Dmytro Semenov

@buffcode @OriginalYellow The module has been created to be used in a single request context and never meant to be used in a long-running process. So, this is not a...

The same issue for me

Given that other reports on similar errors are related to mismatch in index. I found one place that was adding pad token to tokenizer in llama_chat which was not accounted...

Here's my llama2_chat.py version that worked for me, just make sure you remove adding pad token here: [https://github.com/OpenAccess-AI-Collective/axolotl/blob/ec0958f4f846236ac2703dd644f6dac4365f64b4/src/axolotl/utils/models.py#L80](url) And if you still need padding, then better update model separately with...

> @vibhorag101 I'm sorry but I don't feel comfortable sharing it, because I don't really know what I'm doing. I wouldn't want people to waste time on something if I'm...