Using push_to_hub API, notebook crashed and whole process failed

I am using Google Colab to push my fine tuned model to the Hub. When I run the function, a output directory was created on the Colab my drive and the model(Bert-base-uncased) begins to train on my datasets(glue- sst2) for 2 epochs. As the 2nd epoch progress bar run to completion, the training keeps running for hours and eventually crashed and the whole failed, then I won’t be allowed connect GPU as backend for that day. I have tried this twice but keeps getting the same result. Could it be the Colab GPU isn’t enough to push the model to the Hub or am doing something wrong? I followed this process push the model to Hub hugging face push_to_hub api