Open source repositories tagged with #gemma-4, ranked by health score.
Fast and Memory-Efficient Exact Attention for Large Headdim, 1.5x~6x speedup over PyTorch SDPA.
