Nipi64310

Results 6 comments of Nipi64310

@iamjjww 把common那个文件夹放到.py文件夹的位置

@bwpandamb 好像是因为你是 python2的版本,才有这种错误, 可以 把那些字符串转成unicode的

Hi @Yard1 , Thanks a lot for your work. I encountered a RuntimeError: No suitable kernel error while testing the model loading. You can refer to the log below for...

> Yi model is not supported yet @Nipi64310 Yi model structure is now consistent with the Llama model. ![Screenshot_20231202-110417_Samsung Internet](https://github.com/vllm-project/vllm/assets/31469418/3e83298f-339f-49bf-8a95-cbbfec7344a4)

Hi @0xDing ,感谢分享,这2天刷了好多次终于等到评测了,有评测对比7b或chatgpt的吗,毕竟7b跟13b模型大小也差挺多

https://huggingface.co/TMElyralab/lyraChatGLM This chatglm6b model may be optimized based on FasterTransformer ![image](https://github.com/NVIDIA/FasterTransformer/assets/31469418/fca1f0fe-1f4d-43db-bc9a-1b6006193bf2) > 我们基于 FasterTransformer,对 ChatGLM-6B 做了较多的定制化优化,主要有: > > - **特有算子**:ChatGLM-6B 有大量的特有计算,如 2D-RotaryEmbedding、AlphaResidual、Mask 和 PositionIds 的计算等,我们分析了所有的特定计算操作,以 CUDA 为基础重写了这一系列 Kernel。 > -...