Yuxuan Zhang

icon indicating an email [email protected]

icon company ZhipuAI @THUDM @MetaGLM icon location China icon location I am interning at Zhipu AI @THUDM @MetaGLM , researching in LLMs, Agents, and Multimodal Models.

Results 539 comments of Yuxuan Zhang

Getting requirements to build wheel did not run successfully.

哪个依赖无法安装

TypeError: GenerationMixin._extract_past_from_model_output() got an unexpected keyword argument 'standardize_cache_format'

先用transformers 4.40 老模型维护少，GLM-4会跟进维护到4.45

TypeError: GenerationMixin._extract_past_from_model_output() got an unexpected keyword argument 'standardize_cache_format'

这个bug修了，在huggingface中文件更新了

CogVideoX-5B may generate empty videos in some prompts

There is a big problem, your prompt is too short. Please carefully read our readme. We need to use long prompts as input, which requires you to use large language...

请教关于算法原理

我觉得在这里或许能帮助到你查看代码 https://huggingface.co/THUDM/glm-4v-9b/blob/main/visual.py 视觉部分是EVA2CLIPModel，然后encoder采用的LLM模型是什么？这个问题没理解，这是个decoder only模型

NPU out of memory

36g是GPU的，没有测试过NPU的这是消耗了多少G了

app运行错误

需要看一下后台的报错。只有一个前端无法debug

app运行错误

你i应该是用了FP16吧，这个模型仅支持BF16，FP16会出现CUDA错误

CogvideX-Interpolation with CogVideoX1.5-5B-SAT or CogVideoX-5b-I2V

This project is based on the diffusers version; we are about to launch it soon.

CogvideX-Interpolation with CogVideoX1.5-5B-SAT or CogVideoX-5b-I2V

We have released the version of diffusers, now you can give it a try, please download the diffusers library from the source code and run the diffusers model we open-sourced...