├── .gitignore ├── LICENSE ├── README.md ├── classify ├── classifiers │ └── lr_wrapper.py ├── cls_datasets │ ├── cache_jigsaw_datasets.py │ └── cache_sentiment_datasets.py ├── logistic_regression.py └── logistic_regression_jigsaw.py ├── configs ├── config.yaml ├── diffusion │ ├── arch │ │ └── tx-12l-768d-64t.yaml │ ├── default.yaml │ ├── loss │ │ ├── asymmetric_lognormal_v.yaml │ │ └── lognormal_v.yaml │ ├── sampling │ │ └── cosine.yaml │ └── train │ │ └── default.yaml ├── eval │ └── default.yaml └── prompt │ ├── arch │ ├── tx-6l-768d-ft.yaml │ └── tx-6l-768d.yaml │ ├── augmentation │ ├── cosine-scale1.yaml │ ├── cosine-scale2.yaml │ └── cosine-scale3.yaml │ ├── default.yaml │ ├── ft-gpt.yaml │ └── train │ └── default.yaml ├── diffusion ├── diff_utils.py ├── loss_weighting.py ├── noise_schedule.py ├── time_sampler.py └── trainer.py ├── evaluation └── wrappers.py ├── models ├── diff_gpt.py ├── modules │ ├── blocks.py │ ├── diffusion.py │ ├── norm.py │ ├── position.py │ └── transformer.py └── optimization │ └── optimizer.py ├── requirements.txt ├── scripts ├── log_reg │ ├── cache_jigsaw.sh │ ├── cache_sentiment.sh │ ├── train_jigsaw.sh │ └── train_sentiment.sh └── train │ ├── diffusion │ └── default.sh │ └── prompt │ └── default.sh ├── text_datasets ├── CONSTANTS.py ├── data_stats │ └── clean_c4 │ │ ├── mean.pt │ │ ├── stats.json │ │ └── std.pt ├── dataset_utils.py ├── extract_prompt_continuations.py ├── preprocess.py └── scripts │ ├── c4_10mill.sh │ └── extract_eval_text.sh ├── train_diff.py └── utils ├── file_utils.py └── torch_utils.py /.gitignore: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/.gitignore -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/LICENSE -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/README.md -------------------------------------------------------------------------------- /classify/classifiers/lr_wrapper.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/classify/classifiers/lr_wrapper.py -------------------------------------------------------------------------------- /classify/cls_datasets/cache_jigsaw_datasets.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/classify/cls_datasets/cache_jigsaw_datasets.py -------------------------------------------------------------------------------- /classify/cls_datasets/cache_sentiment_datasets.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/classify/cls_datasets/cache_sentiment_datasets.py -------------------------------------------------------------------------------- /classify/logistic_regression.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/classify/logistic_regression.py -------------------------------------------------------------------------------- /classify/logistic_regression_jigsaw.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/classify/logistic_regression_jigsaw.py -------------------------------------------------------------------------------- /configs/config.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/configs/config.yaml -------------------------------------------------------------------------------- /configs/diffusion/arch/tx-12l-768d-64t.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/configs/diffusion/arch/tx-12l-768d-64t.yaml -------------------------------------------------------------------------------- /configs/diffusion/default.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/configs/diffusion/default.yaml -------------------------------------------------------------------------------- /configs/diffusion/loss/asymmetric_lognormal_v.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/configs/diffusion/loss/asymmetric_lognormal_v.yaml -------------------------------------------------------------------------------- /configs/diffusion/loss/lognormal_v.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/configs/diffusion/loss/lognormal_v.yaml -------------------------------------------------------------------------------- /configs/diffusion/sampling/cosine.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/configs/diffusion/sampling/cosine.yaml -------------------------------------------------------------------------------- /configs/diffusion/train/default.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/configs/diffusion/train/default.yaml -------------------------------------------------------------------------------- /configs/eval/default.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/configs/eval/default.yaml -------------------------------------------------------------------------------- /configs/prompt/arch/tx-6l-768d-ft.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/configs/prompt/arch/tx-6l-768d-ft.yaml -------------------------------------------------------------------------------- /configs/prompt/arch/tx-6l-768d.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/configs/prompt/arch/tx-6l-768d.yaml -------------------------------------------------------------------------------- /configs/prompt/augmentation/cosine-scale1.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/configs/prompt/augmentation/cosine-scale1.yaml -------------------------------------------------------------------------------- /configs/prompt/augmentation/cosine-scale2.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/configs/prompt/augmentation/cosine-scale2.yaml -------------------------------------------------------------------------------- /configs/prompt/augmentation/cosine-scale3.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/configs/prompt/augmentation/cosine-scale3.yaml -------------------------------------------------------------------------------- /configs/prompt/default.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/configs/prompt/default.yaml -------------------------------------------------------------------------------- /configs/prompt/ft-gpt.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/configs/prompt/ft-gpt.yaml -------------------------------------------------------------------------------- /configs/prompt/train/default.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/configs/prompt/train/default.yaml -------------------------------------------------------------------------------- /diffusion/diff_utils.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/diffusion/diff_utils.py -------------------------------------------------------------------------------- /diffusion/loss_weighting.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/diffusion/loss_weighting.py -------------------------------------------------------------------------------- /diffusion/noise_schedule.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/diffusion/noise_schedule.py -------------------------------------------------------------------------------- /diffusion/time_sampler.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/diffusion/time_sampler.py -------------------------------------------------------------------------------- /diffusion/trainer.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/diffusion/trainer.py -------------------------------------------------------------------------------- /evaluation/wrappers.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/evaluation/wrappers.py -------------------------------------------------------------------------------- /models/diff_gpt.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/models/diff_gpt.py -------------------------------------------------------------------------------- /models/modules/blocks.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/models/modules/blocks.py -------------------------------------------------------------------------------- /models/modules/diffusion.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/models/modules/diffusion.py -------------------------------------------------------------------------------- /models/modules/norm.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/models/modules/norm.py -------------------------------------------------------------------------------- /models/modules/position.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/models/modules/position.py -------------------------------------------------------------------------------- /models/modules/transformer.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/models/modules/transformer.py -------------------------------------------------------------------------------- /models/optimization/optimizer.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/models/optimization/optimizer.py -------------------------------------------------------------------------------- /requirements.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/requirements.txt -------------------------------------------------------------------------------- /scripts/log_reg/cache_jigsaw.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/scripts/log_reg/cache_jigsaw.sh -------------------------------------------------------------------------------- /scripts/log_reg/cache_sentiment.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/scripts/log_reg/cache_sentiment.sh -------------------------------------------------------------------------------- /scripts/log_reg/train_jigsaw.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/scripts/log_reg/train_jigsaw.sh -------------------------------------------------------------------------------- /scripts/log_reg/train_sentiment.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/scripts/log_reg/train_sentiment.sh -------------------------------------------------------------------------------- /scripts/train/diffusion/default.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/scripts/train/diffusion/default.sh -------------------------------------------------------------------------------- /scripts/train/prompt/default.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/scripts/train/prompt/default.sh -------------------------------------------------------------------------------- /text_datasets/CONSTANTS.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/text_datasets/CONSTANTS.py -------------------------------------------------------------------------------- /text_datasets/data_stats/clean_c4/mean.pt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/text_datasets/data_stats/clean_c4/mean.pt -------------------------------------------------------------------------------- /text_datasets/data_stats/clean_c4/stats.json: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/text_datasets/data_stats/clean_c4/stats.json -------------------------------------------------------------------------------- /text_datasets/data_stats/clean_c4/std.pt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/text_datasets/data_stats/clean_c4/std.pt -------------------------------------------------------------------------------- /text_datasets/dataset_utils.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/text_datasets/dataset_utils.py -------------------------------------------------------------------------------- /text_datasets/extract_prompt_continuations.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/text_datasets/extract_prompt_continuations.py -------------------------------------------------------------------------------- /text_datasets/preprocess.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/text_datasets/preprocess.py -------------------------------------------------------------------------------- /text_datasets/scripts/c4_10mill.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/text_datasets/scripts/c4_10mill.sh -------------------------------------------------------------------------------- /text_datasets/scripts/extract_eval_text.sh: -------------------------------------------------------------------------------- 1 | python extract_prompt_continuations.py --num_threads 4 -------------------------------------------------------------------------------- /train_diff.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/train_diff.py -------------------------------------------------------------------------------- /utils/file_utils.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/utils/file_utils.py -------------------------------------------------------------------------------- /utils/torch_utils.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/justinlovelace/Diffusion-Guided-LM/HEAD/utils/torch_utils.py --------------------------------------------------------------------------------