Hi,
Thanks for your excellent work. I have a question about how to control the protein length. As stated in the paper, "Each sequence is represented by its amino acids followed by an end-of-sequence (EOS) token. All other tokens after EOS are PAD tokens, which are treated as normal tokens for the purposes of noisy observations, predictions and loss."
Does this mean that during the training process, you padded all sequences to the same length, such as 256 for antibodies? Additionally, do you need the model to learn to generate EOS and PAD tokens, and ensure that during sampling, PAD tokens only appear after the EOS token in the sequences?
Best regards.
Hi,
Thanks for your excellent work. I have a question about how to control the protein length. As stated in the paper, "Each sequence is represented by its amino acids followed by an end-of-sequence (EOS) token. All other tokens after EOS are PAD tokens, which are treated as normal tokens for the purposes of noisy observations, predictions and loss."
Does this mean that during the training process, you padded all sequences to the same length, such as 256 for antibodies? Additionally, do you need the model to learn to generate EOS and PAD tokens, and ensure that during sampling, PAD tokens only appear after the EOS token in the sequences?
Best regards.