Mohammad Miadh Angkad f88acf8780 [JIT Kernel] Reland NVFP4 kernels to JIT (#20012)

2026-03-07 10:31:08 +08:00

..

cutlass_moe/w4a8

[perf]optimize w4afp8 kernel on deepseek-v3-0324 (#12921)

2025-12-18 18:13:22 +08:00

cutlass_moe_helper.cu

…

fp8_blockwise_moe_kernel.cu

…

fused_qknorm_rope_kernel.cu

…

kimi_k2_moe_fused_gate.cu

…

moe_align_kernel.cu

…

moe_fused_gate.cu

…

moe_sum_reduce.cu

…

moe_sum.cu

…

moe_topk_sigmoid_kernels.cu

…

moe_topk_softmax_kernels.cu

…

prepare_moe_input.cu

…