Libertai/glm53-flash-vllm-gb10
LibertAI Labs: running GLM-5.3-Flash-NVFP4 under vLLM on 2x GB10 (sm_121). Hand-written sparse-MLA CUDA kernel for NoPE MLA, plus a fix for vLLM's uninitialised NVFP4 MoE activation scale.
Python
6
1 contributors
Apache License 2.0
Social mentions
Recent discussions about this repository across the web
Repository activities
repository's daily and monthly activities across stars, forks, merged PRs, issues, and closed issues