You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When I run the program on 4 NVIDIA 3090 GPUs, how should I set the train_batch_size, train_micro_batch_size_per_gpu, and gradient_accumulation_steps in the dsconfig file? In addition, what should the batch_size_train be set to in the train_bbc.yaml file under fake_config? Are there any other parameters that need to be adjusted?
The text was updated successfully, but these errors were encountered:
When I run the program on 4 NVIDIA 3090 GPUs, how should I set the train_batch_size, train_micro_batch_size_per_gpu, and gradient_accumulation_steps in the dsconfig file? In addition, what should the batch_size_train be set to in the train_bbc.yaml file under fake_config? Are there any other parameters that need to be adjusted?
The text was updated successfully, but these errors were encountered: