increase cuda_cpy block size (ggml/996)

Co-authored-by: bssrdf <bssrdf@gmail.com>

ggml/src/ggml-cuda/cpy.cuh

@@ -1,6 +1,6 @@
 #include "common.cuh"
 
-#define CUDA_CPY_BLOCK_SIZE 32
+#define CUDA_CPY_BLOCK_SIZE 64
 
 void ggml_cuda_cpy(ggml_backend_cuda_context & ctx, const ggml_tensor * src0, ggml_tensor * src1);