Xiangyu Zhao
Xiangyu Zhao
@ganchun1130 @sanshi9523 您好，Qwen2.5VL官方的transformers版本为**4.49.0**，请切换到该版本进行评测。
@shawn0wang 您好，能麻烦share一下您的执行命令和模型的checkpoint name吗
@shawn0wang 您好，您的qwen2.5vl的checkpoint路径是什么呢？上面有同学解决了类似的问题[issue802](https://github.com/open-compass/VLMEvalKit/issues/802)。
> 我尝试发现，在数据读入的时候，好像image的图片只有路径，而且路径有可能是错的，并且type都是text，最后传入的prompt好像就只有路径，没有图片的tokens，我不知道有没有可能是这个问题，大家也可以打印看看，定义model的时候，verbose设置true即可看到输入的message，根本没有图片 您好，图片是会保存到本地的，路径一般是`index.jpg`，使用verbose并不会打印图片路径。不存在图片未输入的情况。
Hi, Thank you for your inquiry! Upon reviewing the [`load_pretrained_model`](https://github.com/Tencent/VITA/blob/cbe20436580daaff9d1d23555b7c61bbccc6c9de/vita/model/builder.py#L146) function in the [VITA repository](https://github.com/Tencent/VITA), I noticed that the original `device_map` configuration is currently set up to support only **two...
Could you please share your execution command?
I know it. When you want to split one model on four devices, do not use `--nproc_per_node=4`. Instead, if there are four devices on your machine, please use `--nproc_per_node=1`.
We recommend that you try using the **original device_map** in combination with the --nproc_per_node=1 flag to verify if this resolves the issue. Set the cuda_device to `0,1`.
@lxysl We’ve encountered some issues related to the VITA-Qwen model. Could you please assist us in resolving them?
您好，请问有完整的报错日志吗？