Inference improvements #130
Description
On inference time, we are currently not batching. So there is room for improvement for speeding up the DeepBLAST.align method in deepblast/trainer.py. I anticipate that we can get a 10-20x improvement in speed. Batching is already being done on training these models, so it is a matter of porting over some of the existing code.
- Batching on protrans models
- Batching on alignments