Upload llama.cpp/ggml/src/ggml-cuda/diagmask.cuh with huggingface_hub

llama.cpp/ggml/src/ggml-cuda/diagmask.cuh

@@ -0,0 +1,5 @@
+#include "common.cuh"
+
+#define CUDA_DIAG_MASK_INF_BLOCK_SIZE 32
+
+void ggml_cuda_op_diag_mask_inf(ggml_backend_cuda_context & ctx, ggml_tensor * dst);