Daniel Han

Oh wait actually if you're using `TextStreamer`, you can use `skip_prompt` as in our Ollama notebook: https://colab.research.google.com/drive/1WZDi7APtQ9VsvOrQSSC5DDtxq159j8iZ?usp=sharing

Hmm good point - on Databricks VMs - someone in our Discord chat said using `ddp_find_unused_parameters=False` in `TrainingArguments` made 1x A10G instance work - unsure if that's the solution

@davanstrien Great idea!! Will add an ORPO example :) I didn't look into it yet since didn't have time to add one, but 100% is a fantastic idea!

@davanstrien Oh wait now that I checked @gagan3012's colab, it's just the DPO trainer, except using a ORPO loss, and no SFT component??!

Oh fantastic! That's very cool and less work for me :)))

Oh was gonna say we just added one! https://colab.research.google.com/drive/11t4njE3c4Lxl-07OD8lJSMKkfyJml3Tn?usp=sharing

@deniz-birlikci Oh my they actually set it to None? Thats weird

Oh seems like this happened for Llama-3 as well - I added this in - credits go to you!! Thanks!

Known problem sadly :( I just didn't know how to fix it - I'm assuming ur calling `get_chat_template` more than once in the notebook correct?

@81549361 Oh that's not good - let me check