├── .github └── workflows │ └── python-publish.yml ├── .gitignore ├── LICENSE ├── README.md ├── benchmarks ├── inference │ ├── amd_rx_7900xtx.txt │ ├── intel_arc_a770.txt │ ├── nvidia_rtx_3090.txt │ ├── nvidia_rtx_4090.txt │ ├── nvidia_rtx_5090.txt │ └── nvidia_rtx_pro_6000.txt └── training │ ├── amd_rx_7900xtx.txt │ ├── intel_arc_a770.txt │ ├── nvidia_rtx_3090.txt │ ├── nvidia_rtx_4090.txt │ ├── nvidia_rtx_5090.txt │ └── nvidia_rtx_pro_6000.txt ├── pyproject.toml ├── scripts ├── benchmark_sdnq_inference.py ├── benchmark_sdnq_training.py └── dequantize_sdnq_training.py └── src └── sdnq ├── __init__.py ├── common.py ├── dequantizer.py ├── file_loader.py ├── forward.py ├── layers ├── conv │ ├── conv_fp16.py │ ├── conv_fp8.py │ ├── conv_fp8_tensorwise.py │ ├── conv_int8.py │ └── forward.py └── linear │ ├── forward.py │ ├── linear_fp16.py │ ├── linear_fp8.py │ ├── linear_fp8_tensorwise.py │ └── linear_int8.py ├── loader.py ├── optim ├── __init__.py ├── adafactor.py ├── adamw.py ├── came.py ├── lion.py ├── muon.py ├── optimizer.py └── utils.py ├── packed_int.py ├── quantizer.py ├── sdnext.py ├── training ├── __init__.py ├── forward.py ├── layers │ └── linear │ │ ├── forward.py │ │ ├── linear_fp16.py │ │ ├── linear_fp16_ckpt.py │ │ ├── linear_fp16_dynamic.py │ │ ├── linear_fp16_dynamic_ckpt.py │ │ ├── linear_fp8.py │ │ ├── linear_fp8_ckpt.py │ │ ├── linear_fp8_dynamic.py │ │ ├── linear_fp8_dynamic_ckpt.py │ │ ├── linear_fp8_tensorwise.py │ │ ├── linear_fp8_tensorwise_ckpt.py │ │ ├── linear_fp8_tensorwise_dynamic.py │ │ ├── linear_fp8_tensorwise_dynamic_ckpt.py │ │ ├── linear_int8.py │ │ ├── linear_int8_ckpt.py │ │ ├── linear_int8_dynamic.py │ │ └── linear_int8_dynamic_ckpt.py └── tensor.py └── triton_mm.py /.github/workflows/python-publish.yml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/.github/workflows/python-publish.yml -------------------------------------------------------------------------------- /.gitignore: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/.gitignore -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/LICENSE -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/README.md -------------------------------------------------------------------------------- /benchmarks/inference/amd_rx_7900xtx.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/benchmarks/inference/amd_rx_7900xtx.txt -------------------------------------------------------------------------------- /benchmarks/inference/intel_arc_a770.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/benchmarks/inference/intel_arc_a770.txt -------------------------------------------------------------------------------- /benchmarks/inference/nvidia_rtx_3090.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/benchmarks/inference/nvidia_rtx_3090.txt -------------------------------------------------------------------------------- /benchmarks/inference/nvidia_rtx_4090.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/benchmarks/inference/nvidia_rtx_4090.txt -------------------------------------------------------------------------------- /benchmarks/inference/nvidia_rtx_5090.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/benchmarks/inference/nvidia_rtx_5090.txt -------------------------------------------------------------------------------- /benchmarks/inference/nvidia_rtx_pro_6000.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/benchmarks/inference/nvidia_rtx_pro_6000.txt -------------------------------------------------------------------------------- /benchmarks/training/amd_rx_7900xtx.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/benchmarks/training/amd_rx_7900xtx.txt -------------------------------------------------------------------------------- /benchmarks/training/intel_arc_a770.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/benchmarks/training/intel_arc_a770.txt -------------------------------------------------------------------------------- /benchmarks/training/nvidia_rtx_3090.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/benchmarks/training/nvidia_rtx_3090.txt -------------------------------------------------------------------------------- /benchmarks/training/nvidia_rtx_4090.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/benchmarks/training/nvidia_rtx_4090.txt -------------------------------------------------------------------------------- /benchmarks/training/nvidia_rtx_5090.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/benchmarks/training/nvidia_rtx_5090.txt -------------------------------------------------------------------------------- /benchmarks/training/nvidia_rtx_pro_6000.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/benchmarks/training/nvidia_rtx_pro_6000.txt -------------------------------------------------------------------------------- /pyproject.toml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/pyproject.toml -------------------------------------------------------------------------------- /scripts/benchmark_sdnq_inference.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/scripts/benchmark_sdnq_inference.py -------------------------------------------------------------------------------- /scripts/benchmark_sdnq_training.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/scripts/benchmark_sdnq_training.py -------------------------------------------------------------------------------- /scripts/dequantize_sdnq_training.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/scripts/dequantize_sdnq_training.py -------------------------------------------------------------------------------- /src/sdnq/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/__init__.py -------------------------------------------------------------------------------- /src/sdnq/common.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/common.py -------------------------------------------------------------------------------- /src/sdnq/dequantizer.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/dequantizer.py -------------------------------------------------------------------------------- /src/sdnq/file_loader.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/file_loader.py -------------------------------------------------------------------------------- /src/sdnq/forward.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/forward.py -------------------------------------------------------------------------------- /src/sdnq/layers/conv/conv_fp16.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/layers/conv/conv_fp16.py -------------------------------------------------------------------------------- /src/sdnq/layers/conv/conv_fp8.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/layers/conv/conv_fp8.py -------------------------------------------------------------------------------- /src/sdnq/layers/conv/conv_fp8_tensorwise.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/layers/conv/conv_fp8_tensorwise.py -------------------------------------------------------------------------------- /src/sdnq/layers/conv/conv_int8.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/layers/conv/conv_int8.py -------------------------------------------------------------------------------- /src/sdnq/layers/conv/forward.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/layers/conv/forward.py -------------------------------------------------------------------------------- /src/sdnq/layers/linear/forward.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/layers/linear/forward.py -------------------------------------------------------------------------------- /src/sdnq/layers/linear/linear_fp16.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/layers/linear/linear_fp16.py -------------------------------------------------------------------------------- /src/sdnq/layers/linear/linear_fp8.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/layers/linear/linear_fp8.py -------------------------------------------------------------------------------- /src/sdnq/layers/linear/linear_fp8_tensorwise.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/layers/linear/linear_fp8_tensorwise.py -------------------------------------------------------------------------------- /src/sdnq/layers/linear/linear_int8.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/layers/linear/linear_int8.py -------------------------------------------------------------------------------- /src/sdnq/loader.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/loader.py -------------------------------------------------------------------------------- /src/sdnq/optim/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/optim/__init__.py -------------------------------------------------------------------------------- /src/sdnq/optim/adafactor.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/optim/adafactor.py -------------------------------------------------------------------------------- /src/sdnq/optim/adamw.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/optim/adamw.py -------------------------------------------------------------------------------- /src/sdnq/optim/came.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/optim/came.py -------------------------------------------------------------------------------- /src/sdnq/optim/lion.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/optim/lion.py -------------------------------------------------------------------------------- /src/sdnq/optim/muon.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/optim/muon.py -------------------------------------------------------------------------------- /src/sdnq/optim/optimizer.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/optim/optimizer.py -------------------------------------------------------------------------------- /src/sdnq/optim/utils.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/optim/utils.py -------------------------------------------------------------------------------- /src/sdnq/packed_int.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/packed_int.py -------------------------------------------------------------------------------- /src/sdnq/quantizer.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/quantizer.py -------------------------------------------------------------------------------- /src/sdnq/sdnext.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/sdnext.py -------------------------------------------------------------------------------- /src/sdnq/training/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/training/__init__.py -------------------------------------------------------------------------------- /src/sdnq/training/forward.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/training/forward.py -------------------------------------------------------------------------------- /src/sdnq/training/layers/linear/forward.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/training/layers/linear/forward.py -------------------------------------------------------------------------------- /src/sdnq/training/layers/linear/linear_fp16.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/training/layers/linear/linear_fp16.py -------------------------------------------------------------------------------- /src/sdnq/training/layers/linear/linear_fp16_ckpt.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/training/layers/linear/linear_fp16_ckpt.py -------------------------------------------------------------------------------- /src/sdnq/training/layers/linear/linear_fp16_dynamic.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/training/layers/linear/linear_fp16_dynamic.py -------------------------------------------------------------------------------- /src/sdnq/training/layers/linear/linear_fp16_dynamic_ckpt.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/training/layers/linear/linear_fp16_dynamic_ckpt.py -------------------------------------------------------------------------------- /src/sdnq/training/layers/linear/linear_fp8.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/training/layers/linear/linear_fp8.py -------------------------------------------------------------------------------- /src/sdnq/training/layers/linear/linear_fp8_ckpt.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/training/layers/linear/linear_fp8_ckpt.py -------------------------------------------------------------------------------- /src/sdnq/training/layers/linear/linear_fp8_dynamic.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/training/layers/linear/linear_fp8_dynamic.py -------------------------------------------------------------------------------- /src/sdnq/training/layers/linear/linear_fp8_dynamic_ckpt.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/training/layers/linear/linear_fp8_dynamic_ckpt.py -------------------------------------------------------------------------------- /src/sdnq/training/layers/linear/linear_fp8_tensorwise.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/training/layers/linear/linear_fp8_tensorwise.py -------------------------------------------------------------------------------- /src/sdnq/training/layers/linear/linear_fp8_tensorwise_ckpt.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/training/layers/linear/linear_fp8_tensorwise_ckpt.py -------------------------------------------------------------------------------- /src/sdnq/training/layers/linear/linear_fp8_tensorwise_dynamic.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/training/layers/linear/linear_fp8_tensorwise_dynamic.py -------------------------------------------------------------------------------- /src/sdnq/training/layers/linear/linear_fp8_tensorwise_dynamic_ckpt.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/training/layers/linear/linear_fp8_tensorwise_dynamic_ckpt.py -------------------------------------------------------------------------------- /src/sdnq/training/layers/linear/linear_int8.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/training/layers/linear/linear_int8.py -------------------------------------------------------------------------------- /src/sdnq/training/layers/linear/linear_int8_ckpt.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/training/layers/linear/linear_int8_ckpt.py -------------------------------------------------------------------------------- /src/sdnq/training/layers/linear/linear_int8_dynamic.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/training/layers/linear/linear_int8_dynamic.py -------------------------------------------------------------------------------- /src/sdnq/training/layers/linear/linear_int8_dynamic_ckpt.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/training/layers/linear/linear_int8_dynamic_ckpt.py -------------------------------------------------------------------------------- /src/sdnq/training/tensor.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/training/tensor.py -------------------------------------------------------------------------------- /src/sdnq/triton_mm.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/Disty0/sdnq/HEAD/src/sdnq/triton_mm.py --------------------------------------------------------------------------------