search

hpcaitech / FastFold

Optimizing AlphaFold Training and Inference on GPU Clusters
Apache License 2.0
556 stars 84 forks source link

AutoChunk #160

Open georgkempf opened 1 year ago

georgkempf commented 1 year ago

Hello, I was wondering whether AutoChunk described in https://arxiv.org/pdf/2203.00854v3.pdf is already implemented? Will this avoid manually setting the chunk_size?

oahzxl commented 1 year ago

Hi, autochunk was implemented in colossalAI, and it has not been added to this repo. So I'm afraid you still need to set it manually now.

ssn1771 commented 6 months ago

@oahzxl Hi, I was curious about the inference overhead caused by using AutoChunk for protein prediction models , did it take some time to find the optimal chunk size？