Hongji Zhu comments

Results 64 comments of


                                            Hongji Zhu

Mac推理报错

No any other error message? Just stuck there or exited?

小米手机运行 MiniCPM-V 2.0 apk，模型回复异常

感谢反馈～由于模型量化损失和实现细节差异，导致有一定概率出现重复文字输出，我们将尽快修复～

我的M3芯片本地运行MiniCPM-Llama3-V-2_5-int4得到了报错Using `bitsandbytes` 8-bit quantization requires Accelerate

由于bitsandbytes (see [issse](https://github.com/TimDettmers/bitsandbytes/issues/252))现在还不支持MPS，int4模型暂时无法在mac运行。

我的M3芯片本地运行MiniCPM-Llama3-V-2_5-int4得到了报错Using `bitsandbytes` 8-bit quantization requires Accelerate

> > 由于bitsandbytes (see [issse](https://github.com/TimDettmers/bitsandbytes/issues/252))现在还不支持MPS，int4模型暂时无法在mac运行。 > > 请问目前支持win10吗？我setup了requirements和bitsandbytes，都提示 raise ImportError( ImportError: Using `bitsandbytes` 8-bit quantization requires Accelerate: `pip install accelerate` and the latest version of bitsandbytes: `pip install -i https://pypi.org/simple/...

macbook m3 运行web_demo_2.5.py，不能返回正常中文信息[BUG] <title>

你好，为方便调试，麻烦提供可复现的图像和prompt～

[BUG] get_vllm_embedding中的patch_attn_mask计算有问题

感谢反馈，我们正在评估影响

[BUG] web_demo_2.6.py RuntimeError: probability tensor contains either `inf`, `nan` or element < 0

请问显卡型号是什么？

[BUG] web_demo_2.6.py RuntimeError: probability tensor contains either `inf`, `nan` or element < 0

请确认web_demo_2.6.py代码没有做任何改动，模型权重是否下载完整(可以检查sha256值跟hf上是否一致)；如果可以麻烦提供可复现的测试原图和prompt

回复异常token answer: '></em>

Please provide code and input data for reproduce

[BUG] Sizes of tensors must match except in dimension 1. Expected size 123 but got size 122 for tensor number 1 in the list.

Looks like too many frames and exceeds the max_inp_length 8192, limit the max frames to a smaller size would be ok.