zzc98 comments

Results 6 comments of


                                            zzc98

Reproduce the results in the paper

I have the same question.

Add specific cases of Ascend acceleration cards

I have the same inquiry—I have not yet been able to locate the relevant documentation for Ascend.

Incorrectly generated lines with arrows

I want to generate a graph without arrows, as follows ![Image](https://github.com/user-attachments/assets/96efb0f8-1126-4919-8b0f-92fcaadcaa56) it can be generated vai https://www.mermaidchart.com/play#pako:eNqrVkrOT0lVslJKL0osyFDwCYrJUwACR11dXVcI0wnIdAZiFwjXFcRVqgUAspQN9w I really want to draw without arrows through python.

支持国产GPU/华为昇腾NPU需要做哪些工作？

还有支持Ascend的计划吗，我看之前说 v0.2.4 就会支持，readme中也写了，但是没有看到任何与ascend相关的代码或文档

[Feature] 请求支持compressed-tensors的各种量化格式

any progress?

Eval bug: CPU usage is abnormal when running deepseek-r1-671B-Q4_0 weights in Atlas 800T a2 and NPU device。

> > [@woshidahunzi1](https://github.com/woshidahunzi1) According to [ollama/ollama#5872](https://github.com/ollama/ollama/pull/5872), Ascend does not support Q4_K_M quantization, so it fall backs to running on the CPU. > > But the model I run is Q4_0...