< img height="1" width="1" style="display:none;" alt="" src="https://px.ads.linkedin.com/collect/?pid=3131724&fmt=gif" />
Last updated:

    Download Models

    Download Methods

    The platform supports three methods to download model files:

    Method Use Case
    HTTPS Git Clone General-purpose download for all users
    SSH Git Clone Password-free download after SSH key configuration
    csghub-cli Supports resumable downloads; ideal for large models

    Prerequisites

    Before downloading model repositories containing large files (e.g., .safetensors, .bin), install Git LFS:

    # Install Git LFS (macOS)
    brew install git-lfs
    
    # Install Git LFS (Ubuntu/Debian)
    sudo apt-get install git-lfs
    
    # Initialize Git LFS
    git lfs install
    

    Download via HTTPS

    # Clone the model repository (LFS files are downloaded automatically)
    git lfs install
    git clone https://<platform-host>/<namespace>/<model-name>
    

    For private models, use an access token as credentials:

    git clone https://<username>:<access-token>@<platform-host>/<namespace>/<model-name>
    

    To skip large file downloads and only get the repository structure:

    GIT_LFS_SKIP_SMUDGE=1 git clone https://<platform-host>/<namespace>/<model-name>
    

    Download via SSH

    After adding your SSH public key in User Settings → SSH Keys:

    git lfs install
    git clone ssh://git@<platform-host>/<namespace>/<model-name>
    

    Download via csghub-cli

    Install csghub-cli:

    pip install csghub-sdk
    

    Download a model:

    # Download the entire model repository
    csghub-cli download <namespace>/<model-name> --repo_type model
    
    # Download a specific branch or revision
    csghub-cli download <namespace>/<model-name> --repo_type model --revision main
    

    Download via Python SDK

    from pycsghub.snapshot_download import snapshot_download
    
    model_path = snapshot_download(
        repo_id="<namespace>/<model-name>",
        repo_type="model",
        endpoint="https://<platform-host>",
        token="<access-token>"  # Required for private models
    )
    print(f"Model downloaded to: {model_path}")
    

    Note

    Access tokens can be generated in User Settings → Access Tokens.