LoRA, GPU & all hyperparameters

#1
by despinapz - opened

Hello, thank you for the provided information. Can you list all the hyperparameters and the GPU you used to achieve this accuracy, please? Also, did you use the LoRA method? I cannot replicate your results using your selected hyperparameters, as the loss is becoming 0.0 after just 0.1 of the first epoch. Thank you.

I used adamw_8bit as the optimizer.

despinapz changed discussion status to closed

Sign up or log in to comment