ComfyUI Extension: Hugging Face LoRA Uploader

Authored by vioaki

Created

Updated

0 stars

Run ComfyUI workflows without the setup

No installs, no CUDA version roulette, no GPU sitting idle on your bill. Bring a workflow and run it in the browser.

A ComfyUI custom node to upload LoRA models to Hugging Face Hub.

Looking for a different extension?

Custom Nodes (2)

README

ComfyUI LoRA Uploaders

Upload your LoRA models to Hugging Face Hub and ModelScope directly from ComfyUI.

Version: v0.2.0

Quick Start

Installation

  1. Clone into ComfyUI's custom nodes directory:
cd ComfyUI/custom_nodes/
git clone https://github.com/vioaki/ComfyUI-HuggingFaceLoraUploader.git
  1. Install dependencies:
cd ComfyUI-HuggingFaceLoraUploader
pip install -r requirements.txt
  1. Restart ComfyUI

Get Your Access Token

  • Hugging Face: Settings → Access Tokens (requires write permission)
  • ModelScope: Personal Center → Access Tokens

Usage

Upload to Hugging Face

  1. Add node: Right-click → Add Node → Uploaders/HuggingFace → Hugging Face LoRA Uploader
  2. Fill in required fields:
    • lora_name: Select your local LoRA file
    • hf_token: Your Hugging Face access token
    • repo_id: Target repository (format: username/repo_name)
    • commit_message: Describe this upload
  3. Run workflow

Upload to ModelScope

  1. Add node: Right-click → Add Node → Uploaders/ModelScope → ModelScope LoRA Uploader (HTTP)
  2. Fill in required fields:
    • lora_name: Select your local LoRA file
    • modelscope_token: Your ModelScope access token
    • repo_id: Target repository (format: username/model_name)
    • commit_message: Describe this upload
    • visibility_str: public or private
  3. Run workflow

Key Features

  • Auto-Detection: Automatically finds LoRA files in your ComfyUI directories
  • Auto-Create Repos: Creates repositories if they don't exist
  • Custom Paths: Upload to specific folders within repositories
  • Status Feedback: See upload results directly in ComfyUI

Important Notes

Security Warning

Access tokens are saved in plain text in your workflow JSON files. Do not share workflow files containing tokens publicly.

Optional Settings

Both uploaders support:

  • path_in_repo: Upload to a subfolder (e.g., loras/)
  • create_repo_if_not_exists: Auto-create repository (default: enabled)
  • private_repo (HF) / visibility_str (ModelScope): Set repository privacy

ModelScope additional options:

  • chinese_name: Chinese display name for your model
  • license_str: Choose a license (Apache 2.0, MIT, etc.)
  • revision: Target branch (default: master)

Troubleshooting

Nodes not appearing?

  • Verify huggingface-hub and modelscope are installed
  • Check ComfyUI console for errors

Upload failed?

  • Confirm your token has write permissions
  • Check repository ID format
  • Verify internet connection
  • See ComfyUI console for detailed error messages

Large files taking long?

  • This is normal - be patient during upload

Technical Details

  • ModelScope v0.2.0 uses HTTP API (no Git configuration needed)
  • Automatically generates configuration.json and README.md for ModelScope uploads
  • Cleans up temporary files after upload

Run ComfyUI workflows without the setup

No installs, no CUDA version roulette, no GPU sitting idle on your bill. Bring a workflow and run it in the browser.

Learn more