Open mo666666 opened 1 year ago
I try to perform the experiments on 8A100 GPUs. However, as I observed, the utilities of GPUs are quite low (<20%). Therefore, I am quite curious about whether tricks exist to further accelerate the training process.
I try to perform the experiments on 8A100 GPUs. However, as I observed, the utilities of GPUs are quite low (<20%). Therefore, I am quite curious about whether tricks exist to further accelerate the training process.