423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-08-23 06:12:50 +08:00
Code Issues Packages Projects Releases Wiki Activity

1. add modelers hub support

Browse Source 
Former-commit-id: 24ebe187e360753666b768685a0dcc78054bb702
This commit is contained in:
huniu20 2024-10-09 17:21:37 +08:00
parent 4464a6ff5b
commit 26e897e861
7 changed files with 92 additions and 21 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
README.md
View File

@ -75,6 +75,8 @@ Compared to ChatGLM's [P-Tuning](https://github.com/THUDM/ChatGLM2-6B/tree/main/
## Changelog
[24/10/09] We supported downloading pre-trained models and datasets from the **[Modelers Hub](https://modelers.cn/models)** for Chinese mainland users. See [this tutorial](#download-from-modelers-hub) for usage.
[24/09/19] We support fine-tuning the **[Qwen2.5](https://qwenlm.github.io/blog/qwen2.5/)** models.
[24/08/30] We support fine-tuning the **[Qwen2-VL](https://qwenlm.github.io/blog/qwen2-vl/)** models. Thank [@simonJJJ](https://github.com/simonJJJ)'s PR.
@ -363,7 +365,7 @@ cd LLaMA-Factory
pip install -e ".[torch,metrics]"
```
Extra dependencies available: torch, torch-npu, metrics, deepspeed, liger-kernel, bitsandbytes, hqq, eetq, gptq, awq, aqlm, vllm, galore, badam, adam-mini, qwen, modelscope, quality
Extra dependencies available: torch, torch-npu, metrics, deepspeed, liger-kernel, bitsandbytes, hqq, eetq, gptq, awq, aqlm, vllm, galore, badam, adam-mini, qwen, modelscope, quality, openmind
> [!TIP]
> Use `pip install --no-deps -e .` to resolve package conflicts.
@ -415,7 +417,7 @@ Download the pre-built Docker images: [32GB](http://mirrors.cn-central-221.ovaij
### Data Preparation
Please refer to [data/README.md](data/README.md) for checking the details about the format of dataset files. You can either use datasets on HuggingFace / ModelScope hub or load the dataset in local disk.
Please refer to [data/README.md](data/README.md) for checking the details about the format of dataset files. You can use datasets on HuggingFace hub, ModelScope hub, modelers hub or load the dataset in local disk.
> [!NOTE]
> Please update `data/dataset_info.json` to use your custom dataset.
@ -483,6 +485,7 @@ docker build -f ./docker/docker-cuda/Dockerfile \
docker run -dit --gpus=all \
-v ./hf_cache:/root/.cache/huggingface \
-v ./ms_cache:/root/.cache/modelscope \
-v ./om_cache:/root/.cache/openmind \
-v ./data:/app/data \
-v ./output:/app/output \
-p 7860:7860 \
@ -507,6 +510,7 @@ docker build -f ./docker/docker-npu/Dockerfile \
docker run -dit \
-v ./hf_cache:/root/.cache/huggingface \
-v ./ms_cache:/root/.cache/modelscope \
-v ./om_cache:/root/.cache/openmind \
-v ./data:/app/data \
-v ./output:/app/output \
-v /usr/local/dcmi:/usr/local/dcmi \
@ -540,6 +544,7 @@ docker build -f ./docker/docker-rocm/Dockerfile \
docker run -dit \
-v ./hf_cache:/root/.cache/huggingface \
-v ./ms_cache:/root/.cache/modelscope \
-v ./om_cache:/root/.cache/openmind \
-v ./data:/app/data \
-v ./output:/app/output \
-v ./saves:/app/saves \
@ -560,6 +565,7 @@ docker exec -it llamafactory bash
- `hf_cache`: Utilize Hugging Face cache on the host machine. Reassignable if a cache already exists in a different directory.
- `ms_cache`: Similar to Hugging Face cache but for ModelScope users.
- `om_cache`: Similar to Hugging Face cache but for Modelers users.
- `data`: Place datasets on this dir of the host machine so that they can be selected on LLaMA Board GUI.
- `output`: Set export dir to this location so that the merged result can be accessed directly on the host machine.
@ -584,6 +590,16 @@ export USE_MODELSCOPE_HUB=1 # `set USE_MODELSCOPE_HUB=1` for Windows
Train the model by specifying a model ID of the ModelScope Hub as the `model_name_or_path`. You can find a full list of model IDs at [ModelScope Hub](https://modelscope.cn/models), e.g., `LLM-Research/Meta-Llama-3-8B-Instruct`.
### Download from Modelers Hub
You can also use Modelers Hub to download models and datasets.
```bash
export USE_OPENMIND_HUB=1 # `set USE_OPENMIND_HUB=1` for Windows
```
Train the model by specifying a model ID of the Modelers Hub as the `model_name_or_path`. You can find a full list of model IDs at [Modelers Hub](https://modelers.cn/models), e.g., `TeleAI/TeleChat-7B-pt`.
### Use W&B Logger
To use [Weights & Biases](https://wandb.ai) for logging experimental results, you need to add the following arguments to yaml files.

20
README_zh.md
View File

@ -75,6 +75,7 @@ https://github.com/user-attachments/assets/e6ce34b0-52d5-4f3e-a830-592106c4c272
</details>
## 更新日志
[24/10/09] 我们支持了从 **[魔乐社区](https://modelers.cn/models)** 下载预训练模型和数据集。详细用法请参照 [此教程](#从魔乐社区下载)。
[24/09/19] 我们支持了 **[Qwen2.5](https://qwenlm.github.io/blog/qwen2.5/)** 模型的微调。
@ -364,7 +365,7 @@ cd LLaMA-Factory
pip install -e ".[torch,metrics]"
```
可选的额外依赖项：torch、torch-npu、metrics、deepspeed、liger-kernel、bitsandbytes、hqq、eetq、gptq、awq、aqlm、vllm、galore、badam、adam-mini、qwen、modelscope、quality
可选的额外依赖项：torch、torch-npu、metrics、deepspeed、liger-kernel、bitsandbytes、hqq、eetq、gptq、awq、aqlm、vllm、galore、badam、adam-mini、qwen、modelscope、quality、openmind
> [!TIP]
> 遇到包冲突时，可使用 `pip install --no-deps -e .` 解决。
@ -416,7 +417,7 @@ source /usr/local/Ascend/ascend-toolkit/set_env.sh
### 数据准备
关于数据集文件的格式，请参考 [data/README_zh.md](data/README_zh.md) 的内容。你可以使用 HuggingFace / ModelScope 上的数据集或加载本地数据集。
关于数据集文件的格式，请参考 [data/README_zh.md](data/README_zh.md) 的内容。你可以使用 HuggingFace , ModelScope 或者 Modelers 上的数据集或加载本地数据集。
> [!NOTE]
> 使用自定义数据集时，请更新 `data/dataset_info.json` 文件。
@ -484,6 +485,7 @@ docker build -f ./docker/docker-cuda/Dockerfile \
docker run -dit --gpus=all \
-v ./hf_cache:/root/.cache/huggingface \
-v ./ms_cache:/root/.cache/modelscope \
-v ./om_cache:/root/.cache/openmind \
-v ./data:/app/data \
-v ./output:/app/output \
-p 7860:7860 \
@ -508,6 +510,7 @@ docker build -f ./docker/docker-npu/Dockerfile \
docker run -dit \
-v ./hf_cache:/root/.cache/huggingface \
-v ./ms_cache:/root/.cache/modelscope \
-v ./om_cache:/root/.cache/openmind \
-v ./data:/app/data \
-v ./output:/app/output \
-v /usr/local/dcmi:/usr/local/dcmi \
@ -541,6 +544,7 @@ docker build -f ./docker/docker-rocm/Dockerfile \
docker run -dit \
-v ./hf_cache:/root/.cache/huggingface \
-v ./ms_cache:/root/.cache/modelscope \
-v ./om_cache:/root/.cache/openmind \
-v ./data:/app/data \
-v ./output:/app/output \
-v ./saves:/app/saves \
@ -561,6 +565,7 @@ docker exec -it llamafactory bash
- `hf_cache`：使用宿主机的 Hugging Face 缓存文件夹，允许更改为新的目录。
- `ms_cache`：类似 Hugging Face 缓存文件夹，为 ModelScope 用户提供。
- `om_cache`：类似 Hugging Face 缓存文件夹，为 Modelers 用户提供。
- `data`：宿主机中存放数据集的文件夹路径。
- `output`：将导出目录设置为该路径后，即可在宿主机中访问导出后的模型。
@ -585,6 +590,17 @@ export USE_MODELSCOPE_HUB=1 # Windows 使用 `set USE_MODELSCOPE_HUB=1`
将 `model_name_or_path` 设置为模型 ID 来加载对应的模型。在[魔搭社区](https://modelscope.cn/models)查看所有可用的模型，例如 `LLM-Research/Meta-Llama-3-8B-Instruct`。
### 从魔乐社区下载
您也可以通过下述方法使用魔乐社区，在魔乐社区上下载数据集和模型。
```bash
export USE_OPENMIND_HUB=1 # Windows 使用 `set USE_OPENMIND_HUB=1`
```
将 `model_name_or_path` 设置为模型 ID 来加载对应的模型。在[魔乐社区](https://modelers.cn/models)查看所有可用的模型，例如 `TeleAI/TeleChat-7B-pt`。
### 使用 W&B 面板
若要使用 [Weights & Biases](https://wandb.ai) 记录实验数据，请在 yaml 文件中添加下面的参数。

1
src/llamafactory/__init__.py
View File

@ -38,6 +38,7 @@ Force check imports: FORCE_CHECK_IMPORTS=1
Force using torchrun: FORCE_TORCHRUN=1
Set logging verbosity: LLAMAFACTORY_VERBOSITY=WARN
Use modelscope: USE_MODELSCOPE_HUB=1
Use openmind: USE_OPENMIND_HUB=1
"""
from .extras.env import VERSION

21
src/llamafactory/data/loader.py
View File

@ -53,7 +53,7 @@ def _load_single_dataset(
"""
logger.info("Loading dataset {}...".format(dataset_attr))
data_path, data_name, data_dir, data_files = None, None, None, None
if dataset_attr.load_from in ["hf_hub", "ms_hub"]:
if dataset_attr.load_from in ["om_hub", "hf_hub", "ms_hub"]:
data_path = dataset_attr.dataset_name
data_name = dataset_attr.subset
data_dir = dataset_attr.folder
@ -84,7 +84,24 @@ def _load_single_dataset(
else:
raise NotImplementedError("Unknown load type: {}.".format(dataset_attr.load_from))
if dataset_attr.load_from == "ms_hub":
if dataset_attr.load_from == "om_hub":
try:
from openmind import OmDataset
from openmind.utils.hub import OM_DATASETS_CACHE
cache_dir = model_args.cache_dir or OM_DATASETS_CACHE
dataset = OmDataset.load_dataset(
path=data_path,
name=data_name,
data_dir=data_dir,
data_files=data_files,
split=data_args.split,
cache_dir=cache_dir,
token=model_args.ms_hub_token,
streaming=(data_args.streaming and (dataset_attr.load_from != "file")),
)
except ImportError:
raise ImportError("Please install openmind via `pip install openmind -U`")
elif dataset_attr.load_from == "ms_hub":
require_version("modelscope>=1.11.0", "To fix: pip install modelscope>=1.11.0")
from modelscope import MsDataset
from modelscope.utils.config_ds import MS_DATASETS_CACHE

18
src/llamafactory/data/parser.py
View File

@ -20,7 +20,7 @@ from typing import Any, Dict, List, Literal, Optional, Sequence
from transformers.utils import cached_file
from ..extras.constants import DATA_CONFIG
from ..extras.misc import use_modelscope
from ..extras.misc import use_openmind, use_modelscope
@dataclass
@ -97,8 +97,13 @@ def get_dataset_list(dataset_names: Optional[Sequence[str]], dataset_dir: str) -
dataset_list: List["DatasetAttr"] = []
for name in dataset_names:
if dataset_info is None: # dataset_dir is ONLINE
load_from = "ms_hub" if use_modelscope() else "hf_hub"
if dataset_info is None: # dataset_dir is ONLINE
if use_openmind():
load_from = "om_hub"
if use_modelscope():
load_from = "ms_hub"
if load_from is None:
load_from = "hf_hub"
dataset_attr = DatasetAttr(load_from, dataset_name=name)
dataset_list.append(dataset_attr)
continue
@ -106,11 +111,14 @@ def get_dataset_list(dataset_names: Optional[Sequence[str]], dataset_dir: str) -
if name not in dataset_info:
raise ValueError("Undefined dataset {} in {}.".format(name, DATA_CONFIG))
has_om_url = "om_hub_url" in dataset_info[name]
has_hf_url = "hf_hub_url" in dataset_info[name]
has_ms_url = "ms_hub_url" in dataset_info[name]
if has_hf_url or has_ms_url:
if (use_modelscope() and has_ms_url) or (not has_hf_url):
if has_om_url or has_hf_url or has_ms_url:
if (use_openmind() and has_om_url) or (not has_hf_url):
dataset_attr = DatasetAttr("om_hub", dataset_name=dataset_info[name]["om_hub_url"])
elif (use_modelscope() and has_ms_url) or (not has_hf_url):
dataset_attr = DatasetAttr("ms_hub", dataset_name=dataset_info[name]["ms_hub_url"])
else:
dataset_attr = DatasetAttr("hf_hub", dataset_name=dataset_info[name]["hf_hub_url"])

29
src/llamafactory/extras/misc.py
View File

@ -231,18 +231,31 @@ def torch_gc() -> None:
torch.cuda.empty_cache()
def try_download_model_from_ms(model_args: "ModelArguments") -> str:
if not use_modelscope() or os.path.exists(model_args.model_name_or_path):
def try_download_model_from_other_hub(model_args: "ModelArguments") -> str:
if (not use_openmind() and not use_modelscope()) or os.path.exists(model_args.model_name_or_path):
return model_args.model_name_or_path
try:
from modelscope import snapshot_download
if use_openmind():
try:
import openmind
from openmind.utils.hub import snapshot_download
revision = "master" if model_args.model_revision == "main" else model_args.model_revision
return snapshot_download(model_args.model_name_or_path, revision=revision, cache_dir=model_args.cache_dir)
except ImportError:
raise ImportError("Please install modelscope via `pip install modelscope -U`")
revision = "main" if model_args.model_revision == "main" else model_args.model_revision
return snapshot_download(model_args.model_name_or_path, revision=revision, cache_dir=model_args.cache_dir)
except ImportError:
raise ImportError("Please install openmind and openmind_hub via `pip install openmind -U`")
if use_modelscope():
try:
from modelscope import snapshot_download
revision = "master" if model_args.model_revision == "main" else model_args.model_revision
return snapshot_download(model_args.model_name_or_path, revision=revision, cache_dir=model_args.cache_dir)
except ImportError:
raise ImportError("Please install modelscope via `pip install modelscope -U`")
def use_openmind() -> bool:
return os.environ.get("USE_OPENMIND_HUB", "0").lower() in ["true", "1"]
def use_modelscope() -> bool:
return os.environ.get("USE_MODELSCOPE_HUB", "0").lower() in ["true", "1"]

4
src/llamafactory/model/loader.py
View File

@ -19,7 +19,7 @@ from transformers import AutoConfig, AutoModelForCausalLM, AutoModelForVision2Se
from trl import AutoModelForCausalLMWithValueHead
from ..extras.logging import get_logger
from ..extras.misc import count_parameters, skip_check_imports, try_download_model_from_ms
from ..extras.misc import count_parameters, skip_check_imports, try_download_model_from_other_hub
from .adapter import init_adapter
from .model_utils.liger_kernel import apply_liger_kernel
from .model_utils.misc import register_autoclass
@ -50,7 +50,7 @@ def _get_init_kwargs(model_args: "ModelArguments") -> Dict[str, Any]:
Note: including inplace operation of model_args.
"""
skip_check_imports()
model_args.model_name_or_path = try_download_model_from_ms(model_args)
model_args.model_name_or_path = try_download_model_from_other_hub(model_args)
return {
"trust_remote_code": True,
"cache_dir": model_args.cache_dir,