Model Hub

The platform provides users with the option to upload a Hugging Face models and use them when generating NLP trials.

Important Points:

Model Hub can only be used for NLP sequence classification and regression models.

How to use

1. Choose a Model from Hugging Face

Suppose we want to use the model albert/albert-base-v1 from Hugging Face.

• Download the model to our file system, using the following command:

  git clone https://huggingface.co/albert/albert-base-v2

2. Compress the Model Folder

Once the model is downloaded, compress the entire folder into a .zip file.

3. Upload the Model to EvoML

1. Navigate to the Models Hub tab in evoML.

2. Click on New Model.

   

3. Fill in the required fields:

   • Namespace (optional)
   • Name
   • Upload the model .zip file

   

4. Once uploaded, the model appears in the Models Hub.

4. Use the Model in a Trial

Our uploaded model now has a unique name: my_namespace/albert_base_1.

To use it in a new trial:

1. Set up the trial configuration.
2. Navigate to the Models tab.
3. Follow these steps:
   • Select NLP Sequence Classifier.
   • Update to a Fixed model.
   • Modify the Values field to include my_namespace/albert_base_1.
   • Ensure the model is selected in the Fixed Value dropdown.

5. Start the Trial and Generate the Pipeline

• Start the new trial.
• Once the trial completes, generate the pipeline, and deploy.