zhongtianHU

Traceback (most recent call last): File "/opt/conda/envs/llava/lib/python3.10/runpy.py", line 196, in _run_module_as_main return _run_code(code, main_globals, None, File "/opt/conda/envs/llava/lib/python3.10/runpy.py", line 86, in _run_code exec(code, run_globals) File "/root/siton-glusterfs-eaxtsxdfs/hzt/projects/LLaVA-main_main/llava/serve/cli.py", line 130, in main(args) File...

我是对qwen-vl-chat进行的微调，使用了我自己的图文对进行训练，vllm可以也可以使用我的图文对测试集进行批量推理么？

那好吧，yi-vl-6B也是不支持的吗?而且训练yi-vl-6B的时候直接在sft.sh中添加--custom_train_dataset_path进行训练报错了 Traceback (most recent call last): File "/root/siton-glusterfs-eaxtsxdfs/hzt/projects/swift/swift/cli/sft.py", line 5, in sft_main() File "/root/siton-glusterfs-eaxtsxdfs/hzt/projects/swift/swift/utils/run_utils.py", line 31, in x_main result = llm_x(args, **kwargs) File "/root/siton-glusterfs-eaxtsxdfs/hzt/projects/swift/swift/llm/sft.py", line 144, in llm_sft td0, tkwargs0...

我看到你们在dataset.py中有一个针对于yi-vl系列的数据集格式的模板但是那是针对官方的coco-mini-en数据集的，但是自定义数据的快速拓展针对yi-vl系列并没有很好的兼容。

好的，感谢，忘记这部分了，我去尝试一下

> yi-vl模型貌似可以用--custom_val_dataset_path数据集批量推理，但推理不全，swift infer命令设置--custom_val_dataset_path参数，只会输出10个推理结果 是的，只会输出10个结果

> 只会输出10个推理结果 这个原因是需要设置一下 val_dataset_sample 参数 设置为-1 val_dataset_sample参数好像并没有找到可以修改的地方

>  Hi, is the problem solved, I had some similar problem today on Yi-VL-6B, the model can't output Chinese and generates it haphazardly. But the day before the problem,...

> 不 OK, maybe there's something wrong with the model itself, or maybe there's something in the reasoning that doesn't fit the conditions

> hello~ > > yi-vl-6b is an excellent performing model, and the ms-swift LLM training framework has incorporated sft for yi-vl. It provides example scripts and supports custom datasets. You...