Olah is self-hosted lightweight huggingface mirror service. Olah means hello in Hilichurlian.

Other languages: 中文

• Models mirror
• Datasets mirror
• Spaces mirror

pip install olah

or:

pip install git+https://github.com/vtuber-plan/olah.git 

1. Clone this repository

git clone https://github.com/vtuber-plan/olah.git
cd olah

2. Install the Package

pip install --upgrade pip
pip install -e .

Run the command in the console:

python -m olah.server

Then set the Environment Variable HF_ENDPOINT to the mirror site (Here is http://localhost:8090).

export HF_ENDPOINT=http://localhost:8090

Starting from now on, all download operations in the HuggingFace library will be proxied through this mirror site.

from huggingface_hub import snapshot_download

snapshot_download(repo_id='Qwen/Qwen-7B', repo_type='model',
                  local_dir='./model_dir', resume_download=True,
                  max_workers=8)

You can check the path ./repos which stores all cached datasets and models.

Run the command in the console:

python -m olah.server

Or you can specify the host address and listening port:

python -m olah.server --host localhost --port 8090

Please remember to change the --mirror-url and --mirror-lfs-url to the actual URLs of the mirror site while modifying the host and port.

The default mirror cache path is ./repos, you can change it by --repos-path parameter:

python -m olah.server --host localhost --port 8090 --repos-path ./hf_mirrors

• Authentication
• Administrator and user system
• OOS backend support
• Mirror Update Schedule Task

olah is released under the MIT License.

• olah-docs
• olah-source