Is AdamW8bit compatible with OSS in fairscale?
LeeDoYup opened this issue · comments
Thank you for the nice project.
When I use AdamW8bit optimizer, i could save the GPU memory.
However, when i combined the optimizer with OSS in fairscale,
the GPU memory is not reduced.
Is not this library compatible with OSS in fairscale. or another issue?
I checked that it is compatible with OSS, and the memory is reduced when i increase batch size.