finetune
#1
by
karsar
- opened
finetune
What method is used to fine-tune this model?
The model was finetuned on conversations, similar to the llama2 chat version. We used the llama2 chat template. We provided both very long conversations between the user and trurl, and short instruction-answer pairs. The whole model was trained, without any layer freezing.
It would be interesting to try to fine tune it using LORA and compare the results.
Nice idea, we will think about it!
AgaMiko
changed discussion status to
closed