RIicardo.M.Lu

> 你好呀， > > 我在 Mac M4 上尝试使用 0.1.27 进行微调时遇到了同样的错误，同时使用了`Qwen2-VL-2B-Instruct-4bit`和`mlx-community/Qwen2-VL-2B-Instruct-bf16`。我也尝试了`meta-llama/Llama-3.2-11B-Vision-Instruct`，但出现了不同的错误（见下文），但我不确定这是否相关。 > > 我对训练和验证数据集使用以下数据格式 > > ``` > { > "images": img_path, > "messages": [ > { > "role": "user", >...

In fact, I did not encounter this problem in versions 0.1.26 and below, but it occurred in versions 0.1.27 and above. Maybe you can try another version.

I also encountered the same problem, have you solved it?