auroralsun

Results 2 comments of auroralsun

微调过程loss为0， grad_norm为nan

I ran into the same situation where I had limited the length of the data set to 3000 and max_length to 4096 ![image](https://github.com/ymcui/Chinese-LLaMA-Alpaca-3/assets/59675069/8cd7d69a-ef8a-4f38-9af9-ffadf6ea40de)

微调过程loss为0， grad_norm为nan

> I ran into the same situation where I had limited the length of the data set to 3000 and max_length to 4096 ![image](https://private-user-images.githubusercontent.com/59675069/329804812-8cd7d69a-ef8a-4f38-9af9-ffadf6ea40de.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.JBpq6zO4U-PMFJMzal_REDUm9NQYX-gJ0YF0Uo4n-50) 用LLAMA-factory训练了，一切正常