Adding Models and LoRAs to Invoke

Invoke provides a set of default models and LoRAs to all users that are popular and effective to create with. However, sometimes a specific fine-tuned model or LoRA is required to get precisely the desired result. 

This article will provide details on how to add Models and LoRAs to Invoke. 

Note: Model uploading requires the Owner or Account Admin role on the Premier or Enterprise plans.

Where to find models for upload

If you are looking for new models to download that aren’t part of our default Starter Models, we suggest browsing https://huggingface.co/models.

Steps to Upload a Model - Professional Edition

You can upload models directly from your hard drive or from a HuggingFace repository. Follow the appropriate steps below for your upload method.

Uploading a Model Directly from Your Hard Drive

• Ensure that your model is saved locally to your computer:
  • Models in safetensors format should be a single .safetensors file.
  • Models in Diffusers format should be saved in a folder containing all necessary information and files.
• Navigate to the Model Upload page (from the studio, click the Invoke icon and select Manage Models -> Manage Account Models -> Model Uploads).
• Drag your model file or folder into the indicated space on the Upload New Model form.
• Double-check the Model Type and Base Model Type are correctly selected. Incorrect details may cause errors during uploading.
  • In unsure, refer to the model's source documentation for guidance.
  • Avoid navigating away or refreshing the page while your model is uploading.
• Once uploaded, our system will process the model. This typically takes 10-15 minutes.
• After processing is complete, the model will be available for use in your projects. For users with multiple projects, you may have to add the model manually to your project on the Model Management screen.

Uploading a Model from a HuggingFace Repository

• Navigate to the HuggingFace model card page of your desired model.
• Note the Repo Name (found at the top-left corner of the model card).
  • In the example below, the repo id would be XLabs-AI/flux-ip-adapter
• If the repository contains multiple files, ensure you specify which file you want to upload using the following format: REPO ID::filename. For example, in the Repo below, you would enter XLabs-AI/flux-ip-adapter::ip_adapter.safetensors
• Navigate to the Model Upload page (from the studio, click the Invoke icon and select Manage Models -> Manage Account Models -> Model Uploads).
• Select "Upload from Hugging Face" on the model upload form. Enter the Repo ID or Repo ID and filename following the formatting above. You may also need to provide a token.
• Double-check the Model Type and Base Model Type are correctly selected. Incorrect details may cause errors during uploading.
  • In unsure, refer to the model's source documentation for guidance.
  • Avoid navigating away or refreshing the page while your model is uploading.
• Click Upload Model. 
  • To upload a model file, ensure that you have the model saved locally to your computer as either a .safetensors file, or if in the Diffusers format, in a folder with all necessary information.
  • Models in safetensors format will be a single file with a .safetensors file extension and diffusers format models are made up of a collection of folders that contain all of the model's files and information.

Steps to Upload a Model - Community Edition

1. You can install model files directly, install a model from a Hugging Face repo, select models from a local folder, or select a model from Invoke’s preloaded Starter Models
   • To install a model file, ensure that you have the model saved locally to your computer as either a .safetensors file, or if in the Diffusers format, in a folder with all necessary information.
   • Models in safetensors format will be a single file with a .safetensors file extension and diffusers format models are made up of a collection of folders that contain all of the model's files and information.
2. Navigate to the ‘Add Model’ section of the Model Manager tab
3. Select the model for installation
   1. For direct installation - Enter URL or path to a local file or folder for a single diffusers model. Ensure that all of the files uploaded are correct.
   2. For installation from Hugging Face - Enter the repo ID of the desired model.
   3. For installation from Scan Folder - Enter path to the local folder and scan folder.
   4. For Starter Models - Install the desired model to your project using the ‘+’ icon.
4. Fill in model details
   • Important note: Double check that you have chosen the correct Model Type and Base Model, as incorrect details will cause errors when attempting to use the model in the creation process.
   • If you're unsure of the Model Type or Base Model, reference the model's source.
5. Wait for the model to be installed - ensure you do not navigate away or refresh your page
6. The model will be processed then ready for use!

Contacting Support

For further assistance with uploaded models, or if you encounter any issues, please contact our support team. You can reach out to us through the 'Help' or 'Support' section in the Studio, or email us directly at [email protected]