For some reason this will not run for me using accelerate, no matter the config transformers will only allocate the model to a single GPU.
Is there a model specific reason for this?
· Sign up or log in to comment