onnx/engine的basename 3 | # ORIGINE_MODEL : 原始onnx文件名称 4 | IMGSIZE=224 5 | MODEL_NAME=Vgg16 6 | ORIGINE_MODEL=vgg16.onnx 7 | 8 | # QUANT CONFIG (仅PRECISION为int8时生效) 9 | # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] 10 | # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) 11 | # QUANT_STEP : 量化步数 12 | # QUANT_SEED : 随机种子 保证量化结果可复现 13 | # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 14 | QUANT_OBSERVER=hist_percentile 15 | QUANT_BATCHSIZE=1 16 | QUANT_STEP=32 17 | QUANT_SEED=42 18 | DISABLE_QUANT_LIST= 19 | QUANT_EXIST_ONNX= 20 | -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/load_ixrt_plugin.py: -------------------------------------------------------------------------------- 1 | from os.path import join, dirname, exists 2 | import tensorrt as trt 3 | import ctypes 4 | 5 | def load_ixrt_plugin(logger=trt.Logger(trt.Logger.WARNING), namespace="", dynamic_path=""): 6 | if not dynamic_path: 7 | dynamic_path = join(dirname(trt.__file__), "lib", "libixrt_plugin.so") 8 | if not exists(dynamic_path): 9 | raise FileNotFoundError( 10 | f"The ixrt_plugin lib {dynamic_path} is not existed, please provided effective plugin path!") 11 | ctypes.CDLL(dynamic_path, mode=ctypes.RTLD_GLOBAL) 12 | trt.init_libnvinfer_plugins(logger, namespace) 13 | print(f"Loaded plugin from {dynamic_path}") -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/refine_utils/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | tabulate 3 | pycuda 4 | onnx 5 | onnxsim 6 | opencv-python==4.6.0.66 -------------------------------------------------------------------------------- /models/cv/classification/ixrt_common/simplify_model.py: -------------------------------------------------------------------------------- 1 | import onnx 2 | import argparse 3 | from onnxsim import simplify 4 | 5 | # Simplify 6 | def simplify_model(args): 7 | onnx_model = onnx.load(args.origin_model) 8 | model_simp, check = simplify(onnx_model) 9 | model_simp = onnx.shape_inference.infer_shapes(model_simp) 10 | onnx.save(model_simp, args.output_model) 11 | print(" Simplify onnx Done.") 12 | 13 | def parse_args(): 14 | parser = argparse.ArgumentParser() 15 | parser.add_argument("--origin_model", type=str) 16 | parser.add_argument("--output_model", type=str) 17 | parser.add_argument("--reshape", action="store_true") 18 | args = parser.parse_args() 19 | return args 20 | 21 | args = parse_args() 22 | simplify_model(args) 23 | 24 | 25 | 26 | 27 | -------------------------------------------------------------------------------- /models/cv/classification/mnasnet0_5/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name mnasnet0_5 --weight mnasnet0.5_top1_67.823-3ffadce67e.pth --output mnasnet0_5.onnx -------------------------------------------------------------------------------- /models/cv/classification/mnasnet0_75/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name mnasnet0_75 --weight mnasnet0_75-7090bc5f.pth --output mnasnet0_75.onnx -------------------------------------------------------------------------------- /models/cv/classification/mnasnet1_0/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name mnasnet1_0 --weight mnasnet1.0_top1_73.512-f206786ef8.pth --output mnasnet1_0.onnx -------------------------------------------------------------------------------- /models/cv/classification/mnasnet1_3/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name mnasnet1_3 --weight mnasnet1_3-a4c69d6f.pth --output mnasnet1_3.onnx -------------------------------------------------------------------------------- /models/cv/classification/mobilenet_v3/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name mobilenet_v3_small --weight mobilenet_v3_small-047dcff4.pth --output mobilenetv3_small.onnx -------------------------------------------------------------------------------- /models/cv/classification/mobilenet_v3_large/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name mobilenet_v3_large --weight mobilenet_v3_large-8738ca79.pth --output mobilenetv3_large.onnx -------------------------------------------------------------------------------- /models/cv/classification/regnet_x_16gf/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name regnet_x_16gf --weight regnet_x_16gf-2007eb11.pth --output regnet_x_16gf.onnx -------------------------------------------------------------------------------- /models/cv/classification/regnet_x_1_6gf/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name regnet_x_1_6gf --weight regnet_x_1_6gf-a12f2b72.pth --output regnet_x_1_6gf.onnx -------------------------------------------------------------------------------- /models/cv/classification/regnet_x_32gf/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name regnet_x_32gf --weight regnet_x_32gf-9d47f8d0.pth --output regnet_x_32gf.onnx 21 | -------------------------------------------------------------------------------- /models/cv/classification/regnet_x_3_2gf/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name regnet_x_3_2gf --weight regnet_x_3_2gf-f342aeae.pth --output regnet_x_3_2gf.onnx -------------------------------------------------------------------------------- /models/cv/classification/regnet_x_400mf/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name regnet_x_400mf --weight regnet_x_400mf-adf1edd5.pth --output regnet_x_400mf.onnx -------------------------------------------------------------------------------- /models/cv/classification/regnet_x_800mf/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name regnet_x_800mf --weight regnet_x_800mf-94a99ebd.pth --output regnet_x_800mf.onnx -------------------------------------------------------------------------------- /models/cv/classification/regnet_x_8gf/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name regnet_x_8gf --weight regnet_x_8gf-03ceed89.pth --output regnet_x_8gf.onnx -------------------------------------------------------------------------------- /models/cv/classification/regnet_y_16gf/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name regnet_y_16gf --weight regnet_y_16gf-9e6ed7dd.pth --output regnet_y_16gf.onnx -------------------------------------------------------------------------------- /models/cv/classification/regnet_y_1_6gf/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name regnet_y_1_6gf --weight regnet_y_1_6gf-b11a554e.pth --output regnet_y_1_6gf.onnx -------------------------------------------------------------------------------- /models/cv/classification/regnet_y_32gf/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name regnet_y_32gf --weight regnet_y_32gf-4dee3f7a.pth --output regnet_y_32gf.onnx 21 | -------------------------------------------------------------------------------- /models/cv/classification/regnet_y_3_2gf/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name regnet_y_3_2gf --weight regnet_y_3_2gf-b5a9779c.pth --output regnet_y_3_2gf.onnx -------------------------------------------------------------------------------- /models/cv/classification/regnet_y_400mf/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name regnet_y_400mf --weight regnet_y_400mf-c65dace8.pth --output regnet_y_400mf.onnx 21 | -------------------------------------------------------------------------------- /models/cv/classification/resnet101/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name resnet101 --weight resnet101-63fe2227.pth --output resnet101.onnx -------------------------------------------------------------------------------- /models/cv/classification/resnet152/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name resnet152 --weight resnet152-394f9c45.pth --output resnet152.onnx -------------------------------------------------------------------------------- /models/cv/classification/resnet18/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name resnet18 --weight resnet18-f37072fd.pth --output resnet18.onnx -------------------------------------------------------------------------------- /models/cv/classification/resnet50/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name resnet50 --weight resnet50-0676ba61.pth --output resnet50.onnx -------------------------------------------------------------------------------- /models/cv/classification/resnext101_32x8d/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name resnext101_32x8d --weight resnext101_32x8d-8ba56ff5.pth --output resnext101_32x8d.onnx -------------------------------------------------------------------------------- /models/cv/classification/resnext101_64x4d/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name resnext101_64x4d --weight resnext101_64x4d-173b62eb.pth --output resnext101_64x4d.onnx -------------------------------------------------------------------------------- /models/cv/classification/resnext50_32x4d/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name resnext50_32x4d --weight resnext50_32x4d-7cdf4587.pth --output resnext50_32x4d.onnx -------------------------------------------------------------------------------- /models/cv/classification/shufflenetv2_x0_5/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name shufflenet_v2_x0_5 --weight shufflenetv2_x0.5-f707e7126e.pth --output shufflenetv2_x0_5.onnx -------------------------------------------------------------------------------- /models/cv/classification/shufflenetv2_x1_0/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip install -r requirements.txt 20 | python3 ../../igie_common/export.py --model-name shufflenet_v2_x1_0 --weight shufflenetv2_x1-5666bf0f80.pth --output shufflenetv2_x1_0.onnx -------------------------------------------------------------------------------- /models/cv/classification/shufflenetv2_x1_5/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name shufflenet_v2_x1_5 --weight shufflenetv2_x1_5-3c479a10.pth --output shufflenetv2_x1_5.onnx -------------------------------------------------------------------------------- /models/cv/classification/shufflenetv2_x2_0/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name shufflenet_v2_x2_0 --weight shufflenetv2_x2_0-8be3c8ee.pth --output shufflenetv2_x2_0.onnx -------------------------------------------------------------------------------- /models/cv/classification/squeezenet_v1_0/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name squeezenet1_0 --weight squeezenet1_0-b66bff10.pth --output squeezenet1_0.onnx -------------------------------------------------------------------------------- /models/cv/classification/squeezenet_v1_1/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name squeezenet1_1 --weight squeezenet1_1-b8a52dc0.pth --output squeezenet1_1.onnx -------------------------------------------------------------------------------- /models/cv/classification/swin_transformer/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r requirements.txt 20 | python3 export.py --output swin_transformer.onnx 21 | onnxsim swin_transformer.onnx swin_transformer_opt.onnx -------------------------------------------------------------------------------- /models/cv/classification/swin_transformer/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnxsim 3 | transformers==4.33.2 4 | -------------------------------------------------------------------------------- /models/cv/classification/swin_transformer_large/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | onnxsim 2 | onnx_graphsurgeon 3 | scikit-learn 4 | tqdm 5 | pycuda 6 | onnx 7 | tabulate 8 | pycocotools 9 | opencv-python==4.6.0.66 10 | typing-extensions==4.12.2 -------------------------------------------------------------------------------- /models/cv/classification/vgg11/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name vgg11 --weight vgg11-8a719046.pth --output vgg11.onnx -------------------------------------------------------------------------------- /models/cv/classification/vgg13/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name vgg13 --weight vgg13-19584684.pth --output vgg13.onnx -------------------------------------------------------------------------------- /models/cv/classification/vgg13_bn/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name vgg13_bn --weight vgg13_bn-abd245e5.pth --output vgg13_bn.onnx -------------------------------------------------------------------------------- /models/cv/classification/vgg16/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name vgg16 --weight vgg16-397923af.pth --output vgg16.onnx -------------------------------------------------------------------------------- /models/cv/classification/vgg19/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name vgg19 --weight vgg19-dcbb9e9d.pth --output vgg19.onnx 21 | -------------------------------------------------------------------------------- /models/cv/classification/vgg19_bn/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name vgg19_bn --weight vgg19_bn-c79401a0.pth --output vgg19_bn.onnx 21 | -------------------------------------------------------------------------------- /models/cv/classification/vit/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r requirements.txt 20 | python3 export.py --output vit.onnx 21 | onnxsim vit.onnx vit_opt.onnx 22 | -------------------------------------------------------------------------------- /models/cv/classification/vit/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | transformers==4.37.1 5 | -------------------------------------------------------------------------------- /models/cv/classification/vit/igie/test.py: -------------------------------------------------------------------------------- 1 | from transformers import ViTImageProcessor, ViTForImageClassification 2 | model = ViTForImageClassification.from_pretrained('google/vit-base-patch16-224') 3 | 4 | processor = ViTImageProcessor.from_pretrained('google/vit-base-patch16-224') 5 | model = ViTForImageClassification.from_pretrained('google/vit-base-patch16-224') 6 | 7 | print(model) 8 | -------------------------------------------------------------------------------- /models/cv/classification/wide_resnet101/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name wide_resnet101_2 --weight wide_resnet101_2-32ee1156.pth --output wide_resnet101.onnx -------------------------------------------------------------------------------- /models/cv/classification/wide_resnet50/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r ../../igie_common/requirements.txt 20 | python3 ../../igie_common/export.py --model-name wide_resnet50_2 --weight wide_resnet50_2-95faca4d.pth --output wide_resnet50.onnx -------------------------------------------------------------------------------- /models/cv/face_recognition/facenet/ixrt/config/FACENET_CONFIG: -------------------------------------------------------------------------------- 1 | # IMGSIZE : 模型输入hw大小 2 | # MODEL_NAME : 生成onnx/engine的basename 3 | # ORIGINE_MODEL : 原始onnx文件名称 4 | IMGSIZE=160 5 | MODEL_NAME=facenet 6 | ORIGINE_MODEL=facenet_export.onnx 7 | 8 | # QUANT CONFIG (仅PRECISION为int8时生效) 9 | # QUANT_OBSERVER : 量化策略，可选 [hist_percentile, percentile, minmax, entropy, ema] 10 | # QUANT_BATCHSIZE : 量化时组dataloader的batchsize, 最好和onnx中的batchsize保持一致，有些op可能推导shape错误(比如Reshape) 11 | # QUANT_STEP : 量化步数 12 | # QUANT_SEED : 随机种子 保证量化结果可复现 13 | # QUANT_EXIST_ONNX : 如果有其他来源的量化模型则填写 14 | QUANT_OBSERVER=hist_percentile 15 | QUANT_BATCHSIZE=64 16 | QUANT_STEP=32 17 | QUANT_SEED=42 18 | DISABLE_QUANT_LIST= 19 | QUANT_EXIST_ONNX= 20 | -------------------------------------------------------------------------------- /models/cv/face_recognition/facenet/ixrt/load_ixrt_plugin.py: -------------------------------------------------------------------------------- 1 | import ctypes 2 | import tensorrt 3 | from os.path import join, dirname, exists 4 | def load_ixrt_plugin(logger=tensorrt.Logger(tensorrt.Logger.INFO), namespace="", dynamic_path=""): 5 | if not dynamic_path: 6 | dynamic_path = join(dirname(tensorrt.__file__), "lib", "libixrt_plugin.so") 7 | if not exists(dynamic_path): 8 | raise FileNotFoundError( 9 | f"The ixrt_plugin lib {dynamic_path} is not existed, please provided effective plugin path!") 10 | ctypes.CDLL(dynamic_path) 11 | tensorrt.init_libnvinfer_plugins(logger, namespace) 12 | print(f"Loaded plugin from {dynamic_path}") -------------------------------------------------------------------------------- /models/cv/face_recognition/facenet/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | tabulate 3 | scipy==1.8.0 4 | scikit-learn 5 | onnx==1.18.0 6 | onnxsim 7 | simplejson 8 | numpy==1.23.5 9 | opencv-python==4.6.0.66 -------------------------------------------------------------------------------- /models/cv/instance_segmentation/mask_rcnn/ixrt/plugins/README.md: -------------------------------------------------------------------------------- 1 | ### MaskRCNN plugins 2 | 3 | modify from https://github.com/wang-xinyu/tensorrtx/tree/master/rcnn 4 | -------------------------------------------------------------------------------- /models/cv/instance_segmentation/mask_rcnn/ixrt/plugins/cmake/FindCompiler.cmake: -------------------------------------------------------------------------------- 1 | if(NOT COMPILER_PATH) 2 | if (EXISTS /opt/sw_home/local/bin/clang++) 3 | set(COMPILER_PATH /opt/sw_home/local/bin) 4 | elseif (EXISTS /usr/local/bin/clang++) 5 | set(COMPILER_PATH /opt/sw_home/local/bin) 6 | else() 7 | message(STATUS "COMPILER_PATH is not set and we couldn't find clang compiler neither, will use system C/C++ compiler") 8 | endif() 9 | endif() 10 | if (COMPILER_PATH) 11 | set(CMAKE_CXX_COMPILER ${COMPILER_PATH}/clang++) 12 | set(CMAKE_C_COMPILER ${COMPILER_PATH}/clang) 13 | endif() 14 | message(STATUS "Use ${CMAKE_CXX_COMPILER} and ${CMAKE_C_COMPILER} as C++ and C compiler") 15 | -------------------------------------------------------------------------------- /models/cv/instance_segmentation/mask_rcnn/ixrt/plugins/cmake/FindIxrt.cmake: -------------------------------------------------------------------------------- 1 | # This cmake file decides how to build with IxRT 2 | # Custom IxRT Path 3 | if(NOT "${IXRT_HOME}" STREQUAL "") 4 | set(IXRT_INCLUDE_DIR ${IXRT_HOME}/include) 5 | set(IXRT_LIB_DIR ${IXRT_HOME}/lib) 6 | # From default paths 7 | else() 8 | set(IXRT_INCLUDE_DIR /usr/local/corex/include) 9 | set(IXRT_LIB_DIR /usr/local/corex/lib) 10 | endif() 11 | 12 | message(STATUS "IXRT_INCLUDE_DIR: ${IXRT_INCLUDE_DIR}") 13 | message(STATUS "IXRT_LIB_DIR: ${IXRT_LIB_DIR}") 14 | 15 | if(EXISTS ${IXRT_INCLUDE_DIR} AND EXISTS ${IXRT_LIB_DIR}) 16 | include_directories(${IXRT_INCLUDE_DIR}) 17 | else() 18 | message( FATAL_ERROR "IxRT library doesn't exist!") 19 | endif() 20 | -------------------------------------------------------------------------------- /models/cv/instance_segmentation/mask_rcnn/ixrt/plugins/cmake/SetFlags.cmake: -------------------------------------------------------------------------------- 1 | option(USE_TRT "Use TensorRT for API comparison" OFF) 2 | set(CMAKE_CXX_STANDARD 17) 3 | set(CMAKE_CXX_STANDARD_REQUIRED ON) 4 | set(CMAKE_CXX_EXTENSIONS OFF) 5 | set(CMAKE_EXPORT_COMPILE_COMMANDS ON) 6 | if (USE_TRT) 7 | set(LIBRT nvinfer) 8 | set(LIBPLUGIN nvinfer_plugin) 9 | set(LIBPARSER nvonnxparser) 10 | string(APPEND CMAKE_CXX_FLAGS " -DUSE_TRT") 11 | else() 12 | set(LIBRT ixrt) 13 | set(LIBPLUGIN ixrt_plugin) 14 | set(LIBPARSER "") 15 | endif () 16 | -------------------------------------------------------------------------------- /models/cv/instance_segmentation/mask_rcnn/ixrt/plugins/macros.h: -------------------------------------------------------------------------------- 1 | #pragma once 2 | 3 | #include 4 | #include 5 | 6 | #define CUDA_11 7 | #define TRT_NOEXCEPT noexcept 8 | #define TRT_CONST_ENQUEUE const 9 | -------------------------------------------------------------------------------- /models/cv/instance_segmentation/mask_rcnn/ixrt/samples/demo.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/models/cv/instance_segmentation/mask_rcnn/ixrt/samples/demo.jpg -------------------------------------------------------------------------------- /models/cv/instance_segmentation/mask_rcnn/ixrt/scripts/infer_mask_rcnn_fp16_performance.sh: -------------------------------------------------------------------------------- 1 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 2 | # All Rights Reserved. 3 | # 4 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 5 | # not use this file except in compliance with the License. You may obtain 6 | # a copy of the License at 7 | # 8 | # http://www.apache.org/licenses/LICENSE-2.0 9 | # 10 | # Unless required by applicable law or agreed to in writing, software 11 | # distributed under the License is distributed on an "AS IS" BASIS, WITHOUT 12 | # WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the 13 | # License for the specific language governing permissions and limitations 14 | # under the License. 15 | 16 | cd ./python && python3 maskrcnn.py perf --engine_file ./maskrcnn.engine --bsz 1 \ 17 | --fps_target 7.3 18 | -------------------------------------------------------------------------------- /models/cv/instance_segmentation/solov1/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx==1.18.0 3 | onnxsim 4 | tabulate 5 | mmdet==2.28.2 6 | addict 7 | yapf 8 | pycuda -------------------------------------------------------------------------------- /models/cv/multi_object_tracking/deepsort/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx==1.18.0 2 | tqdm 3 | onnxsim 4 | -------------------------------------------------------------------------------- /models/cv/multi_object_tracking/deepsort/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx==1.18.0 2 | tqdm 3 | onnxsim 4 | termcolor -------------------------------------------------------------------------------- /models/cv/multi_object_tracking/fastreid/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | onnxoptimizer -------------------------------------------------------------------------------- /models/cv/multi_object_tracking/fastreid/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | onnxoptimizer 5 | termcolor -------------------------------------------------------------------------------- /models/cv/multi_object_tracking/repnet/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r requirements.txt 20 | 21 | python3 export.py --weight epoch_14.pth --output repnet.onnx 22 | 23 | # Use onnxsim optimize onnx model 24 | onnxsim repnet.onnx repnet_opt.onnx -------------------------------------------------------------------------------- /models/cv/multi_object_tracking/repnet/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx==1.18.0 2 | tqdm 3 | onnxsim 4 | -------------------------------------------------------------------------------- /models/cv/object_detection/atss/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | mmdet==3.3.0 5 | mmdeploy==1.3.1 6 | mmengine==0.10.4 7 | -------------------------------------------------------------------------------- /models/cv/object_detection/atss/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | mmdet==3.3.0 5 | mmdeploy==1.3.1 6 | mmengine==0.10.4 7 | -------------------------------------------------------------------------------- /models/cv/object_detection/centernet/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | mmdet 4 | mmdeploy==1.3.1 5 | mmengine 6 | -------------------------------------------------------------------------------- /models/cv/object_detection/centernet/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | mmdet 4 | mmdeploy==1.3.1 5 | mmengine 6 | transformers==4.37.1 -------------------------------------------------------------------------------- /models/cv/object_detection/detr/ixrt/datasets/__init__.py: -------------------------------------------------------------------------------- 1 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 2 | # All Rights Reserved. 3 | # 4 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 5 | # not use this file except in compliance with the License. You may obtain 6 | # a copy of the License at 7 | # 8 | # http://www.apache.org/licenses/LICENSE-2.0 9 | # 10 | # Unless required by applicable law or agreed to in writing, software 11 | # distributed under the License is distributed on an "AS IS" BASIS, WITHOUT 12 | # WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the 13 | # License for the specific language governing permissions and limitations 14 | # under the License. -------------------------------------------------------------------------------- /models/cv/object_detection/detr/ixrt/refine_utils/__init__.py: -------------------------------------------------------------------------------- 1 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 2 | # All Rights Reserved. 3 | # 4 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 5 | # not use this file except in compliance with the License. You may obtain 6 | # a copy of the License at 7 | # 8 | # http://www.apache.org/licenses/LICENSE-2.0 9 | # 10 | # Unless required by applicable law or agreed to in writing, software 11 | # distributed under the License is distributed on an "AS IS" BASIS, WITHOUT 12 | # WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the 13 | # License for the specific language governing permissions and limitations 14 | # under the License. -------------------------------------------------------------------------------- /models/cv/object_detection/detr/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | pycuda 3 | onnx 4 | onnxsim 5 | tabulate 6 | pycocotools 7 | opencv-python==4.6.0.66 -------------------------------------------------------------------------------- /models/cv/object_detection/fcos/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | mmdet==3.3.0 5 | mmdeploy==1.3.1 6 | mmengine==0.10.4 7 | -------------------------------------------------------------------------------- /models/cv/object_detection/fcos/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | yapf==0.40.2 2 | addict==2.4.0 3 | mmdet==3.3.0 4 | tqdm 5 | onnx 6 | onnxsim 7 | pycocotools 8 | opencv-python==4.6.0.66 9 | mmengine -------------------------------------------------------------------------------- /models/cv/object_detection/foveabox/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | mmdet 5 | mmdeploy==1.3.1 6 | mmengine 7 | -------------------------------------------------------------------------------- /models/cv/object_detection/foveabox/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx 3 | onnxsim 4 | ultralytics 5 | pycocotools 6 | mmdeploy==1.3.1 7 | mmdet 8 | opencv-python==4.6.0.66 -------------------------------------------------------------------------------- /models/cv/object_detection/fsaf/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | mmdet==3.3.0 5 | mmdeploy==1.3.1 6 | mmengine==0.10.4 7 | -------------------------------------------------------------------------------- /models/cv/object_detection/fsaf/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | mmdet==3.3.0 5 | mmdeploy==1.3.1 6 | mmengine==0.10.4 7 | -------------------------------------------------------------------------------- /models/cv/object_detection/gfl/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | mmdet==3.3.0 5 | mmdeploy==1.3.1 6 | mmengine==0.10.4 7 | -------------------------------------------------------------------------------- /models/cv/object_detection/hrnet/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | mmdet 5 | mmdeploy==1.3.1 6 | mmengine 7 | -------------------------------------------------------------------------------- /models/cv/object_detection/hrnet/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | mmdet 5 | mmdeploy==1.3.1 6 | mmengine 7 | -------------------------------------------------------------------------------- /models/cv/object_detection/ixrt_common/cut_model.py: -------------------------------------------------------------------------------- 1 | import onnx 2 | import argparse 3 | from onnxsim import simplify 4 | 5 | def parse_args(): 6 | parser = argparse.ArgumentParser() 7 | parser.add_argument("--input_model", type=str) 8 | parser.add_argument("--output_model", type=str) 9 | parser.add_argument("--input_names", nargs='+', type=str) 10 | parser.add_argument("--output_names", nargs='+', type=str) 11 | args = parser.parse_args() 12 | return args 13 | 14 | args = parse_args() 15 | onnx.utils.extract_model(args.input_model, args.output_model, args.input_names, args.output_names) 16 | print(" Cut Model Done.") -------------------------------------------------------------------------------- /models/cv/object_detection/ixrt_common/datasets/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /models/cv/object_detection/ixrt_common/load_ixrt_plugin.py: -------------------------------------------------------------------------------- 1 | import ctypes 2 | import tensorrt 3 | from os.path import join, dirname, exists 4 | def load_ixrt_plugin(logger=tensorrt.Logger(tensorrt.Logger.INFO), namespace="", dynamic_path=""): 5 | if not dynamic_path: 6 | dynamic_path = join(dirname(tensorrt.__file__), "lib", "libixrt_plugin.so") 7 | if not exists(dynamic_path): 8 | raise FileNotFoundError( 9 | f"The ixrt_plugin lib {dynamic_path} is not existed, please provided effective plugin path!") 10 | ctypes.CDLL(dynamic_path) 11 | tensorrt.init_libnvinfer_plugins(logger, namespace) 12 | print(f"Loaded plugin from {dynamic_path}") -------------------------------------------------------------------------------- /models/cv/object_detection/ixrt_common/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx 3 | onnxsim 4 | ultralytics 5 | pycocotools 6 | opencv-python==4.6.0.66 7 | pycuda 8 | seaborn -------------------------------------------------------------------------------- /models/cv/object_detection/ixrt_common/simplify_model.py: -------------------------------------------------------------------------------- 1 | import onnx 2 | import argparse 3 | from onnxsim import simplify 4 | 5 | # Simplify 6 | def simplify_model(args): 7 | onnx_model = onnx.load(args.origin_model) 8 | model_simp, check = simplify(onnx_model) 9 | model_simp = onnx.shape_inference.infer_shapes(model_simp) 10 | onnx.save(model_simp, args.output_model) 11 | print(" Simplify onnx Done.") 12 | 13 | def parse_args(): 14 | parser = argparse.ArgumentParser() 15 | parser.add_argument("--origin_model", type=str) 16 | parser.add_argument("--output_model", type=str) 17 | args = parser.parse_args() 18 | return args 19 | 20 | args = parse_args() 21 | simplify_model(args) -------------------------------------------------------------------------------- /models/cv/object_detection/paa/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | mmdet 5 | mmdeploy==1.3.1 6 | mmengine 7 | -------------------------------------------------------------------------------- /models/cv/object_detection/paa/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | mmdet 5 | mmdeploy==1.3.1 6 | mmengine 7 | -------------------------------------------------------------------------------- /models/cv/object_detection/retinaface/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | opencv-python==4.6.0.66 5 | -------------------------------------------------------------------------------- /models/cv/object_detection/retinaface/igie/widerface_evaluate/README.md: -------------------------------------------------------------------------------- 1 | # WiderFace-Evaluation 2 | Python Evaluation Code for [Wider Face Dataset](http://mmlab.ie.cuhk.edu.hk/projects/WIDERFace/) 3 | 4 | 5 | ## Usage 6 | 7 | 8 | ##### before evaluating .... 9 | 10 | ```` 11 | python3 setup.py build_ext --inplace 12 | ```` 13 | 14 | ##### evaluating 15 | 16 | **GroungTruth:** `wider_face_val.mat`, `wider_easy_val.mat`, `wider_medium_val.mat`,`wider_hard_val.mat` 17 | 18 | ```` 19 | python3 evaluation.py -p -g 20 | ```` 21 | 22 | ## Bugs & Problems 23 | please issue 24 | 25 | ## Acknowledgements 26 | 27 | some code borrowed from Sergey Karayev 28 | -------------------------------------------------------------------------------- /models/cv/object_detection/retinaface/igie/widerface_evaluate/ground_truth/wider_easy_val.mat: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/models/cv/object_detection/retinaface/igie/widerface_evaluate/ground_truth/wider_easy_val.mat -------------------------------------------------------------------------------- /models/cv/object_detection/retinaface/igie/widerface_evaluate/ground_truth/wider_face_val.mat: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/models/cv/object_detection/retinaface/igie/widerface_evaluate/ground_truth/wider_face_val.mat -------------------------------------------------------------------------------- /models/cv/object_detection/retinaface/igie/widerface_evaluate/ground_truth/wider_hard_val.mat: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/models/cv/object_detection/retinaface/igie/widerface_evaluate/ground_truth/wider_hard_val.mat -------------------------------------------------------------------------------- /models/cv/object_detection/retinaface/igie/widerface_evaluate/ground_truth/wider_medium_val.mat: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/models/cv/object_detection/retinaface/igie/widerface_evaluate/ground_truth/wider_medium_val.mat -------------------------------------------------------------------------------- /models/cv/object_detection/retinaface/igie/widerface_evaluate/setup.py: -------------------------------------------------------------------------------- 1 | """ 2 | WiderFace evaluation code 3 | author: wondervictor 4 | mail: tianhengcheng@gmail.com 5 | copyright@wondervictor 6 | """ 7 | 8 | from distutils.core import setup, Extension 9 | from Cython.Build import cythonize 10 | import numpy 11 | 12 | package = Extension('bbox', ['box_overlaps.pyx'], include_dirs=[numpy.get_include()]) 13 | setup(ext_modules=cythonize([package])) 14 | -------------------------------------------------------------------------------- /models/cv/object_detection/retinaface/ixrt/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /models/cv/object_detection/retinaface/ixrt/models/config.py: -------------------------------------------------------------------------------- 1 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 2 | # All Rights Reserved. 3 | # 4 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 5 | # not use this file except in compliance with the License. You may obtain 6 | # a copy of the License at 7 | # 8 | # http://www.apache.org/licenses/LICENSE-2.0 9 | # 10 | # Unless required by applicable law or agreed to in writing, software 11 | # distributed under the License is distributed on an "AS IS" BASIS, WITHOUT 12 | # WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the 13 | # License for the specific language governing permissions and limitations 14 | # under the License. 15 | 16 | cfg_mnet = { 17 | 'min_sizes': [[10, 20], [32, 64], [128, 256]], 18 | 'steps': [8, 16, 32], 19 | 'clip': False, 20 | 'variance': [0.1, 0.2] 21 | } -------------------------------------------------------------------------------- /models/cv/object_detection/retinaface/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | tabulate 3 | cuda-python 4 | onnx 5 | onnxsim 6 | opencv-python==4.6.0.66 -------------------------------------------------------------------------------- /models/cv/object_detection/retinaface/ixrt/setup.py: -------------------------------------------------------------------------------- 1 | """ 2 | WiderFace evaluation code 3 | author: wondervictor 4 | mail: tianhengcheng@gmail.com 5 | copyright@wondervictor 6 | """ 7 | 8 | from distutils.core import setup, Extension 9 | from Cython.Build import cythonize 10 | import numpy 11 | 12 | package = Extension('bbox', ['box_overlaps.pyx'], include_dirs=[numpy.get_include()]) 13 | setup(ext_modules=cythonize([package])) -------------------------------------------------------------------------------- /models/cv/object_detection/retinaface/ixrt/utils/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /models/cv/object_detection/retinanet/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | mmdet 5 | mmdeploy==1.3.1 6 | mmengine 7 | -------------------------------------------------------------------------------- /models/cv/object_detection/retinanet/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | mmdet 5 | mmdeploy==1.3.1 6 | mmengine 7 | -------------------------------------------------------------------------------- /models/cv/object_detection/rtmdet/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | mmdet==3.3.0 5 | mmdeploy==1.3.1 6 | mmengine==0.10.4 7 | -------------------------------------------------------------------------------- /models/cv/object_detection/sabl/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | mmdet==3.3.0 5 | mmdeploy==1.3.1 6 | mmengine==0.10.4 7 | -------------------------------------------------------------------------------- /models/cv/object_detection/ssd/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | mmdet==3.3.0 5 | mmdeploy==1.3.1 6 | mmengine==0.10.4 7 | -------------------------------------------------------------------------------- /models/cv/object_detection/yolof/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | mmdet==3.3.0 5 | mmdeploy==1.3.1 6 | mmengine==0.10.4 7 | -------------------------------------------------------------------------------- /models/cv/object_detection/yolof/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | mmdet==3.3.0 5 | mmdeploy==1.3.1 6 | mmengine 7 | shapel 8 | aenum -------------------------------------------------------------------------------- /models/cv/object_detection/yolov10/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | huggingface_hub==0.25.2 3 | onnx==1.15.0 4 | matplotlib 5 | pycocotools -------------------------------------------------------------------------------- /models/cv/object_detection/yolov10/ixrt/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r requirements.txt 20 | 21 | mkdir checkpoints 22 | mv yolov10s.pt yolov10.pt 23 | python3 export.py --weight yolov10.pt --batch 32 24 | mv yolov10.onnx checkpoints/ 25 | -------------------------------------------------------------------------------- /models/cv/object_detection/yolov10/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx 3 | onnxsim 4 | pycocotools 5 | opencv-python==4.8.0.74 6 | ultralytics==8.2.51 7 | cuda-python -------------------------------------------------------------------------------- /models/cv/object_detection/yolov11/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx==1.16.0 3 | ultralytics==8.3.59 4 | pycocotools -------------------------------------------------------------------------------- /models/cv/object_detection/yolov11/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx==1.18.0 3 | onnxsim==0.4.36 4 | pycocotools 5 | opencv-python==4.8.0.74 6 | ultralytics==8.3.59 7 | cuda-python -------------------------------------------------------------------------------- /models/cv/object_detection/yolov12/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx==1.16.0 3 | huggingface_hub 4 | pycocotools -------------------------------------------------------------------------------- /models/cv/object_detection/yolov12/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx==1.18.0 3 | huggingface_hub -------------------------------------------------------------------------------- /models/cv/object_detection/yolov13/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx==1.16.0 3 | huggingface_hub 4 | ultralytics 5 | pycocotools -------------------------------------------------------------------------------- /models/cv/object_detection/yolov13/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx==1.18.0 3 | huggingface_hub 4 | ultralytics -------------------------------------------------------------------------------- /models/cv/object_detection/yolov3/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx 3 | onnxsim 4 | ultralytics==8.3.97 5 | pycocotools 6 | -------------------------------------------------------------------------------- /models/cv/object_detection/yolov4/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx 3 | onnxsim 4 | pycocotools 5 | -------------------------------------------------------------------------------- /models/cv/object_detection/yolov4/ixrt/cut_model.py: -------------------------------------------------------------------------------- 1 | import onnx 2 | import argparse 3 | from onnxsim import simplify 4 | 5 | def parse_args(): 6 | parser = argparse.ArgumentParser() 7 | parser.add_argument("--input_model", type=str) 8 | parser.add_argument("--output_model", type=str) 9 | parser.add_argument("--input_names", nargs='+', type=str) 10 | parser.add_argument("--output_names", nargs='+', type=str) 11 | args = parser.parse_args() 12 | return args 13 | 14 | args = parse_args() 15 | onnx.utils.extract_model(args.input_model, args.output_model, args.input_names, args.output_names) 16 | print(" Cut Model Done.") -------------------------------------------------------------------------------- /models/cv/object_detection/yolov4/ixrt/load_ixrt_plugin.py: -------------------------------------------------------------------------------- 1 | import ctypes 2 | import tensorrt 3 | from os.path import join, dirname, exists 4 | def load_ixrt_plugin(logger=tensorrt.Logger(tensorrt.Logger.INFO), namespace="", dynamic_path=""): 5 | if not dynamic_path: 6 | dynamic_path = join(dirname(tensorrt.__file__), "lib", "libixrt_plugin.so") 7 | if not exists(dynamic_path): 8 | raise FileNotFoundError( 9 | f"The ixrt_plugin lib {dynamic_path} is not existed, please provided effective plugin path!") 10 | ctypes.CDLL(dynamic_path) 11 | tensorrt.init_libnvinfer_plugins(logger, namespace) 12 | print(f"Loaded plugin from {dynamic_path}") -------------------------------------------------------------------------------- /models/cv/object_detection/yolov4/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx 3 | onnxsim 4 | pycocotools 5 | pycuda -------------------------------------------------------------------------------- /models/cv/object_detection/yolov5/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r requirements.txt 20 | 21 | python3 export.py --weight yolov5m.pt --output yolov5m.onnx 22 | 23 | # Use onnxsim optimize onnx model 24 | onnxsim yolov5m.onnx yolov5m_opt.onnx 25 | -------------------------------------------------------------------------------- /models/cv/object_detection/yolov5/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx 3 | onnxsim 4 | ultralytics==8.3.97 5 | pycocotools 6 | -------------------------------------------------------------------------------- /models/cv/object_detection/yolov5s/igie/cut_model.py: -------------------------------------------------------------------------------- 1 | import onnx 2 | import argparse 3 | from onnxsim import simplify 4 | 5 | def parse_args(): 6 | parser = argparse.ArgumentParser() 7 | parser.add_argument("--input_model", type=str) 8 | parser.add_argument("--output_model", type=str) 9 | parser.add_argument("--input_names", nargs='+', type=str) 10 | parser.add_argument("--output_names", nargs='+', type=str) 11 | args = parser.parse_args() 12 | return args 13 | 14 | args = parse_args() 15 | onnx.utils.extract_model(args.input_model, args.output_model, args.input_names, args.output_names) 16 | print(" Cut Model Done.") -------------------------------------------------------------------------------- /models/cv/object_detection/yolov5s/igie/datasets/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /models/cv/object_detection/yolov5s/igie/load_ixrt_plugin.py: -------------------------------------------------------------------------------- 1 | import ctypes 2 | import tensorrt 3 | from os.path import join, dirname, exists 4 | def load_ixrt_plugin(logger=tensorrt.Logger(tensorrt.Logger.INFO), namespace="", dynamic_path=""): 5 | if not dynamic_path: 6 | dynamic_path = join(dirname(tensorrt.__file__), "lib", "libixrt_plugin.so") 7 | if not exists(dynamic_path): 8 | raise FileNotFoundError( 9 | f"The ixrt_plugin lib {dynamic_path} is not existed, please provided effective plugin path!") 10 | ctypes.CDLL(dynamic_path) 11 | tensorrt.init_libnvinfer_plugins(logger, namespace) 12 | print(f"Loaded plugin from {dynamic_path}") -------------------------------------------------------------------------------- /models/cv/object_detection/yolov5s/igie/simplify_model.py: -------------------------------------------------------------------------------- 1 | import onnx 2 | import argparse 3 | from onnxsim import simplify 4 | 5 | # Simplify 6 | def simplify_model(args): 7 | onnx_model = onnx.load(args.origin_model) 8 | model_simp, check = simplify(onnx_model) 9 | model_simp = onnx.shape_inference.infer_shapes(model_simp) 10 | onnx.save(model_simp, args.output_model) 11 | print(" Simplify onnx Done.") 12 | 13 | def parse_args(): 14 | parser = argparse.ArgumentParser() 15 | parser.add_argument("--origin_model", type=str) 16 | parser.add_argument("--output_model", type=str) 17 | args = parser.parse_args() 18 | return args 19 | 20 | args = parse_args() 21 | simplify_model(args) -------------------------------------------------------------------------------- /models/cv/object_detection/yolov6/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx 3 | onnxsim 4 | pycocotools 5 | requests -------------------------------------------------------------------------------- /models/cv/object_detection/yolov6/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx 3 | onnxsim 4 | pycocotools 5 | pycuda 6 | numpy==1.24.0 7 | requests -------------------------------------------------------------------------------- /models/cv/object_detection/yolov7/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx 3 | onnxsim 4 | pycocotools 5 | seaborn 6 | requests -------------------------------------------------------------------------------- /models/cv/object_detection/yolov8/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx 3 | pycocotools 4 | # FAILed in 8.2.51 5 | ultralytics==8.1.34 6 | -------------------------------------------------------------------------------- /models/cv/object_detection/yolov8/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx 3 | onnxsim 4 | pycocotools 5 | opencv-python==4.8.0.74 6 | ultralytics==8.2.51 7 | cuda-python -------------------------------------------------------------------------------- /models/cv/object_detection/yolov8n/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx 3 | pycocotools 4 | # FAILed in 8.2.51 5 | ultralytics==8.1.34 6 | opencv-python 7 | -------------------------------------------------------------------------------- /models/cv/object_detection/yolov9/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | ultralytics==8.2.51 5 | pycocotools -------------------------------------------------------------------------------- /models/cv/object_detection/yolov9/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx 3 | onnxsim 4 | pycocotools 5 | opencv-python==4.8.0.74 6 | ultralytics==8.2.51 7 | cuda-python -------------------------------------------------------------------------------- /models/cv/object_detection/yolox/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx==1.18.0 3 | onnxsim 4 | pycocotools 5 | ninja==1.11.1.3 -------------------------------------------------------------------------------- /models/cv/object_detection/yolox/ixrt/datasets/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /models/cv/object_detection/yolox/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx 3 | onnxsim 4 | tabulate 5 | pycocotools 6 | ppq 7 | pycuda 8 | protobuf==3.20.0 9 | opencv-python==4.6.0.66 10 | ninja==1.11.1.3 -------------------------------------------------------------------------------- /models/cv/ocr/kie_layoutxlm/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx 3 | onnxsim 4 | Polygon3 5 | paddlenlp==2.8.1 6 | lanms-neo==1.0.2 7 | paddleocr==2.6.0 8 | paddle2onnx==1.3.0 9 | python-bidi 10 | protobuf==3.20.3 11 | aistudio-sdk==0.2.6 12 | numpy==1.26.4 -------------------------------------------------------------------------------- /models/cv/ocr/svtr/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | tqdm 2 | onnx==1.18.0 3 | onnxsim 4 | numpy==1.24.0 5 | Polygon3 6 | paddlenlp==2.8.1 7 | lanms-neo==1.0.2 8 | paddleocr==2.8.1 9 | paddle2onnx==1.3.1 10 | python-bidi 11 | protobuf==3.20.3 -------------------------------------------------------------------------------- /models/cv/pose_estimation/hrnetpose/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx==1.18.0 2 | tqdm 3 | onnxsim 4 | mmdet==3.3.0 5 | mmdeploy==1.3.1 6 | mmengine==0.10.4 7 | -------------------------------------------------------------------------------- /models/cv/pose_estimation/lightweight_openpose/ixrt/datasets/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /models/cv/pose_estimation/lightweight_openpose/ixrt/modules/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /models/cv/pose_estimation/lightweight_openpose/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx==1.18.0 2 | tqdm 3 | onnxsim 4 | simplejson 5 | opencv-python==4.6.0.66 6 | mmcv==1.5.3 7 | pycocotools -------------------------------------------------------------------------------- /models/cv/pose_estimation/rtmpose/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | mmdet==3.3.0 5 | mmdeploy==1.3.1 6 | mmengine==0.10.4 7 | -------------------------------------------------------------------------------- /models/cv/pose_estimation/rtmpose/ixrt/demo/demo.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/models/cv/pose_estimation/rtmpose/ixrt/demo/demo.jpg -------------------------------------------------------------------------------- /models/cv/pose_estimation/rtmpose/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | onnxsim 4 | mmdet==3.3.0 5 | mmdeploy==1.3.1 6 | mmengine==0.10.4 -------------------------------------------------------------------------------- /models/cv/pose_estimation/rtmpose/ixrt/result.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/models/cv/pose_estimation/rtmpose/ixrt/result.jpg -------------------------------------------------------------------------------- /models/cv/semantic_segmentation/unet/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx==1.18.0 2 | tqdm 3 | onnxsim 4 | mmsegmentation==1.2.2 5 | mmengine==0.10.7 6 | mmdeploy==1.3.1 7 | ftfy 8 | regex 9 | prettytable 10 | mmdet==3.3.0 11 | shapely 12 | aenum -------------------------------------------------------------------------------- /models/multimodal/diffusion_model/stable-diffusion-1.5/diffusers/requirements.txt: -------------------------------------------------------------------------------- 1 | contourpy==1.2.1 2 | cycler==0.12.1 3 | et-xmlfile==1.1.0 4 | fonttools==4.53.0 5 | kiwisolver==1.4.5 6 | matplotlib==3.9.0 7 | numpy==1.26.4 8 | opencv-python==4.10.0.82 9 | openpyxl==3.1.4 10 | packaging==24.1 11 | pandas==2.2.2 12 | pillow==10.3.0 13 | pyparsing==3.1.2 14 | python-dateutil==2.9.0.post0 15 | pytz==2024.1 16 | scipy==1.13.1 17 | six==1.16.0 18 | tzdata==2024.1 19 | transformers==4.39.3 20 | accelerate==0.29.0 21 | peft==0.13.2 22 | safetensors -------------------------------------------------------------------------------- /models/multimodal/vision_language_model/e5-v/vllm/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | cp -r ../../vllm_public_assets/ ./ -------------------------------------------------------------------------------- /models/multimodal/vision_language_model/glm-4v/vllm/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | cp -r ../../vllm_public_assets/ ./ 20 | -------------------------------------------------------------------------------- /models/multimodal/vision_language_model/llava_next_video_7b/vllm/video-example-data/sample_demo_1.mp4: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/models/multimodal/vision_language_model/llava_next_video_7b/vllm/video-example-data/sample_demo_1.mp4 -------------------------------------------------------------------------------- /models/multimodal/vision_language_model/minicpm_o/vllm/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | cp -r ../../vllm_public_assets/ ./ 19 | pip install /mnt/deepspark/install/transformers-4.45.2+corex.4.3.0-py3-none-any.whl 20 | -------------------------------------------------------------------------------- /models/multimodal/vision_language_model/nvlm/vllm/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | cp -r ../../vllm_public_assets/ ./ 20 | -------------------------------------------------------------------------------- /models/multimodal/vision_language_model/paligemma/vllm/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | cp -r ../../vllm_public_assets/ ./ 20 | -------------------------------------------------------------------------------- /models/multimodal/vision_language_model/phi3_v/vllm/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | cp -r ../../vllm_public_assets/ ./ 20 | -------------------------------------------------------------------------------- /models/multimodal/vision_language_model/qwen2_5_vl/vllm/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | cp -r ../../vllm_public_assets/ ./ 20 | pip install transformers==4.50.3 21 | -------------------------------------------------------------------------------- /models/multimodal/vision_language_model/qwen2_vl/vllm/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | cp -r ../../vllm_public_assets/ ./ 20 | pip install transformers==4.50.3 21 | -------------------------------------------------------------------------------- /models/multimodal/vision_language_model/qwen_vl/vllm/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | cp -r ../../vllm_public_assets/ ./ 20 | pip install matplotlib 21 | -------------------------------------------------------------------------------- /models/multimodal/vision_language_model/step3/vllm/requirements.txt: -------------------------------------------------------------------------------- 1 | fire 2 | tqdm 3 | loguru 4 | requests 5 | tabulate -------------------------------------------------------------------------------- /models/multimodal/vision_language_model/vllm_public_assets/American_Eskimo_Dog.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/models/multimodal/vision_language_model/vllm_public_assets/American_Eskimo_Dog.jpg -------------------------------------------------------------------------------- /models/multimodal/vision_language_model/vllm_public_assets/cherry_blossom.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/models/multimodal/vision_language_model/vllm_public_assets/cherry_blossom.jpg -------------------------------------------------------------------------------- /models/multimodal/vision_language_model/xlmroberta/vllm/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x -------------------------------------------------------------------------------- /models/nlp/llm/baichuan2-7b/vllm/template_baichuan.jinja: -------------------------------------------------------------------------------- 1 | {{ (messages|selectattr('role', 'equalto', 'system')|list|last).content|trim if (messages|selectattr('role', 'equalto', 'system')|list) else '' }} 2 | 3 | {%- for message in messages -%} 4 | {%- if message['role'] == 'user' -%} 5 | {{- '' + message['content'] -}} 6 | {%- elif message['role'] == 'assistant' -%} 7 | {{- '' + message['content'] -}} 8 | {%- endif -%} 9 | {%- endfor -%} 10 | 11 | {%- if add_generation_prompt and messages[-1]['role'] != 'assistant' -%} 12 | {{- '' -}} 13 | {% endif %} 14 | -------------------------------------------------------------------------------- /models/nlp/llm/ernie-4.5-21b-a3b/fastdeploy/run_demo.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | export PADDLE_XCCL_BACKEND=iluvatar_gpu 18 | export INFERENCE_MSG_QUEUE_ID=232132 19 | export LD_PRELOAD=/usr/local/corex/lib64/libcuda.so.1 20 | export FD_DEBUG=1 21 | 22 | python3 run_demo.py 23 | -------------------------------------------------------------------------------- /models/nlp/llm/ernie-4.5-300b-a47b/fastdeploy/run_demo.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | export PADDLE_XCCL_BACKEND=iluvatar_gpu 18 | export INFERENCE_MSG_QUEUE_ID=232132 19 | export LD_PRELOAD=/usr/local/corex/lib64/libcuda.so.1 20 | export FD_DEBUG=1 21 | 22 | python3 run_demo.py 23 | -------------------------------------------------------------------------------- /models/nlp/llm/internlm3/lmdeploy/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | pip install /mnt/deepspark/data/install/lmdeploy-0.7.2+corex.4.3.0-py3-none-any.whl -------------------------------------------------------------------------------- /models/nlp/llm/llama2-13b/trtllm/scripts/set_environment.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, WITHOUT 13 | # WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the 14 | # License for the specific language governing permissions and limitations 15 | # under the License. 16 | 17 | set -e 18 | 19 | PROJECT_DIR=$1 20 | 21 | pip3 install -r "$PROJECT_DIR/scripts/requirements.txt" 22 | 23 | -------------------------------------------------------------------------------- /models/nlp/llm/llama2-70b/trtllm/scripts/set_environment.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | set -e 3 | 4 | PROJECT_DIR=$1 5 | 6 | pip3 install -r "$PROJECT_DIR/scripts/requirements.txt" 7 | 8 | -------------------------------------------------------------------------------- /models/nlp/llm/llama2-7b/trtllm/scripts/set_environment.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | set -e 3 | 4 | PROJECT_DIR=$1 5 | 6 | pip3 install -r "$PROJECT_DIR/scripts/requirements.txt" 7 | 8 | -------------------------------------------------------------------------------- /models/nlp/llm/llama2-7b/vllm/template_llama.jinja: -------------------------------------------------------------------------------- 1 | {{- '[INST] ' -}} 2 | 3 | {%- for message in messages -%} 4 | {%- if loop.first and message['role'] == 'system' -%} 5 | {{- '<>\n' + message['content'] + '\n<>\n' -}} 6 | {%- elif message['role'] == 'user' and loop.index <= 2 -%} 7 | {{- message['content'] + ' [/INST]' -}} 8 | {%- elif message['role'] == 'user' -%} 9 | {{- '[INST] ' + message['content'] + ' [/INST]' -}} 10 | {%- elif message['role'] == 'assistant' -%} 11 | {{- ' ' + message['content'] + ' ' -}} 12 | {%- endif -%} 13 | {%- endfor -%} 14 | -------------------------------------------------------------------------------- /models/nlp/plm/albert/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | onnxsim 2 | onnx_graphsurgeon 3 | scikit-learn 4 | tqdm 5 | pycuda 6 | onnx 7 | tabulate 8 | pycocotools 9 | opencv-python==4.6.0.66 10 | transformers==4.33.3 11 | typing-extensions==4.12.2 -------------------------------------------------------------------------------- /models/nlp/plm/bert_base_ner/igie/Int8QAT/bert4torch/__init__.py: -------------------------------------------------------------------------------- 1 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 2 | # All Rights Reserved. 3 | # 4 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 5 | # not use this file except in compliance with the License. You may obtain 6 | # a copy of the License at 7 | # 8 | # http://www.apache.org/licenses/LICENSE-2.0 9 | # 10 | # Unless required by applicable law or agreed to in writing, software 11 | # distributed under the License is distributed on an "AS IS" BASIS, WITHOUT 12 | # WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the 13 | # License for the specific language governing permissions and limitations 14 | # under the License. 15 | 16 | __version__ = "0.1.9" 17 | -------------------------------------------------------------------------------- /models/nlp/plm/bert_base_ner/igie/Int8QAT/pytorch_quantization/__init__.py: -------------------------------------------------------------------------------- 1 | # 2 | # Copyright (c) 2022 Bytedance Inc. All rights reserved. 3 | # Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); 6 | # you may not use this file except in compliance with the License. 7 | # You may obtain a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | # 17 | 18 | # we modified pytorch_quantization in TensorRT(https://github.com/NVIDIA/TensorRT) 19 | # of commit 42805f0 20 | 21 | from .version import __version__ 22 | -------------------------------------------------------------------------------- /models/nlp/plm/bert_base_ner/igie/Int8QAT/pytorch_quantization/nn/_functions/__init__.py: -------------------------------------------------------------------------------- 1 | # 2 | # Copyright (c) 2022 Bytedance Inc. All rights reserved. 3 | # Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); 6 | # you may not use this file except in compliance with the License. 7 | # You may obtain a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | # 17 | -------------------------------------------------------------------------------- /models/nlp/plm/bert_base_ner/igie/Int8QAT/pytorch_quantization/nn/modules/__init__.py: -------------------------------------------------------------------------------- 1 | # 2 | # Copyright (c) 2022 Bytedance Inc. All rights reserved. 3 | # Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); 6 | # you may not use this file except in compliance with the License. 7 | # You may obtain a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | # 17 | -------------------------------------------------------------------------------- /models/nlp/plm/bert_base_ner/igie/Int8QAT/pytorch_quantization/optim/__init__.py: -------------------------------------------------------------------------------- 1 | # 2 | # Copyright (c) 2022 Bytedance Inc. All rights reserved. 3 | # Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); 6 | # you may not use this file except in compliance with the License. 7 | # You may obtain a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | # 17 | -------------------------------------------------------------------------------- /models/nlp/plm/bert_base_ner/igie/Int8QAT/pytorch_quantization/utils/__init__.py: -------------------------------------------------------------------------------- 1 | # 2 | # Copyright (c) 2022 Bytedance Inc. All rights reserved. 3 | # Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); 6 | # you may not use this file except in compliance with the License. 7 | # You may obtain a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | # 17 | 18 | 19 | """Main entry of all utils""" 20 | 21 | from .reduce_amax import reduce_amax 22 | -------------------------------------------------------------------------------- /models/nlp/plm/bert_base_ner/igie/Int8QAT/pytorch_quantization/version.py: -------------------------------------------------------------------------------- 1 | # 2 | # Copyright (c) 2022 Bytedance Inc. All rights reserved. 3 | # Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); 6 | # you may not use this file except in compliance with the License. 7 | # You may obtain a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | # 17 | __version__ = "2.1.2" 18 | -------------------------------------------------------------------------------- /models/nlp/plm/bert_base_ner/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | transformers 4 | bert4torch 5 | numpy==1.23.5 6 | h5py -------------------------------------------------------------------------------- /models/nlp/plm/bert_base_squad/igie/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install -r requirements.txt 20 | 21 | python3 export.py --output bert-base-uncased-squad-v1.onnx -------------------------------------------------------------------------------- /models/nlp/plm/bert_base_squad/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx==1.18.0 2 | tqdm 3 | transformers==4.37.1 4 | -------------------------------------------------------------------------------- /models/nlp/plm/bert_base_squad/ixrt/helpers/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /models/nlp/plm/bert_base_squad/ixrt/load_ixrt_plugin.py: -------------------------------------------------------------------------------- 1 | from os.path import join, dirname, exists 2 | import tensorrt as trt 3 | import ctypes 4 | 5 | def load_ixrt_plugin(logger=trt.Logger(trt.Logger.WARNING), namespace="", dynamic_path=""): 6 | if not dynamic_path: 7 | dynamic_path = join(dirname(trt.__file__), "lib", "libixrt_plugin.so") 8 | if not exists(dynamic_path): 9 | raise FileNotFoundError( 10 | f"The ixrt_plugin lib {dynamic_path} is not existed, please provided effective plugin path!") 11 | ctypes.CDLL(dynamic_path, mode=ctypes.RTLD_GLOBAL) 12 | trt.init_libnvinfer_plugins(logger, namespace) 13 | print(f"Loaded plugin from {dynamic_path}") -------------------------------------------------------------------------------- /models/nlp/plm/bert_base_squad/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx==1.18.0 2 | pycuda 3 | six 4 | tqdm -------------------------------------------------------------------------------- /models/nlp/plm/bert_large_squad/igie/Int8QAT/pytorch_quantization/__init__.py: -------------------------------------------------------------------------------- 1 | # 2 | # Copyright (c) 2022 Bytedance Inc. All rights reserved. 3 | # Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); 6 | # you may not use this file except in compliance with the License. 7 | # You may obtain a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | # 17 | 18 | # we modified pytorch_quantization in TensorRT(https://github.com/NVIDIA/TensorRT) 19 | # of commit 42805f0 20 | 21 | from .version import __version__ 22 | -------------------------------------------------------------------------------- /models/nlp/plm/bert_large_squad/igie/Int8QAT/pytorch_quantization/nn/_functions/__init__.py: -------------------------------------------------------------------------------- 1 | # 2 | # Copyright (c) 2022 Bytedance Inc. All rights reserved. 3 | # Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); 6 | # you may not use this file except in compliance with the License. 7 | # You may obtain a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | # 17 | -------------------------------------------------------------------------------- /models/nlp/plm/bert_large_squad/igie/Int8QAT/pytorch_quantization/nn/modules/__init__.py: -------------------------------------------------------------------------------- 1 | # 2 | # Copyright (c) 2022 Bytedance Inc. All rights reserved. 3 | # Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); 6 | # you may not use this file except in compliance with the License. 7 | # You may obtain a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | # 17 | -------------------------------------------------------------------------------- /models/nlp/plm/bert_large_squad/igie/Int8QAT/pytorch_quantization/optim/__init__.py: -------------------------------------------------------------------------------- 1 | # 2 | # Copyright (c) 2022 Bytedance Inc. All rights reserved. 3 | # Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); 6 | # you may not use this file except in compliance with the License. 7 | # You may obtain a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | # 17 | -------------------------------------------------------------------------------- /models/nlp/plm/bert_large_squad/igie/Int8QAT/pytorch_quantization/utils/__init__.py: -------------------------------------------------------------------------------- 1 | # 2 | # Copyright (c) 2022 Bytedance Inc. All rights reserved. 3 | # Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); 6 | # you may not use this file except in compliance with the License. 7 | # You may obtain a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | # 17 | 18 | 19 | """Main entry of all utils""" 20 | 21 | from .reduce_amax import reduce_amax 22 | -------------------------------------------------------------------------------- /models/nlp/plm/bert_large_squad/igie/Int8QAT/pytorch_quantization/version.py: -------------------------------------------------------------------------------- 1 | # 2 | # Copyright (c) 2022 Bytedance Inc. All rights reserved. 3 | # Copyright (c) 2021, NVIDIA CORPORATION. All rights reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); 6 | # you may not use this file except in compliance with the License. 7 | # You may obtain a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | # 17 | __version__ = "2.1.2" 18 | -------------------------------------------------------------------------------- /models/nlp/plm/bert_large_squad/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx 2 | tqdm 3 | numpy==1.23.5 4 | transformers==4.37.1 5 | h5py -------------------------------------------------------------------------------- /models/nlp/plm/bert_large_squad/ixrt/helpers/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /models/nlp/plm/bert_large_squad/ixrt/load_ixrt_plugin.py: -------------------------------------------------------------------------------- 1 | from os.path import join, dirname, exists 2 | import tensorrt as trt 3 | import ctypes 4 | 5 | def load_ixrt_plugin(logger=trt.Logger(trt.Logger.WARNING), namespace="", dynamic_path=""): 6 | if not dynamic_path: 7 | dynamic_path = join(dirname(trt.__file__), "lib", "libixrt_plugin.so") 8 | if not exists(dynamic_path): 9 | raise FileNotFoundError( 10 | f"The ixrt_plugin lib {dynamic_path} is not existed, please provided effective plugin path!") 11 | ctypes.CDLL(dynamic_path, mode=ctypes.RTLD_GLOBAL) 12 | trt.init_libnvinfer_plugins(logger, namespace) 13 | print(f"Loaded plugin from {dynamic_path}") 14 | -------------------------------------------------------------------------------- /models/nlp/plm/deberta/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | onnxsim 2 | onnx_graphsurgeon 3 | scikit-learn 4 | tqdm 5 | pycuda 6 | onnx 7 | tabulate 8 | pycocotools 9 | opencv-python==4.6.0.66 10 | tf2onnx 11 | transformers==4.33.3 12 | typing-extensions==4.12.2 -------------------------------------------------------------------------------- /models/nlp/plm/roberta/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | onnxsim 2 | py-libnuma==1.2 3 | bert 4 | pycuda 5 | transformers==4.33.3 6 | typing-extensions==4.12.2 -------------------------------------------------------------------------------- /models/nlp/plm/roformer/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | tf2onnx 2 | pycuda 3 | onnxsim 4 | py-libnuma==1.2 5 | typing-extensions==4.12.2 -------------------------------------------------------------------------------- /models/nlp/plm/transformer/build_helpers/__init__.py: -------------------------------------------------------------------------------- 1 | # Order of imports is important (see warning otherwise when running tests) 2 | import setuptools # isort:skip # noqa 3 | import distutils # isort:skip # noqa 4 | -------------------------------------------------------------------------------- /models/nlp/plm/transformer/igie/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /models/nlp/plm/transformer/igie/requirements.txt: -------------------------------------------------------------------------------- 1 | numpy==1.26.4 2 | cython 3 | antlr4-python3-runtime==4.9.3 4 | sacrebleu==2.5.1 5 | bitarray 6 | scikit-learn -------------------------------------------------------------------------------- /models/nlp/plm/transformer/igie/transformer_cfg.py: -------------------------------------------------------------------------------- 1 | import json 2 | class TransformerBaseConfig: 3 | def __init__(self, config_path, use_fp16=True): 4 | with open(config_path, "r") as f: 5 | data = json.load(f) 6 | self.num_attention_heads = data["num_attention_heads"] 7 | self.hidden_size = data["hidden_size"] 8 | self.intermediate_size = data["intermediate_size"] 9 | self.num_hidden_layers = data["num_hidden_layers"] 10 | self.head_size = self.hidden_size // self.num_attention_heads 11 | self.tgt_vocab_size = data["tgt_vocab_size"] 12 | self.max_sequence_length = data["max_sequence_length"] 13 | self.sos_token_id = data["sos_token_id"] 14 | self.eos_token_id = data["eos_token_id"] 15 | self.use_fp16 = use_fp16 -------------------------------------------------------------------------------- /models/nlp/plm/transformer/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | numpy==1.26.4 2 | cython 3 | antlr4-python3-runtime==4.9.3 4 | sacrebleu==2.5.1 5 | bitarray 6 | scikit-learn -------------------------------------------------------------------------------- /models/nlp/plm/transformer/plugin/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /models/nlp/plm/transformer/plugin/transformer_cfg.py: -------------------------------------------------------------------------------- 1 | import json 2 | class TransformerBaseConfig: 3 | def __init__(self, config_path, use_fp16=True): 4 | with open(config_path, "r") as f: 5 | data = json.load(f) 6 | self.num_attention_heads = data["num_attention_heads"] 7 | self.hidden_size = data["hidden_size"] 8 | self.intermediate_size = data["intermediate_size"] 9 | self.num_hidden_layers = data["num_hidden_layers"] 10 | self.head_size = self.hidden_size // self.num_attention_heads 11 | self.tgt_vocab_size = data["tgt_vocab_size"] 12 | self.max_sequence_length = data["max_sequence_length"] 13 | self.sos_token_id = data["sos_token_id"] 14 | self.eos_token_id = data["eos_token_id"] 15 | self.use_fp16 = use_fp16 -------------------------------------------------------------------------------- /models/nlp/plm/videobert/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | onnxsim 2 | onnx_graphsurgeon 3 | scikit-learn 4 | tqdm 5 | pycuda 6 | onnx 7 | tabulate 8 | pycocotools 9 | opencv-python==4.6.0.66 10 | transformers==4.33.3 11 | typing-extensions==4.12.2 -------------------------------------------------------------------------------- /models/others/recommendation/wide_and_deep/ixrt/requirements.txt: -------------------------------------------------------------------------------- 1 | tf2onnx 2 | pycuda 3 | onnxsim 4 | py-libnuma==1.2 -------------------------------------------------------------------------------- /models/speech/asr/ultravox/vllm/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | -------------------------------------------------------------------------------- /models/speech/asr/whisper/vllm/ci/prepare.sh: -------------------------------------------------------------------------------- 1 | #!/bin/bash 2 | # Copyright (c) 2025, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 3 | # All Rights Reserved. 4 | # 5 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 6 | # not use this file except in compliance with the License. You may obtain 7 | # a copy of the License at 8 | # 9 | # http://www.apache.org/licenses/LICENSE-2.0 10 | # 11 | # Unless required by applicable law or agreed to in writing, software 12 | # distributed under the License is distributed on an "AS IS" BASIS, 13 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 14 | # See the License for the specific language governing permissions and 15 | # limitations under the License. 16 | 17 | set -x 18 | 19 | pip3 install transformers==4.50.3 20 | pip3 install librosa -------------------------------------------------------------------------------- /tests/README.md: -------------------------------------------------------------------------------- 1 | # Introduction 2 | 3 | This script is just used for ci runner. -------------------------------------------------------------------------------- /tests/utils.py: -------------------------------------------------------------------------------- 1 | # Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. 2 | # All Rights Reserved. 3 | # 4 | # Licensed under the Apache License, Version 2.0 (the "License"); you may 5 | # not use this file except in compliance with the License. You may obtain 6 | # a copy of the License at 7 | # 8 | # http://www.apache.org/licenses/LICENSE-2.0 9 | # 10 | # Unless required by applicable law or agreed to in writing, software 11 | # distributed under the License is distributed on an "AS IS" BASIS, 12 | # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 13 | # See the License for the specific language governing permissions and 14 | # limitations under the License. 15 | 16 | import os 17 | 18 | def is_debug(): 19 | is_debug_flag = os.environ.get("IS_DEBUG") 20 | if is_debug_flag and is_debug_flag.lower()=="true": 21 | return True 22 | else: 23 | return False -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/.gitignore: -------------------------------------------------------------------------------- 1 | __pycache__ 2 | *.pyc 3 | *.prototxt 4 | *.deploy 5 | .vscode/ 6 | *.npy 7 | *.tar 8 | span.log 9 | byte_micro_perf/backends/*/venv/ 10 | byte_micro_perf/reports/ 11 | byte_infer_perf/general_perf/tools/venv/ 12 | byte_infer_perf/general_perf/backends/*/venv/ 13 | byte_infer_perf/general_perf/model_zoo/* 14 | !byte_infer_perf/general_perf/model_zoo/*.json 15 | byte_infer_perf/general_perf/download/*.* 16 | !byte_infer_perf/general_perf/download/README.md 17 | byte_infer_perf/general_perf/datasets/open_imagenet/preprocessed/ 18 | # byte_infer_perf/general_perf/datasets/* 19 | !byte_infer_perf/general_perf/datasets/fake_dataset 20 | !*.py 21 | byte_infer_perf/general_perf/reports/* 22 | !byte_infer_perf/general_perf/_inference/general_perf/reports/README 23 | format_code.sh 24 | init_env.sh 25 | 26 | byte_infer_perf/llm_perf/download 27 | byte_infer_perf/llm_perf/model_zoo/sota 28 | byte_infer_perf/llm_perf/reports -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/NOTICE: -------------------------------------------------------------------------------- 1 | ByteMLPerf 2 | Copyright 2023 ByteDance Ltd. and/or its affiliates. -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/backends/CPU/calculate_cpu_diff.sh: -------------------------------------------------------------------------------- 1 | #！bin/bash 2 | if [ ! -d "general_perf/backends/CPU/venv" ];then 3 | virtualenv -p python3 general_perf/backends/CPU/venv 4 | source general_perf/backends/CPU/venv/bin/activate 5 | general_perf/backends/CPU/venv/bin/python3 -m pip install --upgrade pip -q 6 | general_perf/backends/CPU/venv/bin/python3 -m pip install -r general_perf/backends/CPU/requirements.txt -q 7 | else 8 | source general_perf/backends/CPU/venv/bin/activate 9 | general_perf/backends/CPU/venv/bin/python3 -m pip install -r general_perf/backends/CPU/requirements.txt -q 10 | fi 11 | 12 | python3 general_perf/backends/CPU/calculate_cpu_diff.py --task $1 --batch_size $2 13 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/backends/CPU/requirements.txt: -------------------------------------------------------------------------------- 1 | matplotlib 2 | scikit-learn 3 | opencv-python-headless 4 | transformers 5 | tokenization 6 | bert-tensorflow==1.0.1 7 | torchvision 8 | onnx 9 | numpy==1.19.2 10 | tensorflow==2.4.0 11 | onnxruntime 12 | torch==1.13.1 13 | sentencepiece==0.1.96 14 | pandas==1.3.3 15 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/backends/ILUVATAR/requirements.txt: -------------------------------------------------------------------------------- 1 | onnx==1.15.0 2 | onnxsim==0.4.36 3 | tf2onnx==1.16.1 4 | onnxruntime 5 | onnxoptimizer==0.3.13 6 | bert-tensorflow==1.0.1 7 | 8 | pandas==2.1.1 9 | numpy==1.23.0 10 | matplotlib 11 | scikit-learn 12 | opencv-python==4.6.0.66 13 | opencv-python-headless 14 | tokenization==1.0.7 15 | tokenizers==0.13.3 16 | sentencepiece==0.1.96 17 | typing_extensions==4.10.0 18 | 19 | py-libnuma==1.2 20 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/backends/ILUVATAR/utils/__init__.py: -------------------------------------------------------------------------------- 1 | from .file import load_json, save_json 2 | from .timer import Timer 3 | 4 | 5 | from .argument import get_args 6 | from .import_model import import_model_to_igie 7 | from .target import get_target 8 | 9 | from .dataloader import get_dataloader_from_args, download_builtin_data 10 | 11 | 12 | from .imagenet_metric import get_topk_accuracy 13 | from .coco_metric import COCO2017Evaluator, COCO2017EvaluatorForYolox, COCO2017EvaluatorForYolov4 14 | 15 | from .quantization import igie_quantize_model_from_args, onnx_quantize_model_from_args 16 | 17 | from .mod_rewriter import modify_seq_len_for_nlp 18 | from .stauts_checker import check_status 19 | 20 | from .compile_engine import compile_engine_from_args -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/backends/ILUVATAR/utils/compile_engine.py: -------------------------------------------------------------------------------- 1 | import os 2 | import tvm 3 | 4 | from .import_model import import_model_to_igie 5 | from .target import get_target 6 | 7 | 8 | # a simple wrapper for compile engine and get module 9 | def compile_engine_from_args(args): 10 | target, device = get_target(args.target) 11 | 12 | if not os.path.exists(args.engine_path): 13 | mod, params = import_model_to_igie(args.model_path, args.input_dict, args.model_framework) 14 | lib = tvm.relay.build(mod, target=target, params=params, precision=args.precision, verbose=args.verbose, required_pass=args.required_pass) 15 | lib.export_library(args.engine_path) 16 | else: 17 | lib = tvm.runtime.load_module(args.engine_path) 18 | module = tvm.contrib.graph_executor.GraphModule(lib["default"](device)) 19 | return module -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/backends/ILUVATAR/utils/fastCoCoeval/__init__.py: -------------------------------------------------------------------------------- 1 | # import torch first to make jit op work without `ImportError of libc10.so` 2 | import torch 3 | 4 | from .jit_ops import FastCOCOEvalOp, JitOp 5 | 6 | try: 7 | from .fast_coco_eval_api import COCOeval_opt 8 | except ImportError: # exception will be raised when users build yolox from source 9 | pass 10 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/backends/ILUVATAR/utils/file.py: -------------------------------------------------------------------------------- 1 | import os 2 | import json 3 | 4 | def download_file(src_url, save_path): 5 | if not os.path.exists(save_path): 6 | cmd = f"wget {src_url}" 7 | os.system(cmd) 8 | 9 | assert os.path.exists(save_path) 10 | 11 | 12 | def load_json(path): 13 | with open(path, "r") as f: 14 | data = json.load(f) 15 | return data 16 | 17 | 18 | def save_json(data, path): 19 | with open(path, "w") as f: 20 | json.dump(data, f, indent=4) 21 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/backends/ILUVATAR/utils/imagenet_metric.py: -------------------------------------------------------------------------------- 1 | import torch 2 | import numpy as np 3 | 4 | 5 | def get_topk_accuracy(pred, label): 6 | if isinstance(pred, np.ndarray): 7 | pred = torch.from_numpy(pred) 8 | 9 | if isinstance(label, np.ndarray): 10 | label = torch.from_numpy(label) 11 | 12 | top1_acc = 0 13 | top5_acc = 0 14 | for idx in range(len(label)): 15 | label_value = label[idx] 16 | if label_value == torch.topk(pred[idx].float(), 1).indices.data: 17 | top1_acc += 1 18 | top5_acc += 1 19 | 20 | elif label_value in torch.topk(pred[idx].float(), 5).indices.data: 21 | top5_acc += 1 22 | 23 | return top1_acc, top5_acc -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/backends/ILUVATAR/utils/stauts_checker.py: -------------------------------------------------------------------------------- 1 | def check_status(result_dict, args): 2 | is_valid = True 3 | if args.acc_target is not None: 4 | acc_result = result_dict["acc_result"] 5 | if acc_result < args.acc_target: 6 | print(f"Expected acc_target is {args.acc_target}, got {acc_result}") 7 | is_valid = False 8 | 9 | if args.fps_target is not None: 10 | fps_result = result_dict["fps_result"] 11 | if fps_result < args.fps_target: 12 | print(f"Expected fps_target is {args.fps_target}, got {fps_result}") 13 | is_valid = False 14 | 15 | if is_valid: 16 | print("\n====Test Success!====\n") 17 | else: 18 | print("\n====Test failed!====\n") 19 | exit(1) 20 | 21 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/backends/ILUVATAR/utils/target.py: -------------------------------------------------------------------------------- 1 | import tvm 2 | 3 | def get_target(target_name): 4 | 5 | target = None 6 | if target_name == "llvm": 7 | target = tvm.target.Target(target_name) 8 | 9 | elif target_name == "iluvatar": 10 | target = tvm.target.iluvatar(model="MR") 11 | 12 | elif target_name == "iluvatar_with_cudnn_cublas": 13 | target = tvm.target.iluvatar(model="MR", options="-libs=cudnn,cublas") 14 | elif target_name == "iluvatar_with_ixinfer": 15 | target = tvm.target.iluvatar(model="MR", options="-libs=ixinfer") 16 | elif target_name == "iluvatar_with_all_libs": 17 | target = tvm.target.iluvatar(model="MR", options="-libs=cudnn,cublas,ixinfer") 18 | 19 | else: 20 | raise Exception(f"Unsupport Target name: {target_name}!") 21 | 22 | device = tvm.device(target.kind.name, 0) 23 | 24 | return target, device 25 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/core/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/core/configs/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/model_zoo/albert-torch-fp32.json: -------------------------------------------------------------------------------- 1 | { 2 | "model": "albert-torch-fp32", 3 | "model_path": "general_perf/model_zoo/popular/open_albert/albert-base-squad.pt", 4 | "framework": "Pytorch", 5 | "framework_version": "1.9.1", 6 | "model_format": "pt", 7 | "model_precision": "FP32", 8 | "inputs":"input_ids.1,attention_mask.1,token_type_ids.1", 9 | "outputs":"start_logits,end_logits", 10 | "input_shape": {"input_ids.1": [1,384], "attention_mask.1": [1,384], "token_type_ids.1": [1,384]}, 11 | "input_type": "LONG,LONG,LONG", 12 | "dataset_name": "open_squad", 13 | "max_batch_size": 64, 14 | "is_quantized": false 15 | } -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/model_zoo/bert-tf-fp32.json: -------------------------------------------------------------------------------- 1 | { 2 | "model": "bert-tf-fp32", 3 | "model_path": "general_perf/model_zoo/regular/open_bert/bert_fp32", 4 | "framework": "Tensorflow", 5 | "framework_version": "2.4.0", 6 | "model_format": "saved_model", 7 | "model_precision": "FP32", 8 | "inputs":"input_ids:0,input_mask:0,segment_ids:0", 9 | "outputs":"logits:0", 10 | "input_shape": {"input_ids:0": [1,384], "input_mask:0": [1,384], "segment_ids:0": [1,384]}, 11 | "input_type": "INT32,INT32,INT32", 12 | "dataset_name": "open_squad", 13 | "max_batch_size": 64, 14 | "is_quantized": false 15 | } -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/model_zoo/bert-torch-fp32.json: -------------------------------------------------------------------------------- 1 | { 2 | "model": "bert-torch-fp32", 3 | "model_path": "general_perf/model_zoo/regular/open_bert/bert-base-uncased-squad-v1.pt", 4 | "framework": "Pytorch", 5 | "framework_version": "1.9.1", 6 | "model_format": "pt", 7 | "model_precision": "FP32", 8 | "inputs":"input_ids.1,attention_mask.1,token_type_ids.1", 9 | "outputs":"start_logits,end_logits", 10 | "input_shape": {"input_ids.1": [1,384], "attention_mask.1": [1,384], "token_type_ids.1": [1,384]}, 11 | "input_type": "LONG,LONG,LONG", 12 | "dataset_name": "open_squad", 13 | "max_batch_size": 64, 14 | "is_quantized": false 15 | } 16 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/model_zoo/chatglm2-6b-torch-fp16.json: -------------------------------------------------------------------------------- 1 | { 2 | "model": "chatglm2-6b-torch-fp16", 3 | "model_path": "general_perf/model_zoo/sota/chatglm2-6b", 4 | "framework": "Pytorch", 5 | "framework_version": "1.13.1", 6 | "model_format": "bin", 7 | "model_precision": "FP16", 8 | "inputs": null, 9 | "outputs": null, 10 | "input_shape": null, 11 | "input_type": null, 12 | "dataset_name": null, 13 | "max_batch_size": null, 14 | "is_quantized": false 15 | } -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/model_zoo/clip-onnx-fp32.json: -------------------------------------------------------------------------------- 1 | { 2 | "model": "clip-onnx-fp32", 3 | "model_path": "general_perf/model_zoo/sota/stable_diffusion/clip-text-encoder.onnx", 4 | "framework": "Onnx", 5 | "framework_version": "1.12.0", 6 | "model_format": "onnx", 7 | "model_precision": "FP32", 8 | "inputs":"input_ids", 9 | "outputs":"last_hidden_state,pooler_output", 10 | "input_shape": {"input_ids": [1,77]}, 11 | "input_type": "INT32", 12 | "dataset_name": null, 13 | "max_batch_size": 64, 14 | "is_quantized": false 15 | } -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/model_zoo/conformer-encoder-onnx-fp32.json: -------------------------------------------------------------------------------- 1 | { 2 | "model": "conformer-encoder-onnx-fp32", 3 | "model_path": "general_perf/model_zoo/popular/open_conformer/conformer_encoder.onnx", 4 | "framework": "Onnx", 5 | "framework_version": "1.6", 6 | "model_format": "onnx", 7 | "model_precision": "FP32", 8 | "inputs":"src,src_pad_mask", 9 | "outputs":"memory", 10 | "input_shape": {"src": [1, 3, 64, 512], "src_pad_mask": [1, 128]}, 11 | "input_type": "FLOAT32,BOOL", 12 | "dataset_name": null, 13 | "max_batch_size": 64, 14 | "is_quantized": false 15 | } -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/model_zoo/deberta-torch-fp32.json: -------------------------------------------------------------------------------- 1 | { 2 | "model": "deberta-torch-fp32", 3 | "model_path": "general_perf/model_zoo/popular/open_deberta/deberta-base-squad.pt", 4 | "framework": "Pytorch", 5 | "framework_version": "1.10.0", 6 | "model_format": "pt", 7 | "model_precision": "FP32", 8 | "inputs":"input_ids.1,attention_mask.1,token_type_ids.1", 9 | "outputs":"start_logits,end_logits", 10 | "input_shape": {"input_ids.1": [1,384], "attention_mask.1": [1,384], "token_type_ids.1": [1,384]}, 11 | "input_type": "LONG,LONG,LONG", 12 | "dataset_name": "open_squad", 13 | "max_batch_size": 64, 14 | "is_quantized": false 15 | } 16 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/model_zoo/gpt2-torch-fp32.json: -------------------------------------------------------------------------------- 1 | { 2 | "model": "gpt2-torch-fp32", 3 | "model_path": "general_perf/model_zoo/sota/traced_gpt2/traced_gpt2.pt", 4 | "framework": "Pytorch", 5 | "framework_version": "1.8.1", 6 | "model_format": "pt", 7 | "model_precision": "FP32", 8 | "inputs":"input_ids", 9 | "outputs":"outputs", 10 | "input_shape": {"input_ids": [1,384]}, 11 | "input_type": "LONG", 12 | "dataset_name": null, 13 | "max_batch_size": 64, 14 | "is_quantized": false 15 | } 16 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/model_zoo/llama2-7b-torch-fp16.json: -------------------------------------------------------------------------------- 1 | { 2 | "model": "llama2-7b-torch-fp16", 3 | "model_path": "general_perf/model_zoo/sota/llama/llama-2-7b", 4 | "framework": "Pytorch", 5 | "framework_version": "1.13.1", 6 | "model_format": "pth", 7 | "model_precision": "FP16", 8 | "inputs": null, 9 | "outputs": null, 10 | "input_shape": null, 11 | "input_type": null, 12 | "dataset_name": null, 13 | "max_batch_size": null, 14 | "is_quantized": false 15 | } -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/model_zoo/resnet50-tf-fp32.json: -------------------------------------------------------------------------------- 1 | { 2 | "model": "resnet50-tf-fp32", 3 | "model_path": "general_perf/model_zoo/regular/open_resnet50/resnet50-fp32", 4 | "framework": "Tensorflow", 5 | "framework_version": "2.4.0", 6 | "model_format": "saved_model", 7 | "model_precision": "FP32", 8 | "inputs": "input_tensor:0", 9 | "outputs": "softmax_tensor:0", 10 | "input_shape": {"input_tensor:0": [1, 224, 224, 3]}, 11 | "input_type": "FLOAT32", 12 | "dataset_name": "open_imagenet", 13 | "max_batch_size": 64, 14 | "layout": "NHWC" 15 | } -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/model_zoo/resnet50-torch-fp32.json: -------------------------------------------------------------------------------- 1 | { 2 | "model": "resnet50-torch-fp32", 3 | "model_path": "general_perf/model_zoo/regular/open_resnet50/resnet50.pt", 4 | "framework": "Pytorch", 5 | "framework_version": "1.8.1", 6 | "model_format": "pt", 7 | "model_precision": "FP32", 8 | "inputs": "input_1.1", 9 | "outputs": "softmax_tensor:0", 10 | "input_shape": {"input_1.1": [1, 3, 224, 224]}, 11 | "input_type": "FLOAT32", 12 | "dataset_name": "open_imagenet", 13 | "max_batch_size": 64, 14 | "layout": "NCHW" 15 | } 16 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/model_zoo/roberta-torch-fp32.json: -------------------------------------------------------------------------------- 1 | { 2 | "model": "roberta-torch-fp32", 3 | "model_path": "general_perf/model_zoo/popular/open_roberta/roberta-base-squad.pt", 4 | "framework": "Pytorch", 5 | "framework_version": "1.9.1", 6 | "model_format": "pt", 7 | "model_precision": "FP32", 8 | "inputs":"input_ids.1,attention_mask.1,token_type_ids.1", 9 | "outputs":"start_logits,end_logits", 10 | "input_shape": {"input_ids.1": [1,384], "attention_mask.1": [1,384], "token_type_ids.1": [1,384]}, 11 | "input_type": "LONG,LONG,LONG", 12 | "dataset_name": "open_squad", 13 | "max_batch_size": 64, 14 | "is_quantized": false 15 | } -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/model_zoo/roformer-tf-fp32.json: -------------------------------------------------------------------------------- 1 | { 2 | "model": "roformer-tf-fp32", 3 | "model_path": "general_perf/model_zoo/popular/open_roformer", 4 | "framework": "Tensorflow", 5 | "framework_version": "2.4.0", 6 | "model_format": "saved_model", 7 | "model_precision": "FP32", 8 | "inputs": "input_segment0,input_token0", 9 | "outputs": "Identity:0", 10 | "input_shape": {"input_segment0": [1, 1024], "input_token0": [1, 1024]}, 11 | "input_type": "FLOAT32,FLOAT32", 12 | "dataset_name": "open_cail2019", 13 | "max_batch_size": 128 14 | } 15 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/model_zoo/swin-large-torch-fp32.json: -------------------------------------------------------------------------------- 1 | { 2 | "model": "swin-large-torch-fp32", 3 | "model_path": "general_perf/model_zoo/popular/swin-large/swin-transformer-large.pt", 4 | "framework": "Pytorch", 5 | "framework_version": "1.12.0", 6 | "model_format": "pt", 7 | "model_precision": "FP32", 8 | "inputs":"pixel_values.1", 9 | "outputs":"logits", 10 | "input_shape": {"pixel_values.1": [1,3,384,384]}, 11 | "input_type": "FLOAT32", 12 | "dataset_name": "open_imagenet", 13 | "max_batch_size": 64, 14 | "is_quantized": false 15 | } 16 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/model_zoo/unet-onnx-fp32.json: -------------------------------------------------------------------------------- 1 | { 2 | "model": "unet-onnx-fp32", 3 | "model_path": "general_perf/model_zoo/sota/unet/model.onnx", 4 | "framework": "Onnx", 5 | "framework_version": "1.12.0", 6 | "model_format": "onnx", 7 | "model_precision": "FP32", 8 | "inputs":"sample,timestep,encoder_hidden_states", 9 | "outputs":"out_sample", 10 | "input_shape": {"sample": [1,4,32,32],"timestep":[1],"encoder_hidden_states":[1,77,768]}, 11 | "input_type": "FLOAT32,INT64,FLOAT32", 12 | "dataset_name": null, 13 | "max_batch_size": 64, 14 | "is_quantized": false 15 | } -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/model_zoo/vae-decoder-onnx-fp32.json: -------------------------------------------------------------------------------- 1 | { 2 | "model": "vae-decoder-onnx-fp32", 3 | "model_path": "general_perf/model_zoo/sota/stable_diffusion/vae-decoder.onnx", 4 | "framework": "Onnx", 5 | "framework_version": "1.12.0", 6 | "model_format": "onnx", 7 | "model_precision": "FP32", 8 | "inputs":"latent_sample", 9 | "outputs":"Convsample_dim_0,Convsample_dim_1,Convsample_dim_2,Convsample_dim_3", 10 | "input_shape": {"latent_sample": [1,4,32,32]}, 11 | "input_type": "FLOAT32", 12 | "dataset_name": null, 13 | "max_batch_size": 64, 14 | "is_quantized": false 15 | } 16 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/model_zoo/vae-encoder-onnx-fp32.json: -------------------------------------------------------------------------------- 1 | { 2 | "model": "vae-encoder-onnx-fp32", 3 | "model_path": "general_perf/model_zoo/sota/stable_diffusion/vae-encoder.onnx", 4 | "framework": "Onnx", 5 | "framework_version": "1.12.0", 6 | "model_format": "onnx", 7 | "model_precision": "FP32", 8 | "inputs":"sample", 9 | "outputs":"latent_sample", 10 | "input_shape": {"sample": [1,3,256,256]}, 11 | "input_type": "FLOAT32", 12 | "dataset_name": null, 13 | "max_batch_size": 64, 14 | "is_quantized": false 15 | } -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/model_zoo/videobert-onnx-fp32.json: -------------------------------------------------------------------------------- 1 | { 2 | "model": "videobert-onnx-fp32", 3 | "model_path": "general_perf/model_zoo/popular/open_videobert/video-bert.onnx", 4 | "framework": "Onnx", 5 | "framework_version": "1.8.1", 6 | "model_format": "onnx", 7 | "model_precision": "FP32", 8 | "inputs":"image,text", 9 | "outputs":"output", 10 | "input_shape": {"image": [1,3,224,224], "text": [100, 77]}, 11 | "input_type": "FLOAT32,LONG", 12 | "dataset_name": "open_cifar", 13 | "max_batch_size": 64, 14 | "is_quantized": false 15 | } -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/model_zoo/widedeep-tf-fp32.json: -------------------------------------------------------------------------------- 1 | { 2 | "model": "widedeep-tf-fp32", 3 | "model_path": "general_perf/model_zoo/regular/open_wide_deep_saved_model", 4 | "framework": "Tensorflow", 5 | "framework_version": "2.4.0", 6 | "model_format": "saved_model", 7 | "model_precision": "FP32", 8 | "inputs": "new_categorical_placeholder:0,new_numeric_placeholder:0", 9 | "outputs": "import/head/predictions/probabilities:0", 10 | "input_shape": {"new_categorical_placeholder:0": [26, 2], "new_numeric_placeholder:0": [1, 13]}, 11 | "input_type": "INT64,FLOAT32", 12 | "dataset_name": "open_criteo_kaggle", 13 | "max_batch_size": 16384 14 | } -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/model_zoo/yolov5-onnx-fp32.json: -------------------------------------------------------------------------------- 1 | { 2 | "model": "yolov5-onnx-fp32", 3 | "model_path": "general_perf/model_zoo/popular/open_yolov5/yolov5s.onnx", 4 | "framework": "Onnx", 5 | "framework_version": "1.10.2", 6 | "model_format": "onnx", 7 | "model_precision": "FP32", 8 | "inputs":"images", 9 | "outputs":"output,345,403,461", 10 | "input_shape": {"images": [1,3,640,640]}, 11 | "input_type": "FLOAT32", 12 | "dataset_name": null, 13 | "max_batch_size": 64 14 | } -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/requirements.txt: -------------------------------------------------------------------------------- 1 | matplotlib 2 | pandas 3 | virtualenv==16.7.9 4 | scikit-learn 5 | prompt_toolkit 6 | tqdm 7 | opencv-python 8 | transformers 9 | tokenization 10 | fpdf 11 | typing-extensions==4.12.2 12 | numpy==1.23.0 13 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/tools/convert.sh: -------------------------------------------------------------------------------- 1 | #！bin/bash 2 | if [ ! -d "tools/venv" ]; then 3 | python3 -m virtualenv tools/venv 4 | source tools/venv/bin/activate 5 | tools/venv/bin/python3 -m pip install --upgrade pip -q 6 | tools/venv/bin/python3 -m pip install -r tools/requirements.txt -q 7 | else 8 | source tools/venv/bin/activate 9 | fi 10 | 11 | if [ "$3" == "pt2onnx" ];then 12 | python3 tools/torch_to_onnx.py --model_path $1 --output_path $2 13 | elif [ "$3" == "saved2onnx" ];then 14 | python3 tools/saved_to_onnx.py --model_path $1 --output_path $2 15 | elif [ "$3" == "saved2frozen" ];then 16 | python3 tools/saved_to_frozen.py --model_path $1 --output_path $2 17 | fi 18 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/tools/requirements.txt: -------------------------------------------------------------------------------- 1 | tensorflow>=2.6.0 2 | tf2onnx 3 | numpy 4 | torch==1.9.1 -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_infer_perf/general_perf/version.py: -------------------------------------------------------------------------------- 1 | __version__ = '1.0.0' 2 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_micro_perf/backends/GPU/requirements.txt: -------------------------------------------------------------------------------- 1 | torch==2.1.0 2 | nvidia-cutlass 3 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_micro_perf/backends/ILUVATAR/requirements.txt: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_micro_perf/backends/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_micro_perf/compiled_cache.db: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/byte_micro_perf/compiled_cache.db -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/byte_micro_perf/requirements.txt: -------------------------------------------------------------------------------- 1 | matplotlib 2 | pandas 3 | virtualenv==16.7.12 4 | scikit-learn 5 | prompt_toolkit 6 | tqdm 7 | opencv-python 8 | transformers 9 | tokenization 10 | fpdf 11 | attrs 12 | decorator 13 | typing-extensions 14 | pydot -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/docs/images/flowchat.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/docs/images/flowchat.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/docs/images/icon.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/docs/images/icon.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/AWS/mem_16.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/AWS/mem_16.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/AWS/mem_17.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/AWS/mem_17.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/Graphcore/image_12.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/Graphcore/image_12.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/Graphcore/mem_12.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/Graphcore/mem_12.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/Habana/image_14.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/Habana/image_14.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/Habana/mem_14.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/Habana/mem_14.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/Habana/pe_14.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/Habana/pe_14.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/Moffett/image_7.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/Moffett/image_7.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/Moffett/image_8.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/Moffett/image_8.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/Moffett/image_9.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/Moffett/image_9.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/Moffett/mem_7.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/Moffett/mem_7.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/Moffett/mem_8.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/Moffett/mem_8.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/Moffett/mem_9.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/Moffett/mem_9.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/Moffett/pe_7.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/Moffett/pe_7.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/Moffett/pe_9.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/Moffett/pe_9.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/NVIDIA/image_0.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/NVIDIA/image_0.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/NVIDIA/image_1.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/NVIDIA/image_1.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/NVIDIA/image_2.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/NVIDIA/image_2.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/NVIDIA/image_3.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/NVIDIA/image_3.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/NVIDIA/image_4.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/NVIDIA/image_4.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/NVIDIA/image_5.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/NVIDIA/image_5.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/NVIDIA/image_6.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/NVIDIA/image_6.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/QUALCOMM/image_10.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/QUALCOMM/image_10.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/QUALCOMM/mem_10.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/QUALCOMM/mem_10.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/QUALCOMM/pe_10.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/QUALCOMM/pe_10.png -------------------------------------------------------------------------------- /toolbox/ByteMLPerf/vendor_zoo/Stream/image_13.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Deep-Spark/DeepSparkInference/HEAD/toolbox/ByteMLPerf/vendor_zoo/Stream/image_13.png --------------------------------------------------------------------------------