Model Hub

Learn how to upload, version, and manage model repositories on the platform.

What is a Model Repository?

A model repository is a container for your machine learning models, including:

Model weights - The trained parameters of your model
Configuration files - Model architecture and hyperparameters
Tokenizer files - For NLP models
README - Documentation and usage instructions
Metadata - Tags, tasks, frameworks, and licensing info

Creating a Model

Create a new model repository through the UI or API:

Via API:

POST /api/models
Authorization: Bearer <your-token>
Content-Type: application/json

{
  "name": "bert-finetuned-sentiment",
  "namespace": "username",
  "visibility": "public",
  "task": "text-classification",
  "framework": "transformers",
  "tags": ["bert", "sentiment-analysis", "nlp"],
  "license": "apache-2.0",
  "description": "BERT model fine-tuned on sentiment analysis"
}

Naming conventions:

Use lowercase letters, numbers, hyphens, and underscores
No spaces or special characters
Be descriptive: include the base model and fine-tuning task
Example: llama-3-8b-instruct

Model Visibility

Public

Anyone can view and download your model. Perfect for open-source contributions.

Private

Only you and your organization members can access. Requires Pro or Team plan.

Gated (Coming Soon)

Users must request access before downloading. Useful for compliance requirements.

Uploading Model Files

Upload your model files to the repository:

Common file structure:

model-name/
├── README.md
├── config.json
├── pytorch_model.bin
├── tokenizer.json
├── tokenizer_config.json
└── special_tokens_map.json

Upload methods:

Web interface - drag and drop files
Git LFS - for version control
API upload - programmatic file management
CLI tool - command-line uploads (coming soon)

Searching & Filtering Models

Find models using various filters:

Search API:

GET /api/models?task=text-classification&framework=transformers&sort=downloads

Query parameters:
- search: text search in names and descriptions
- task: filter by task type
- framework: filter by framework
- tags: comma-separated list of tags
- sort: downloads, likes, created, updated
- limit: results per page (default: 20)
- offset: pagination offset

Model Metadata & Tags

Rich metadata helps users discover and understand your model:

Tasks:

text-generationtext-classificationimage-classificationobject-detectionspeech-recognition

Frameworks:

transformerspytorchtensorflowjaxonnx

Updating Models

Update model metadata or upload new versions:

PATCH /api/models/:namespace/:name
Authorization: Bearer <your-token>
Content-Type: application/json

{
  "description": "Updated description",
  "tags": ["transformer", "nlp", "updated"],
  "visibility": "private"
}

Deleting Models

⚠️ Permanent Action

Deleting a model is permanent and cannot be undone. All files and metadata will be removed.

DELETE /api/models/:namespace/:name
Authorization: Bearer <your-token>

Best Practices

✅Write comprehensive READMEs - Include usage examples, training details, limitations
✅Tag appropriately - Use relevant tags to improve discoverability
✅Include licenses - Always specify the license for your model
✅Version your models - Use clear versioning for model updates
✅Document limitations - Be transparent about what your model can and can't do