LLM: LoRA

1. peft库

1.1 环境安装

conda create -n peft python=3.10
conda activate peft

# 安装相应的包
pip install git+https://github.com/huggingface/transformers
pip install git+https://github.com/huggingface/accelerate
pip install git+https://github.com/huggingface/peft
conda install pytorch==2.1.0 torchvision==0.16.0 torchaudio==2.1.0 pytorch-cuda=11.8 -c pytorch -c nvidia

LoRA代码运行

bash create_datasets.sh

2. huggingface下载模型和数据方法一览

2.1 模型下载

huggingface官网链接：https://huggingface.co/docs/huggingface_hub/v0.19.3/guides/download

下载一整个仓库

from huggingface_hub import snapshot_download

repo_id = "gpt2-medium"  # 模型在huggingface上的名称
local_dir = "/data/wangyh/mllms/LoRA-main/examples/NLG/pretrained_checkpoints/"  # 本地模型存储的地址
local_dir_use_symlinks = False  # 本地模型使用文件保存，而非blob形式保存
# token = "XXX"  # 在hugging face上生成的 access token

# # 如果需要代理的话
# proxies = {
#     'http': 'XXXX',
#     'https': 'XXXX',
# }

snapshot_download(
    repo_id=repo_id,
    local_dir=local_dir,
    local_dir_use_symlinks=local_dir_use_symlinks,
    # token=token,
    # proxies=proxies
)

下载仓库中的某个文件

from huggingface_hub import hf_hub_download

repo_id = "gpt2-medium"  # 模型在huggingface上的名称
file_name = "pytorch_model.bin"
local_dir = "/data/wangyh/mllms/LoRA-main/examples/NLG/pretrained_checkpoints_2/"  # 本地模型存储的地址
local_dir_use_symlinks = False  # 本地模型使用文件保存，而非blob形式保存
# token = "XXX"  # 在hugging face上生成的 access token

# # 如果需要代理的话
# proxies = {
#     'http': 'XXXX',
#     'https': 'XXXX',
# }

hf_hub_download(
    repo_id=repo_id,
    filename=file_name,
    local_dir=local_dir,
    local_dir_use_symlinks=local_dir_use_symlinks,
    # token=token,
    # proxies=proxies
)