Nonuniform-Tensor-Parallelism: Mitigating GPU failure impact for Scaled-up LLM Training

Re-assign community