From 66e6aa8f3735d12d646151437a12aef2f8f42324 Mon Sep 17 00:00:00 2001 From: hiyouga Date: Sat, 27 Dec 2025 07:35:18 +0800 Subject: [PATCH] fixup --- Makefile | 21 ++++++++++++--------- README.md | 14 ++++---------- README_zh.md | 18 +++++++----------- docker/docker-cuda/Dockerfile | 2 +- docker/docker-cuda/Dockerfile.megatron | 2 +- docker/docker-cuda/docker-compose.yml | 1 - docker/docker-npu/Dockerfile | 2 +- docker/docker-npu/docker-compose.yml | 1 - docker/docker-rocm/Dockerfile | 2 +- docker/docker-rocm/docker-compose.yml | 1 - examples/requirements/dev.txt | 4 ---- pyproject.toml | 24 +++++++++++------------- src/llamafactory/data/mm_plugin.py | 8 ++++++-- src/llamafactory/extras/misc.py | 4 ++++ src/llamafactory/model/loader.py | 2 +- src/llamafactory/model/patcher.py | 15 ++++++--------- src/llamafactory/train/test_utils.py | 2 +- src/llamafactory/webui/locales.py | 15 ++++++++++----- 18 files changed, 66 insertions(+), 72 deletions(-) delete mode 100644 examples/requirements/dev.txt diff --git a/Makefile b/Makefile index 4c3a3216d..f30b15c1a 100644 --- a/Makefile +++ b/Makefile @@ -2,23 +2,26 @@ check_dirs := scripts src tests tests_v1 +RUN := $(shell command -v uv >/dev/null 2>&1 && echo "uv run" || echo "") +BUILD := $(shell command -v uv >/dev/null 2>&1 && echo "uv build" || echo "python -m build") + build: - uv build + $(BUILD) commit: - uv run pre-commit install - uv run pre-commit run --all-files + $(RUN) pre-commit install + $(RUN) pre-commit run --all-files license: - uv run python tests/check_license.py $(check_dirs) + $(RUN) python3 tests/check_license.py $(check_dirs) quality: - uv run ruff check $(check_dirs) - uv run ruff format --check $(check_dirs) + $(RUN) ruff check $(check_dirs) + $(RUN) ruff format --check $(check_dirs) style: - uv run ruff check $(check_dirs) --fix - uv run ruff format $(check_dirs) + $(RUN) ruff check $(check_dirs) --fix + $(RUN) ruff format $(check_dirs) test: - WANDB_DISABLED=true uv run pytest -vv --import-mode=importlib tests/ tests_v1/ + WANDB_DISABLED=true $(RUN) pytest -vv --import-mode=importlib tests/ tests_v1/ diff --git a/README.md b/README.md index 919cb2061..d1f0f1176 100644 --- a/README.md +++ b/README.md @@ -538,13 +538,7 @@ Please refer to [build docker](#build-docker) to build the image yourself. Create an isolated Python environment with [uv](https://github.com/astral-sh/uv): ```bash -uv sync --extra torch --extra metrics --prerelease=allow -``` - -Run LLaMA-Factory in the isolated environment: - -```bash -uv run --prerelease=allow llamafactory-cli train examples/train_lora/llama3_lora_pretrain.yaml +uv run llamafactory-cli webui ``` @@ -581,7 +575,7 @@ To enable FlashAttention-2 on the Windows platform, please use the script from [
For Ascend NPU users -To install LLaMA Factory on Ascend NPU devices, please upgrade Python to version 3.10 or higher: `pip install -e "."`. Additionally, you need to install the **[Ascend CANN Toolkit and Kernels](https://www.hiascend.com/developer/download/community/result?module=cann)**. Please follow the [installation tutorial](https://www.hiascend.com/document/detail/en/CANNCommunityEdition/600alphaX/softwareinstall/instg/atlasdeploy_03_0031.html) or use the following commands: +To install LLaMA Factory on Ascend NPU devices, please upgrade Python to version 3.10 or higher: `pip install -e . torch-npu==2.7.1`. Additionally, you need to install the **[Ascend CANN Toolkit and Kernels](https://www.hiascend.com/developer/download/community/result?module=cann)**. Please follow the [installation tutorial](https://www.hiascend.com/document/detail/en/CANNCommunityEdition/600alphaX/softwareinstall/instg/atlasdeploy_03_0031.html) or use the following commands: ```bash # replace the url according to your CANN version and devices @@ -600,8 +594,8 @@ source /usr/local/Ascend/ascend-toolkit/set_env.sh | Requirement | Minimum | Recommend | | ------------ | ------- | -------------- | | CANN | 8.0.RC1 | 8.0.0.alpha002 | -| torch | 2.1.0 | 2.4.0 | -| torch-npu | 2.1.0 | 2.4.0.post2 | +| torch | 2.1.0 | 2.7.1 | +| torch-npu | 2.1.0 | 2.7.1 | | deepspeed | 0.13.2 | 0.13.2 | | vllm-ascend | - | 0.7.3 | diff --git a/README_zh.md b/README_zh.md index 7995c88ce..ca782885b 100644 --- a/README_zh.md +++ b/README_zh.md @@ -519,7 +519,9 @@ cd LLaMA-Factory pip install -e ".[torch,metrics]" --no-build-isolation ``` -可选的额外依赖项：torch、torch-npu、metrics、deepspeed、liger-kernel、bitsandbytes、hqq、eetq、gptq、aqlm、vllm、sglang、galore、apollo、badam、adam-mini、qwen、minicpm_v、openmind、swanlab、dev +可选的额外依赖项：`metrics`、`deepspeed`。使用 `pip install -e ".[metrics,deepspeed]"` 安装。 + +其他可选依赖项请参考 `examples/requirements/` 目录下的文件。 #### 从镜像安装 @@ -538,13 +540,7 @@ docker run -it --rm --gpus=all --ipc=host hiyouga/llamafactory:latest 使用 [uv](https://github.com/astral-sh/uv) 创建隔离的 Python 环境： ```bash -uv sync --extra torch --extra metrics --prerelease=allow -``` - -在环境中运行 LLaMA-Factory： - -```bash -uv run --prerelease=allow llamafactory-cli train examples/train_lora/llama3_lora_pretrain.yaml +uv run llamafactory-cli webui ```
@@ -581,7 +577,7 @@ pip install https://github.com/jllllll/bitsandbytes-windows-webui/releases/downl
昇腾 NPU 用户指南 -在昇腾 NPU 设备上安装 LLaMA Factory 时，请升级 Python 到 3.10 及以上，并需要指定额外依赖项，使用 `pip install -e ".[torch-npu,metrics]"` 命令安装。此外，还需要安装 **[Ascend CANN Toolkit 与 Kernels](https://www.hiascend.com/developer/download/community/result?module=cann)**，安装方法请参考[安装教程](https://www.hiascend.com/document/detail/zh/CANNCommunityEdition/80RC2alpha002/quickstart/quickstart/quickstart_18_0004.html)或使用以下命令： +在昇腾 NPU 设备上安装 LLaMA Factory 时，请升级 Python 到 3.10 及以上，并需要指定额外依赖项，使用 `pip install -e . torch-npu==2.7.1` 命令安装。此外，还需要安装 **[Ascend CANN Toolkit 与 Kernels](https://www.hiascend.com/developer/download/community/result?module=cann)**，安装方法请参考[安装教程](https://www.hiascend.com/document/detail/zh/CANNCommunityEdition/80RC2alpha002/quickstart/quickstart/quickstart_18_0004.html)或使用以下命令： ```bash # 请替换 URL 为 CANN 版本和设备型号对应的 URL @@ -600,8 +596,8 @@ source /usr/local/Ascend/ascend-toolkit/set_env.sh | 依赖项 | 至少 | 推荐 | | ------------ | ------- | -------------- | | CANN | 8.0.RC1 | 8.0.0.alpha002 | -| torch | 2.1.0 | 2.4.0 | -| torch-npu | 2.1.0 | 2.4.0.post2 | +| torch | 2.1.0 | 2.7.1 | +| torch-npu | 2.1.0 | 2.7.1 | | deepspeed | 0.13.2 | 0.13.2 | | vllm-ascend | - | 0.7.3 | diff --git a/docker/docker-cuda/Dockerfile b/docker/docker-cuda/Dockerfile index 94184e791..2315a85b7 100644 --- a/docker/docker-cuda/Dockerfile +++ b/docker/docker-cuda/Dockerfile @@ -32,7 +32,7 @@ RUN pip config set global.index-url "${PIP_INDEX}" && \ COPY . /app # Install LLaMA Factory -RUN pip install --no-cache-dir -e "." --no-build-isolation +RUN pip install --no-cache-dir -e ".[metrics,deepspeed]" --no-build-isolation # Rebuild flash attention RUN if [ "${INSTALL_FLASHATTN}" == "true" ]; then \ diff --git a/docker/docker-cuda/Dockerfile.megatron b/docker/docker-cuda/Dockerfile.megatron index ee6913192..7f457fefd 100644 --- a/docker/docker-cuda/Dockerfile.megatron +++ b/docker/docker-cuda/Dockerfile.megatron @@ -60,7 +60,7 @@ WORKDIR /app COPY . /app # Install LLaMA Factory -RUN pip install --no-cache-dir -e "." --no-build-isolation +RUN pip install --no-cache-dir -e ".[metrics]" --no-build-isolation RUN pip install "git+https://github.com/alibaba/roll.git#subdirectory=mcore_adapter" diff --git a/docker/docker-cuda/docker-compose.yml b/docker/docker-cuda/docker-compose.yml index ab0da4d87..eb4250ce6 100644 --- a/docker/docker-cuda/docker-compose.yml +++ b/docker/docker-cuda/docker-compose.yml @@ -5,7 +5,6 @@ services: context: ../.. args: PIP_INDEX: https://pypi.org/simple - EXTRAS: metrics container_name: llamafactory ports: - "7860:7860" diff --git a/docker/docker-npu/Dockerfile b/docker/docker-npu/Dockerfile index 95da49715..ef4e2627c 100644 --- a/docker/docker-npu/Dockerfile +++ b/docker/docker-npu/Dockerfile @@ -37,7 +37,7 @@ RUN pip uninstall -y torch torchvision torchaudio && \ COPY . /app # Install LLaMA Factory -RUN pip install --no-cache-dir -e "." --no-build-isolation +RUN pip install --no-cache-dir -e ".[metrics,deepspeed]" --no-build-isolation # Set up volumes # VOLUME [ "/root/.cache/huggingface", "/app/shared_data", "/app/output" ] diff --git a/docker/docker-npu/docker-compose.yml b/docker/docker-npu/docker-compose.yml index 8530efafd..621d7177e 100644 --- a/docker/docker-npu/docker-compose.yml +++ b/docker/docker-npu/docker-compose.yml @@ -5,7 +5,6 @@ services: context: ../.. args: PIP_INDEX: https://pypi.org/simple - EXTRAS: torch-npu,metrics container_name: llamafactory-a2 image: llamafactory:npu-a2 volumes: diff --git a/docker/docker-rocm/Dockerfile b/docker/docker-rocm/Dockerfile index c5af6aa27..93e44e535 100644 --- a/docker/docker-rocm/Dockerfile +++ b/docker/docker-rocm/Dockerfile @@ -37,7 +37,7 @@ RUN pip uninstall -y torch torchvision torchaudio && \ COPY . /app # Install LLaMA Factory -RUN pip install --no-cache-dir -e "." --no-build-isolation +RUN pip install --no-cache-dir -e ".[metrics,deepspeed]" --no-build-isolation # Rebuild flash attention RUN if [ "${INSTALL_FLASHATTN}" == "true" ]; then \ diff --git a/docker/docker-rocm/docker-compose.yml b/docker/docker-rocm/docker-compose.yml index 32cdf5633..7e6c83bfb 100644 --- a/docker/docker-rocm/docker-compose.yml +++ b/docker/docker-rocm/docker-compose.yml @@ -5,7 +5,6 @@ services: context: ../.. args: PIP_INDEX: https://pypi.org/simple - EXTRAS: metrics container_name: llamafactory ports: - "7860:7860" diff --git a/examples/requirements/dev.txt b/examples/requirements/dev.txt deleted file mode 100644 index f0fc17fcc..000000000 --- a/examples/requirements/dev.txt +++ /dev/null @@ -1,4 +0,0 @@ -pre-commit -ruff -pytest -build diff --git a/pyproject.toml b/pyproject.toml index ec8473eb1..6cdd1b289 100644 --- a/pyproject.toml +++ b/pyproject.toml @@ -38,50 +38,48 @@ classifiers = [ ] dependencies = [ # core deps + "torch>=2.4.0", + "torchvision>=0.19.0", "transformers>=4.49.0,<=4.56.2,!=4.52.0; python_version < '3.10'", "transformers>=4.49.0,<=4.57.1,!=4.52.0,!=4.57.0; python_version >= '3.10'", "datasets>=2.16.0,<=4.0.0", "accelerate>=1.3.0,<=1.11.0", "peft>=0.14.0,<=0.17.1", "trl>=0.8.6,<=0.9.6", - "torchdata", - # torch - "torch>=2.0.0", - "torchvision>=0.15.0", + "torchdata>=0.10.0,<=0.11.0", # gui - "gradio>=4.38.0,<=5.45.0", + "gradio>=4.38.0,<=6.2.0", "matplotlib>=3.7.0", "tyro<0.9.0", # ops "einops", - "numpy<2.0.0", - "pandas>=2.0.0", + "numpy", + "pandas", "scipy", # model and tokenizer "sentencepiece", "tiktoken", - "modelscope>=1.14.0", + "modelscope", "hf-transfer", - "safetensors<=0.5.3", + "safetensors", # python "fire", "omegaconf", "packaging", "protobuf", "pyyaml", - "pydantic<=2.10.6", + "pydantic", # api "uvicorn", "fastapi", "sse-starlette", # media "av", - "librosa", - # yanked - "propcache!=0.4.0" + "librosa" ] [project.optional-dependencies] +dev = ["pre-commit", "ruff", "pytest", "build"] metrics = ["nltk", "jieba", "rouge-chinese"] deepspeed = ["deepspeed>=0.10.0,<=0.16.9"] diff --git a/src/llamafactory/data/mm_plugin.py b/src/llamafactory/data/mm_plugin.py index 291554021..44df263ab 100644 --- a/src/llamafactory/data/mm_plugin.py +++ b/src/llamafactory/data/mm_plugin.py @@ -500,13 +500,17 @@ class ErnieVLPlugin(BasePlugin): while IMAGE_PLACEHOLDER in content: image_seqlen = image_grid_thw[image_idx].prod() // merge_length if self.expand_mm_tokens else 1 content = content.replace( - IMAGE_PLACEHOLDER, f"Picture {image_idx + 1}:<|IMAGE_START|>{image_token * image_seqlen}<|IMAGE_END|>", 1 + IMAGE_PLACEHOLDER, + f"Picture {image_idx + 1}:<|IMAGE_START|>{image_token * image_seqlen}<|IMAGE_END|>", + 1, ) image_idx += 1 while VIDEO_PLACEHOLDER in content: video_seqlen = video_grid_thw[video_idx].prod() // merge_length if self.expand_mm_tokens else 1 content = content.replace( - VIDEO_PLACEHOLDER, f"Video {video_idx + 1}:<|VIDEO_START|>{video_token * video_seqlen}<|VIDEO_END|>", 1 + VIDEO_PLACEHOLDER, + f"Video {video_idx + 1}:<|VIDEO_START|>{video_token * video_seqlen}<|VIDEO_END|>", + 1, ) video_idx += 1 message["content"] = content diff --git a/src/llamafactory/extras/misc.py b/src/llamafactory/extras/misc.py index 5c4c24787..ba35bafb2 100644 --- a/src/llamafactory/extras/misc.py +++ b/src/llamafactory/extras/misc.py @@ -332,3 +332,7 @@ def fix_proxy(ipv6_enabled: bool = False) -> None: if ipv6_enabled: os.environ.pop("http_proxy", None) os.environ.pop("HTTP_PROXY", None) + os.environ.pop("https_proxy", None) + os.environ.pop("HTTPS_PROXY", None) + os.environ.pop("all_proxy", None) + os.environ.pop("ALL_PROXY", None) diff --git a/src/llamafactory/model/loader.py b/src/llamafactory/model/loader.py index 72f510a44..ef4f3f134 100644 --- a/src/llamafactory/model/loader.py +++ b/src/llamafactory/model/loader.py @@ -15,7 +15,6 @@ import os from typing import TYPE_CHECKING, Any, Optional, TypedDict -import torch from transformers import ( AutoConfig, AutoModelForCausalLM, @@ -158,6 +157,7 @@ def load_model( if model is None and not lazy_load: init_kwargs["config"] = config init_kwargs["pretrained_model_name_or_path"] = model_args.model_name_or_path + init_kwargs["torch_dtype"] = "auto" if model_args.mixture_of_depths == "load": model = load_mod_pretrained_model(**init_kwargs) diff --git a/src/llamafactory/model/patcher.py b/src/llamafactory/model/patcher.py index 7401641aa..67eee886b 100644 --- a/src/llamafactory/model/patcher.py +++ b/src/llamafactory/model/patcher.py @@ -156,16 +156,13 @@ def patch_config( # deepspeed zero3 is not compatible with low_cpu_mem_usage init_kwargs["low_cpu_mem_usage"] = model_args.low_cpu_mem_usage and (not is_deepspeed_zero3_enabled()) - # do not cast data type of the model deepspeed zero3 without qlora - if not (is_deepspeed_zero3_enabled() and model_args.quantization_bit is None): - init_kwargs["torch_dtype"] = "auto" + # fsdp/deepspeed zero3 does not need device map + if not (is_deepspeed_zero3_enabled() or is_fsdp_enabled()) and init_kwargs["low_cpu_mem_usage"]: + if "device_map" not in init_kwargs and model_args.device_map: + init_kwargs["device_map"] = model_args.device_map # device map requires low_cpu_mem_usage=True - if init_kwargs["low_cpu_mem_usage"] and not is_fsdp_enabled(): # fsdp does not need device map - if "device_map" not in init_kwargs and model_args.device_map: - init_kwargs["device_map"] = model_args.device_map # device map requires low_cpu_mem_usage=True - - if init_kwargs.get("device_map", None) == "auto": - init_kwargs["offload_folder"] = model_args.offload_folder + if init_kwargs.get("device_map", None) == "auto": + init_kwargs["offload_folder"] = model_args.offload_folder def patch_model( diff --git a/src/llamafactory/train/test_utils.py b/src/llamafactory/train/test_utils.py index 0f73d1c5e..f31b3d2fc 100644 --- a/src/llamafactory/train/test_utils.py +++ b/src/llamafactory/train/test_utils.py @@ -84,7 +84,7 @@ def load_reference_model( model: AutoModelForCausalLMWithValueHead = AutoModelForCausalLMWithValueHead.from_pretrained( model_path, torch_dtype=torch.float16, device_map="auto" ) - + return model model = AutoModelForCausalLM.from_pretrained(model_path, torch_dtype=torch.float16, device_map="auto") diff --git a/src/llamafactory/webui/locales.py b/src/llamafactory/webui/locales.py index 7051b30e8..06e64e4a7 100644 --- a/src/llamafactory/webui/locales.py +++ b/src/llamafactory/webui/locales.py @@ -35,35 +35,40 @@ LOCALES = { "value": ( "

Visit " "GitHub Page " - "Documentation

" + "Documentation " + "Blog" ), }, "ru": { "value": ( "

Посетить " "страницу GitHub " - "Документацию

" + "Документацию " + "Блог" ), }, "zh": { "value": ( "

访问 " "GitHub 主页 " - "官方文档

" + "官方文档 " + "博客" ), }, "ko": { "value": ( "

" "GitHub 페이지 " - "공식 문서를 방문하세요.

" + "공식 문서 " + "블로그를 방문하세요." ), }, "ja": { "value": ( "

" "GitHub ページ " - "ドキュメントにアクセスする

" + "ドキュメント " + "ブログにアクセスする" ), }, },