alexazel

Reproducibility environment for DoRA fused Triton kernels: CUDA 13.1, PyTorch 2.10, Triton 3.6, vLLM