How can I re-train dolly in a clean A100 Linux machine?

Open biduan opened this issue 2 years ago • 1 comments

Hi, I have one Linux server with 4 A100 GPU. run a smoke test to inference model which is successful But when try to run the script "train_dolly.py“ to re-train model, it was fail. It seems like it need to install Databricks Runtime 12.2 ML but I didn't found any available installer. and The document didn't mention how to setup a training environment. Could anyone give me some advice how to re-train dolly in a clean A100 single machine please?

Apr 26 '23 14:04 biduan

You haven't said anything about the problem. You dont' need Databricks, but you would have to change the references to dbutils and the %pip install command.

Apr 26 '23 14:04 srowen