mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-12 09:16:52 +00:00
* Hybrid direct + implicit GEMM forward convolution NCHWc v5r1. Input tensor bypass LDS. Support fp32/fp16/int8
* Hybrid direct + implicit GEMM forward convolution NCHWc v5r1. Input tensor bypass LDS. Support fp32/fp16/int8