Milan comments

Repositories
Issues
Comments

Results 14 comments of


                                            Milan

Turbo-V3

any idea, how can I run it faster using apple silicon, as i have an M2 pro machine.

Trained deepseek qwen 32b R1 model giving rubbish output, though training went fine

> Is it inference with Unsloth or inference outside of unsloth? see: https://docs.unsloth.ai/basics/running-and-saving-models/troubleshooting i have tried with both, unsloth and outside unsloth. llama.cpp, transformers, unsloth inference, these i havr tried,...

doesnt work when deployed in streamlit community cloud

it works if i copy link address and then paste it in new window.

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",])

were you able to solve it? i m getting same issue with a base model.