From 3d59ec5935ea1d33e9d51060a8dd737169b9b89b Mon Sep 17 00:00:00 2001
From: Howard Su <howard0su@gmail.com>
Date: Sat, 17 Jun 2023 23:46:15 +0800
Subject: [PATCH] ggml : fix warnings under MSVC (#1908)

---
 ggml-cuda.cu    | 4 ++++
 ggml-opencl.cpp | 4 ++++
 llama.cpp       | 2 +-
 3 files changed, 9 insertions(+), 1 deletion(-)

diff --git a/ggml-cuda.cu b/ggml-cuda.cu
index 7edd1a9f8..fed2a7ce1 100644
--- a/ggml-cuda.cu
+++ b/ggml-cuda.cu
@@ -13,6 +13,10 @@
 #include "ggml-cuda.h"
 #include "ggml.h"
 
+#if defined(_MSC_VER)
+#pragma warning(disable: 4244 4267) // possible loss of data
+#endif
+
 static_assert(sizeof(half) == sizeof(ggml_fp16_t), "wrong fp16 size");
 
 #define CUDA_CHECK(err)                                                                 \
diff --git a/ggml-opencl.cpp b/ggml-opencl.cpp
index 1d4db96ee..95f4cec6d 100644
--- a/ggml-opencl.cpp
+++ b/ggml-opencl.cpp
@@ -15,6 +15,10 @@
 
 #include "ggml.h"
 
+#if defined(_MSC_VER)
+#pragma warning(disable: 4244 4267) // possible loss of data
+#endif
+
 #define CL_DMMV_BLOCK_SIZE 32
 
 #define MULTILINE_QUOTE(...) #__VA_ARGS__
diff --git a/llama.cpp b/llama.cpp
index 81f047ed2..a50846f71 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -1253,7 +1253,7 @@ static void llama_model_load_internal(
             vram_scratch = n_batch * MB;
             ggml_cuda_set_scratch_size(vram_scratch);
             if (n_gpu_layers > 0) {
-                fprintf(stderr, "%s: allocating batch_size x 1 MB = %ld MB VRAM for the scratch buffer\n",
+                fprintf(stderr, "%s: allocating batch_size x 1 MB = %zd MB VRAM for the scratch buffer\n",
                         __func__, vram_scratch / MB);
             }
         }