Want us to help migrate your models?

You can book a 60 min call with our engineering team, and we’ll get your models running by the end of the call.

Porting a model generally takes 10-20 minutes of refactoring. At a high level, the steps are:

  1. Create an empty sandbox
  2. Install Python Requirements
  3. (Optional) Upload pre-trained weights to the data folder (if you want to avoiding connecting your dataset and retraining the model from scratch)
  4. Copy the model’s code into the sandbox, returning your model class from the train.py function
  5. Train the model
  6. Create your model handler
  7. Package and deploy

​
Example: Basic Scikit-Learn Model

Connecting to Google BigQueryUsing Pre-Trained Models