Antonios Papadakis comments

Results 6 comments of


                                            Antonios Papadakis

fix a bug

to me this looks like no change at all :(

build

> He should publish release this is a header-only library. That would be useless. You're supposed to get your depencies and then reference the source files and compile

[BUG: Error during training

Did you remove the `no_eval` declaration that is already included in the `#other` section of the examle YAML? This seemed to solve that error for me.

Qwen-2.5 Coder-7B-Instruct: ValueError: Unsloth: Untrained tokens found, but embed_tokens & lm_head not trainable, causing NaNs. Restart then add `embed_tokens` & `lm_head` to `FastLanguageModel.get_peft_model(target_modules = [..., "embed_tokens", "lm_head",])

@danielhanchen This didn't fix the issue. I'm trying to train Mistral-12B-NeMo-Instruct.

Qwen-2.5 Coder-7B-Instruct: ValueError: Unsloth: Untrained tokens found, but embed_tokens & lm_head not trainable, causing NaNs. Restart then add `embed_tokens` & `lm_head` to `FastLanguageModel.get_peft_model(target_modules = [..., "embed_tokens", "lm_head",])

It worked when I trained my first LoRA, but is now failing on my second dataset.

NotImplementedError: Unsloth: unsloth/Llama-3.2-11B-Vision-Instruct-bnb-4bit not supported yet!

@danielhanchen thank you :) I'm not strictly wanting to use 3.2 11B because of the vision. It has 3B more parameters than 3.1 8B and I don't have enough computing...