Nova-Prime-Kernels / crates /blitz-kernels /src /cuda /mamba_ssd_1024tile_32warps.cu
Antigravity Agent
Blitz: Final 3.7x Artisan Source Sync
f6e23b0
raw
history blame contribute delete
294 Bytes
// Auto-Generated Mamba Kernel
// Tile: 1024, Warps: 32 (Threads: 1024), Items: 1, Registers: Max
#include "../csrc/selective_scan/selective_scan_fwd_kernel.cuh"
// Instantiation for Tile 1024
template void selective_scan_fwd_cuda<at::Half, float>(SSMParamsBase &params, cudaStream_t stream);