online-softmax

Here are 5 public repositories matching this topic...

Learn Triton by building FlashAttention from scratch — V2 kernels, persistent threads, mask DSL, profiling toolkit, bilingual docs

Hydra-2P, KDA, and CADR: fused PyTorch/Triton kernels for routing across model depth and sequence time.

Runnable C++20 experiments for RoPE, KV cache, speculative decoding, attention layouts, online softmax, and GEMM.

cpp rope gemm systems-programming cpp20 kv-cache llm llm-inference speculative-decoding online-softmax

Row-wise CUDA softmax kernels: shared-memory reduction, warp-shuffle reduction, and online softmax benchmarked against cuDNN.

cuda cudnn softmax gpu-performance nsight-compute online-softmax

Benchmark-driven Triton kernel experiments covering vector operations, reductions, online softmax, and GPU optimization.

Add a description, image, and links to the online-softmax topic page so that developers can more easily learn about it.

To associate your repository with the online-softmax topic, visit your repo's landing page and select "manage topics."