Open WilTay1 opened 2 weeks ago
this is my machine 2 A100 40G for each
how to make the model run both GPUs?
Thanks for your interest! We run I-Max on A100 with 80G VRAM so haven't encountered this problem. But you can try changing query_batch = False
to query_batch = True
in attention_processor.py
to reduce the peak VRAM. Good luck!
what is maximum resolution 4090 can process without OOM?
this is my machine 2 A100 40G for each
how to make the model run both GPUs?