Supported Frameworks

Below are the supported frameworks for models that can be uploaded into Inferless.

  • Pytorch

  • Tensorflow

  • ONNX (This is currently not supported)

Any other framework is not currently supported as of now. We will keep adding more frameworks as the support for them is added to our platform

File structure requirements

In the case of model in Sagemaker being:

- PyTorch

  • The model file needs to have a model.pt file

  • By default, AWS Sagemaker creates it as model.pth.

To make sure that the model is saved as a model.pt, please follow the below instructions.

  • The Training script used in sagemaker typically saves the model using torch.save
commandpath = os.path.join(model_dir, "model.pth")
torch.save(model.cpu().state_dict(), path)
  • But for Inferless pytorch runtime we need model.pt file which can be created by modifying the training script with the following instead of the above
code.path = os.path.join(model_dir, "model.pt")
torch.jit.save(model, path)
  • Once this is done the estimator.fit function trains the model and saves the trained model in model.pt format.

  • This can then be uploaded to s3 using the estimator.deploy function

  • You can now upload the Model by entering the ARN link to inferless using the model import option.

You can view a example below on how we stored a file as per the steps above:

<Video here>

- Tensorflow

  • There is no need for any additional steps/file modifications. The ARN link of the generated model can directly be used for import