Releases: alchemab/parapred-pytorch
Releases · alchemab/parapred-pytorch
v1.0.2
v1.0.1
Updated docs for running Parapred on a batch of sequences. This is due to PyTorch's pack_padded_sequence
function preferring length-sorted sequences.
The model's forward
call also has an update to expect length-sorted sequences.
v1.0.0
PyTorch implementation of Parapred with features such as:
- Use of sigmoid activation as opposed to hard sigmoid (default)
- JSON outputs
- Paratyping [Richardson et al., 2021, mAbs]
At this moment, we have no plans to include code for training Parapred from scratch.