RVC-Project/Retrieval-based-Voice-Conversion-WebUI

Slower Multi-GPU training with 2x the number of GPUs and 4x the amount of VRAM

Aberta

#244 aberto em 7 de mai. de 2023

 (1 comentário) (1 reação) (0 responsável)Python (2.849 forks)batch import
help wantedquestion

Métricas do repositório

Stars
 (18.427 estrelas)
Métricas de merge de PR
 (Nenhuma PRs mesclada em 30d)

Description

I have two systems training on identical datasets

System A has 4 x NVIDIA RTX A5000 (24GB VRAM per GPU), and a batch size of 12 per GPU.

System B has 7 x NVIDIA RTX A6000 (48GB VRAM per GPU), and a batch size of 18 per GPU.

I would expect System B to train much faster. However...

  • System A (96GB total VRAM, batch size 12) takes 11 seconds per epoch.

  • System B (336GB total VRAM, batch size 18) takes 13 seconds per epoch.

I'm wondering if this is down to the overhead of multi-GPU training, or if there's something I'm missing here?

Thank you

Guia do colaborador