Add optimized CUDA kernels (RMSNorm, MLP) and patched Qwen3 modeling file e36ec14 verified Khanin Udomchoksakul commited on 4 days ago
