Training Process Stuck with Zero GPU Utilization

[FrankLiu1102]

I am trying to train the model using the following command:

bash scripts/train_llama_arxiv.sh 2

The training starts successfully, and the logs indicate that the model has loaded the data and entered the training phase. However, the process gets stuck and does not proceed further. Additionally, GPU utilization remains at 0% throughout the execution.

I am new to this field and may not fully understand some implementation details. I would greatly appreciate any guidance or explanations to resolve this issue.

Thank you!