Model Hub

Overview

The Model Hub page serves as a central repository where users can browse, import, and manage AI models. It features sections for both Public Models (available to all users) and Your Models (models uploaded or customized by the user). The interface allows for easy exploration and interaction with the latest AI models.

Public Models

Public models are the Segmind offered models, available to all the users. The user can select a model of their choice and it will directly take them to the model inference page.

Your Models

Your models page shows the list of models that the user has imported from external sources. These models remain private and are only visible to the specific user.

The user can import any Model by clicking on the "Import Model" button on the top right.

Import Model

The Import Model page allows users to import a LORA (Low-Rank Adaptation) model from Hugging Face or provide a direct file upload. These models will be deployed on the platform's GPUs. Currently, only FLUX.Dev, FLUX Schnell, SDXL, SD2.1, and SD1.5 Lora models are supported.

1. Base Model Dropdown

   • A dropdown menu where users select the base model on which the imported LORA model will be built.

   • Example option: FLUX DEV.

   • This field is mandatory.

2. Safetensor File Upload

   • Users can upload a model file directly in supported formats such as .safetensors, .sft, or .ckpt.

   • This field is mandatory.

3. OR - File URL

   • An alternative to uploading a file is entering the URL of the model’s safetensor file.

   • The field accepts the URL where the model can be fetched from.

   • This field is mandatory if the file is not uploaded directly.

4. Model Information Fields

   • Name of the Model: Users must provide a name for the imported model. This is a required field.

   • Default Prompt: A text field where the default prompt for the model is defined. This is also required for proper model usage.

5. Model Description

   • A description box where users provide details about the model.

   • The input should be in Markdown format, allowing for structured documentation.

   • Example markdown structure could include:

     • Model Name: Brief title or heading for the model.

     • Description: Details about the model and its capabilities.

     • Use Cases: List of potential uses for the model.

   • Though optional, providing a model description enhances clarity and usage understanding.

6. Confirm Button

   • Once all mandatory fields are filled in, the user can press the Confirm button to finalize the import process.

Once the import process is finished, the status will be shown as "Available" and the user can click on the Segmind logo button in the "Actions" field to take them directly to the inference page.