Cui Junbo comments

Results 71 comments of


                                            Cui Junbo

💡 [REQUEST] - 请问可以提供一下minicpm-v2.6跑OCRBench的cot prompt吗，想复现一下实验结果，感谢

请参考 https://github.com/open-compass/VLMEvalKit

[BUG] <title>Input Error: Only 4D input Tensors are supported (got 3D)

您好，可以尝试使用我们最新的教程https://huggingface.co/openbmb/MiniCPM-V-2_6#chat-with-multiple-images

💡 [REQUEST] - Simultaneous multimodal inputs

@HwwwwwwwH Is there a doc for the OmniStreaming mode for VLLM？ Also In omni mode, some instructions can be challenging to follow. (our training data includes only user inputs that...

How to use Gradio with GGUF model?

https://github.com/OpenBMB/MiniCPM-o/blob/main/web_demos/minicpm-o_2.6/model_server.py#L96 try to change this line～

请问多图训练数据的格式是什么样的？

已开放

💡 [REQUEST] - 如何复现 OCRBench 精度，是否有专门的提示词？

没有的/ 我们的测试环境时，使用 https://github.com/open-compass/VLMEvalKit/tree/main/vlmeval 在 A100 *8 卡下进行推理得到的

💡 [REQUEST] - 如何复现 OCRBench 精度，是否有专门的提示词？

在OCR训练中未涉及过多指令，可能需要尝试较多的prompt，或者可以尝试使用few shot的方法实现您想要的结果。关于VLLM上复现OCRBench，请 @HwwwwwwwH 来看一下吧～

💡 [REQUEST] - <title>视频推理时，采样逻辑是怎么样的？

@lihytotoro

[BUG] <title>MiniCPM_V_2_6_INT4 版本执行提示 tensor 异常

@tc-mb 来看看

support MiniCPM-V-2.5

> This looks very promising! > > I've been looking into the conversion issue and made a bit of progress on that end: > > 1. convert.py is not supported...