Call this once at init time before running any ggml_flash_attn_sparse graphs.
| 29 | |
| 30 | // Call this once at init time before running any ggml_flash_attn_sparse graphs. |
| 31 | void pflash_register_ggml_kernel() { |
| 32 | ggml_cuda_flash_attn_sparse_set_kernel(&pflash_adapter); |
| 33 | } |