Metadata-Version: 2.1
Name: fastckpt
Version: 0.0.3
Summary: A fast gradient checkpointing strategy for training with memory-efficient attention (e.g., FlashAttention).
Project-URL: Homepage, https://github.com/RulinShao/fast-ckpt
Project-URL: Bug Tracker, https://github.com/RulinShao/fast-ckpt/issues
Classifier: Programming Language :: Python :: 3
Classifier: License :: OSI Approved :: Apache Software License
Requires-Python: >=3.8
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: aiohttp
Requires-Dist: fastapi
Requires-Dist: httpx
Requires-Dist: markdown2[all]
Requires-Dist: nh3
Requires-Dist: numpy
Requires-Dist: prompt_toolkit>=3.0.0
Requires-Dist: pydantic<2,>=1
Requires-Dist: requests
Requires-Dist: rich>=10.0.0
Requires-Dist: shortuuid
Requires-Dist: tiktoken
Requires-Dist: uvicorn
Provides-Extra: model-worker
Requires-Dist: accelerate>=0.21; extra == "model-worker"
Requires-Dist: peft; extra == "model-worker"
Requires-Dist: sentencepiece; extra == "model-worker"
Requires-Dist: torch; extra == "model-worker"
Requires-Dist: transformers>=4.31.0; extra == "model-worker"
Requires-Dist: protobuf; extra == "model-worker"
Provides-Extra: webui
Requires-Dist: gradio; extra == "webui"
Provides-Extra: train
Requires-Dist: einops; extra == "train"
Requires-Dist: flash-attn>=2.0; extra == "train"
Requires-Dist: wandb; extra == "train"
Provides-Extra: llm-judge
Requires-Dist: openai; extra == "llm-judge"
Requires-Dist: anthropic>=0.3; extra == "llm-judge"
Requires-Dist: ray; extra == "llm-judge"
Provides-Extra: dev
Requires-Dist: black==23.3.0; extra == "dev"
Requires-Dist: pylint==2.8.2; extra == "dev"

# FastCkpt: fast gradient checkpoint for training with FlashAttention

This is a test repo in development.
