Train arguement --devices does not work
Closed this issue · 1 comments
Harsh-Gill commented
it seems if i type
python -m src.train --devices 0,1,2,3 --width 48 --arch EquiUnet
The speed is the same as
it seems if i select python -m src.train --devices 0 -width 48 -arch EquiUnet
Is this devices option untested?
Harsh-Gill commented
This was due to the batch size being limited to 1 , when i increase batch size to 4 or more , it is able to utilize more GPUs