Expecting New NVFP4 for GLM-5.1 ASAP, Thanks!
Expecting New NVFP4 for GLM-5.1 ASAP, Thanks!
Hi, We are working on it and plan to share asap.
Is it also supported by vLLM?
We see a GLM-5.1 NVFP4 model shared from third-party: lukealonso/GLM-5.1-NVFP4, which seems to work pretty well:
https://github.com/microsoft/Tutel?tab=readme-ov-file#steps-for-glm-551-claude-code-mode
Although lukealonso/GLM-5.1-NVFP4 works, the precision after quantization doesn't seem nice. Hope to get a well-tuned NVFP4 format.
Although lukealonso/GLM-5.1-NVFP4 works, the precision after quantization doesn't seem nice. Hope to get a well-tuned NVFP4 format.
In the spirit of learning, please can you explain what you mean by the precision does not seem nice? Thank you in advance.
the score of aime is lower than the glm-5.
Is there an expected release of glm-5.1-NVFP4?
any news?
any updates?