sei-framework
sei-framework copied to clipboard
CUDA out of memory.
Very good work. I'm using your data to train the model following your tutorials and I'm running out of memory on the GPU, how can I modify the train.yml file? I don't have a GPU cluster, only a 24GB Nvidia 3090. hope you can give me a new train.yml file to help to finish the training.
Often times this is an issue where the batch size is too large for the GPU, you can try modifying the YAML with a very small batch size to see if it's better (like batch size 2 or 4).