minor question on extract sequence embeddings

[Azai-yx]

Thanks for adopting the Enformer model to PyTorch.
Recently, I've needed to extract sequence embeddings from Enformer. Do I have to use one-hot encoding as input, or can I directly use sequences instead?

[jeffhsu3]

The model itself takes one-hot encodings.