flashrt/flashrt-fp8-ffn
• Updated • 48 • 5
AI & ML interests
training & inference infra
Recent Activity
flashrt/flashrt-residual-norm-quant
• Updated • 56 • 2
flashrt/fp4-fused-ops
• Updated • 49 • 1
flashrt/flashrt-qkv-cache-rope
• Updated • 44 • 2
flashrt/fa4-cute-runtime
• Updated
flashrt/fp4-gemm
• Updated • 142 • 3
flashrt/transformer-fused-ops
• Updated • 87 • 1
flashrt/masked-mha-runtime
Updated • 15
flashrt/grouped-moe-gemm
• Updated • 5 • 2
flashrt/fp8-kv-attention
• Updated • 23 • 1
flashrt/gated-delta-attention
• Updated • 81 • 3
flashrt/grouped-moe-gemv
• Updated • 56 • 2
flashrt/world-model-conv
• Updated • 23 • 1
flashrt/smallm-ffn-megakernels-blackwell
• Updated • 19 • 1
flashrt/flashrt-nvfp4
• Updated • 26 • 3
flashrt/flashrt-fused-quant
• Updated • 26 • 3
flashrt/adaptive-layernorm-producers
• Updated • 75 • 1
flashrt/weight-only-ffn
• Updated • 26 • 1
flashrt/fp8-gemm
• Updated • 22 • 2
flashrt/bf16-linear-gemv
• Updated • 29 • 1
flashrt/transformer-layout-primitives
• Updated • 31 • 2
flashrt/fp8-cross-attention-blackwell
• Updated • 8 • 1
flashrt/fused-mlp-megakernels-blackwell
• Updated • 8 • 2
flashrt/audio-codebook-primitives
• Updated • 3 • 2
flashrt/flashrt-vla-residual-gates
• Updated • 29 • 3
flashrt/blockwise-fp8-producers
• Updated • 9 • 1
flashrt/diffusion-step-ops
• Updated • 35 • 1
flashrt/flashrt-spatiotemporal-layout
• Updated • 29 • 2
flashrt/padded-fp8-producers
• Updated • 3 • 1
flashrt/fp8-prefill-attention-blackwell
• Updated • 3 • 1