Open Zxb-QHU opened 6 months ago
Hi, you can first use the pre-train script to train the checkpoint, then put the checkpoint following the guide in readme.
Thank you for your response. I have another issue: during training, the GPU is not heavily utilized, but the CPU is maxed out, preventing the program from proceeding. Only when reading CSV files, reading 10 samples allows the training to occur.
Follow your steps to pre-train T5, the code won't run. Is there any error in the code in the repository.
Thank you for your answer.