Grass: Compute Efficient Low-Memory LLM Training with Structured Sparse Gradients

Re-assign community