enable all inference and train on Gaudi/Gaudi2 with optimized perf with latest base

hpcaitech / FastFold

Optimizing AlphaFold Training and Inference on GPU Clusters

Apache License 2.0

557 stars 86 forks source link

enable all inference and train on Gaudi/Gaudi2 with optimized perf with latest base #139

Closed LeoZhao-Intel closed 1 year ago

KatarinaYuan commented 1 year ago

Hi, Thanks for this amazing work! If possible, could you please share some benchmarking for the difference of the training and inference time on Inter Habana Cluster or not?

LeoZhao-Intel commented 1 year ago

https://www.hpc-ai.tech/blog/intel-habana here is blog for some details. It could run on both Gaudi and Gaudi2 for train and inference, we would publish more details and benchmarks in the future.