I build sglang from source code of your forked sglang/dots.vlm1 branch.
we start a sglang server but the EXCEPTION occurs (see title).
if I remove the --chat-template dots-vlm from the bash command, the server starts as expected (seems without chat template?).
We test the pure text prompt, which is ok. We test the multimodal prompt, which results in a response that seems the given image is missing.
Can you please release the environments/requirements in your docker? I cannot use the docker directly due to my offline workplace.
我把sglang/dots.vlm1分支下载下来并且用源码安装了，如果直接按照readme中的例子起服务，会报如题错误。如果去掉对话模板指定，则错误消失，服务正常启动，但似乎发送的多模态请求中的图像部分模型不可见，回答内容完全和图像内容无关。由于我的环境是内部离线环境，机器的cuda版本较高，无法直接使用你们提供的容器。请问是否可以提供一个详细的环境或依赖的配置文件？谢谢！
I build sglang from source code of your forked sglang/dots.vlm1 branch.
we start a sglang server but the EXCEPTION occurs (see title).
if I remove the
--chat-template dots-vlmfrom the bash command, the server starts as expected (seems without chat template?).We test the pure text prompt, which is ok. We test the multimodal prompt, which results in a response that seems the given image is missing.
Can you please release the environments/requirements in your docker? I cannot use the docker directly due to my offline workplace.
我把sglang/dots.vlm1分支下载下来并且用源码安装了，如果直接按照readme中的例子起服务，会报如题错误。如果去掉对话模板指定，则错误消失，服务正常启动，但似乎发送的多模态请求中的图像部分模型不可见，回答内容完全和图像内容无关。由于我的环境是内部离线环境，机器的cuda版本较高，无法直接使用你们提供的容器。请问是否可以提供一个详细的环境或依赖的配置文件？谢谢！