cold-compress
Cold Compress is a hackable, lightweight, and open-source toolkit for creating and benchmarking cache compression methods built on top of GPT-Fast, a simple, PyTorch-native generation codebase.
파일 탐색기
최종 버전 다운로드 (.zip)- debug_heavy_hitter.yaml
- fastgen.yaml
- full.yaml
- heavy_hitter.yaml
- heavy_hitter_funnel.yaml
- heavy_hitter_pyramid.yaml
- hybrid.yaml
- keep_it_odd.yaml
- l2.yaml
- local_global.yaml
- random.yaml
- recent_global.yaml
- task_stats.csv
- dolomites_BertScore_f1.jpg
- dolomites_LLM-Rouge_llm_rouge.jpg
- dolomites_Rouge_rougeL.jpg
- musique_BertScore_f1.jpg
- musique_LLM-Rouge_llm_rouge.jpg
- musique_Rouge_rougeL.jpg
- qmsum_BertScore_f1.jpg
- qmsum_LLM-Rouge_llm_rouge.jpg
- qmsum_Rouge_rougeL.jpg
- reprobench_ExactMatch.jpg
- reprobench_Levenshtein.jpg
- rulercwe_StringMatch_score.jpg
- rulerniah_StringMatch_score.jpg
- rulerqa_StringMatch_score.jpg
- rulervt_StringMatch_score.jpg
- scrollsquality_Accuracy.jpg
- squality_BertScore_f1.jpg
- squality_LLM-Rouge_llm_rouge.jpg
- squality_Rouge_rougeL.jpg
- truthfulqa_Accuracy.jpg
- dolomites_BertScore_f1.jpg
- dolomites_LLM-Rouge_llm_rouge.jpg
- dolomites_Rouge_rougeL.jpg
- musique_BertScore_f1.jpg
- musique_LLM-Rouge_llm_rouge.jpg
- musique_Rouge_rougeL.jpg
- qmsum_BertScore_f1.jpg
- qmsum_LLM-Rouge_llm_rouge.jpg
- qmsum_Rouge_rougeL.jpg
- reprobench_ExactMatch.jpg
- reprobench_Levenshtein.jpg
- rulercwe_StringMatch_score.jpg
- rulerniah_StringMatch_score.jpg
- rulerqa_StringMatch_score.jpg
- rulervt_StringMatch_score.jpg
- scrollsquality_Accuracy.jpg
- squality_BertScore_f1.jpg
- squality_LLM-Rouge_llm_rouge.jpg
- squality_Rouge_rougeL.jpg
- truthfulqa_Accuracy.jpg
- attention_loss.png
- attention_loss.py
- blogpost_perf.py
- attention_loss.sh
- eval_all.sh
- multi_strategy.sh
- multi_strategy.txt
- variable_compression.sh
- attention_loss_concept.png
- attention_loss_pg19.png
- cold_compress_logo.jpg
- kv_cache_compression.png
- kv_cache_flow.png
- local_global_from_character_ai.png
- long_prompt_long_output.txt
- long_prompt_short_output.txt
- noise_qa.txt
- README.md
- reverse_list.txt
- short_prompt_long_output.txt
- short_prompt_short_output.txt
- convert_hf_checkpoint.py
- download.py
- prepare.sh
- prepare_llama2.sh
- prepare_llama3.sh
- prepare_llama31.sh
- prepare_qwen2.sh
- .gitignore
- attention_utils.py
- BENCHMARK.md
- cache.py
- DISCLAIMER.md
- eval.py
- eval_multi.py
- generate.py
- generation_utils.py
- GPTQ.py
- LICENSE
- metric.py
- model.py
- parallelize_evals.py
- prompt_compression.py
- quantization_utils.py
- quantize.py
- README.md
- requirements.txt
- setup.py
- task.py
- tokenizer.py
- tp.py
# 설치 가이드
1. 코드 내려받기
git clone https://github.com/AnswerDotAI/cold-compress
깃허브에서 프로젝트 코드 전체를 내 컴퓨터로 내려받습니다.
cd cold-compress
방금 내려받은 프로젝트 폴더 안으로 이동합니다.
2. Python
쉬움 추천사전 준비물
pip install -r requirements.txt --extra-index-url https://download.pytorch.org/whl/nightly/
requirements.txt 등에 명시된 파이썬 라이브러리를 설치합니다.
에러 메시지 없이 실행되고 터미널에 안내 문구가 출력되면 정상입니다.
이 레포의 README에 적힌 실제 명령어를 그대로 가져왔습니다.
// repository documentation
Was this content helpful?
(0 ratings)
