RuntimeError: CUDA out of memory, in t1_train task.
lianghrx opened this issue · 1 comments
lianghrx commented
Our gpu devices are RTX 2080TIs with only 11GB memory, but the training data requires 10-12GB GPU memory (as far as I know).
Is there any configuration to reduce the runtime memory requirements?
akshitac8 commented
Please try with reducing the batch size and other hyper-parameters which can help you scale the code for RTX 2080.