Dear authors,
I was curious about whether you have tried running the scripts with 13B.
I'm using 4x80GB A100, and it gives me OOM for train_generator.sh - and changing the batch size didn't particularly work for me 😢
In case of you have succeeded, could you kindly share the code for 13B?
Dear authors, I was curious about whether you have tried running the scripts with 13B. I'm using 4x80GB A100, and it gives me OOM for train_generator.sh - and changing the batch size didn't particularly work for me 😢
In case of you have succeeded, could you kindly share the code for 13B?