├── FC-generation ├── README.md ├── data_generation.py └── data_preprocess.ipynb ├── IF-generation ├── .gitignore ├── LICENSE ├── README.md ├── Self-play_with_Execution_Feedback_Improving_Instruction-following_Capabilities_of_Large_Language Models.pdf ├── code │ ├── 1_RFT.py │ ├── 1_RFT_with_kd_gpt.py │ ├── 2_verification_funcs_cases_generation.py │ ├── 2_verification_funcs_cases_generation_with_kd.py │ ├── 3_cross_validation.py │ ├── 4_check_func_backtranslator.py │ ├── 4_eval_func_backtranslator.py │ ├── 5_eval_func_backtranslator_filter.py │ ├── 6_concat_sharegpt_query.py │ ├── 7_query_vertification.py │ ├── 8_query_score_filiter.py │ ├── 9_sft_data_construction.py │ ├── __init__.py │ ├── api_request_parallel_processor.py │ └── utils.py ├── generation_example │ ├── 1_rft_prompts.jsonl │ ├── 1_rft_prompts_results.jsonl │ ├── 2_eval_func_rft_prompts.jsonl │ ├── 2_eval_func_rft_prompts_results.jsonl │ ├── 3_cross_validation.jsonl │ ├── 4_back_trans.jsonl │ ├── 4_backtranslate_prompt.jsonl │ ├── 4_backtranslate_prompt_results.jsonl │ ├── 5_back_trans_fliter.jsonl │ ├── 6_instruction_filtered_query_prompt.jsonl │ ├── 6_instruction_filtered_query_prompt_results.jsonl │ ├── 7_query_need_quality_score_prompt.jsonl │ ├── 7_query_need_quality_score_prompt_results.jsonl │ ├── 7_query_wo_score.jsonl │ └── 8_query_score_filter.jsonl ├── optional_clean.py ├── requirements.txt ├── run.sh └── sample_data │ ├── augment_instructions.txt │ ├── back_trans.jsonl │ ├── dpo_query_eval_score_results.jsonl │ ├── dpo_query_w_funcs.jsonl │ ├── eval_func_rft.jsonl │ ├── query_rft.jsonl │ ├── query_rft_score.jsonl │ ├── query_score_filter.jsonl │ ├── seed_ifeval.txt │ └── seed_instruction.txt ├── README.md └── assets └── flywheel.jpeg /FC-generation/README.md: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/FC-generation/README.md -------------------------------------------------------------------------------- /FC-generation/data_generation.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/FC-generation/data_generation.py -------------------------------------------------------------------------------- /FC-generation/data_preprocess.ipynb: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/FC-generation/data_preprocess.ipynb -------------------------------------------------------------------------------- /IF-generation/.gitignore: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/.gitignore -------------------------------------------------------------------------------- /IF-generation/LICENSE: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/LICENSE -------------------------------------------------------------------------------- /IF-generation/README.md: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/README.md -------------------------------------------------------------------------------- /IF-generation/Self-play_with_Execution_Feedback_Improving_Instruction-following_Capabilities_of_Large_Language Models.pdf: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/Self-play_with_Execution_Feedback_Improving_Instruction-following_Capabilities_of_Large_Language Models.pdf -------------------------------------------------------------------------------- /IF-generation/code/1_RFT.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/code/1_RFT.py -------------------------------------------------------------------------------- /IF-generation/code/1_RFT_with_kd_gpt.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/code/1_RFT_with_kd_gpt.py -------------------------------------------------------------------------------- /IF-generation/code/2_verification_funcs_cases_generation.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/code/2_verification_funcs_cases_generation.py -------------------------------------------------------------------------------- /IF-generation/code/2_verification_funcs_cases_generation_with_kd.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/code/2_verification_funcs_cases_generation_with_kd.py -------------------------------------------------------------------------------- /IF-generation/code/3_cross_validation.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/code/3_cross_validation.py -------------------------------------------------------------------------------- /IF-generation/code/4_check_func_backtranslator.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/code/4_check_func_backtranslator.py -------------------------------------------------------------------------------- /IF-generation/code/4_eval_func_backtranslator.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/code/4_eval_func_backtranslator.py -------------------------------------------------------------------------------- /IF-generation/code/5_eval_func_backtranslator_filter.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/code/5_eval_func_backtranslator_filter.py -------------------------------------------------------------------------------- /IF-generation/code/6_concat_sharegpt_query.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/code/6_concat_sharegpt_query.py -------------------------------------------------------------------------------- /IF-generation/code/7_query_vertification.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/code/7_query_vertification.py -------------------------------------------------------------------------------- /IF-generation/code/8_query_score_filiter.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/code/8_query_score_filiter.py -------------------------------------------------------------------------------- /IF-generation/code/9_sft_data_construction.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/code/9_sft_data_construction.py -------------------------------------------------------------------------------- /IF-generation/code/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /IF-generation/code/api_request_parallel_processor.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/code/api_request_parallel_processor.py -------------------------------------------------------------------------------- /IF-generation/code/utils.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/code/utils.py -------------------------------------------------------------------------------- /IF-generation/generation_example/1_rft_prompts.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/generation_example/1_rft_prompts.jsonl -------------------------------------------------------------------------------- /IF-generation/generation_example/1_rft_prompts_results.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/generation_example/1_rft_prompts_results.jsonl -------------------------------------------------------------------------------- /IF-generation/generation_example/2_eval_func_rft_prompts.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/generation_example/2_eval_func_rft_prompts.jsonl -------------------------------------------------------------------------------- /IF-generation/generation_example/2_eval_func_rft_prompts_results.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/generation_example/2_eval_func_rft_prompts_results.jsonl -------------------------------------------------------------------------------- /IF-generation/generation_example/3_cross_validation.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/generation_example/3_cross_validation.jsonl -------------------------------------------------------------------------------- /IF-generation/generation_example/4_back_trans.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/generation_example/4_back_trans.jsonl -------------------------------------------------------------------------------- /IF-generation/generation_example/4_backtranslate_prompt.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/generation_example/4_backtranslate_prompt.jsonl -------------------------------------------------------------------------------- /IF-generation/generation_example/4_backtranslate_prompt_results.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/generation_example/4_backtranslate_prompt_results.jsonl -------------------------------------------------------------------------------- /IF-generation/generation_example/5_back_trans_fliter.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/generation_example/5_back_trans_fliter.jsonl -------------------------------------------------------------------------------- /IF-generation/generation_example/6_instruction_filtered_query_prompt.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/generation_example/6_instruction_filtered_query_prompt.jsonl -------------------------------------------------------------------------------- /IF-generation/generation_example/6_instruction_filtered_query_prompt_results.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/generation_example/6_instruction_filtered_query_prompt_results.jsonl -------------------------------------------------------------------------------- /IF-generation/generation_example/7_query_need_quality_score_prompt.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/generation_example/7_query_need_quality_score_prompt.jsonl -------------------------------------------------------------------------------- /IF-generation/generation_example/7_query_need_quality_score_prompt_results.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/generation_example/7_query_need_quality_score_prompt_results.jsonl -------------------------------------------------------------------------------- /IF-generation/generation_example/7_query_wo_score.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/generation_example/7_query_wo_score.jsonl -------------------------------------------------------------------------------- /IF-generation/generation_example/8_query_score_filter.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/generation_example/8_query_score_filter.jsonl -------------------------------------------------------------------------------- /IF-generation/optional_clean.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/optional_clean.py -------------------------------------------------------------------------------- /IF-generation/requirements.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/requirements.txt -------------------------------------------------------------------------------- /IF-generation/run.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/run.sh -------------------------------------------------------------------------------- /IF-generation/sample_data/augment_instructions.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/sample_data/augment_instructions.txt -------------------------------------------------------------------------------- /IF-generation/sample_data/back_trans.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/sample_data/back_trans.jsonl -------------------------------------------------------------------------------- /IF-generation/sample_data/dpo_query_eval_score_results.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/sample_data/dpo_query_eval_score_results.jsonl -------------------------------------------------------------------------------- /IF-generation/sample_data/dpo_query_w_funcs.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/sample_data/dpo_query_w_funcs.jsonl -------------------------------------------------------------------------------- /IF-generation/sample_data/eval_func_rft.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/sample_data/eval_func_rft.jsonl -------------------------------------------------------------------------------- /IF-generation/sample_data/query_rft.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/sample_data/query_rft.jsonl -------------------------------------------------------------------------------- /IF-generation/sample_data/query_rft_score.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/sample_data/query_rft_score.jsonl -------------------------------------------------------------------------------- /IF-generation/sample_data/query_score_filter.jsonl: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/sample_data/query_score_filter.jsonl -------------------------------------------------------------------------------- /IF-generation/sample_data/seed_ifeval.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/sample_data/seed_ifeval.txt -------------------------------------------------------------------------------- /IF-generation/sample_data/seed_instruction.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/IF-generation/sample_data/seed_instruction.txt -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/README.md -------------------------------------------------------------------------------- /assets/flywheel.jpeg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/shizhediao/Post-Training-Data-Flywheel/HEAD/assets/flywheel.jpeg --------------------------------------------------------------------------------