From: Georgi Gerganov <redacted>
Date: Sat, 6 Apr 2024 14:50:21 +0000 (+0300)
Subject: sync : llama.cpp (skip)
X-Git-Tag: upstream/0.0.1642~786
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=765d08f7b40134d0705cc964d7e222a0da9ca7ae;p=pkg%2Fggml%2Fsources%2Fggml

sync : llama.cpp (skip)

ggml-ci
---

diff --git a/scripts/sync-llama-am.sh b/scripts/sync-llama-am.sh
index f81477c3..a113e0b9 100755
--- a/scripts/sync-llama-am.sh
+++ b/scripts/sync-llama-am.sh
@@ -59,6 +59,7 @@ while read c; do
         ggml*.m \
         ggml*.metal \
         ggml*.cu \
+        ggml-cuda/* \
         tests/test-opt.cpp \
         tests/test-grad0.cpp \
         tests/test-quantize-fns.cpp \
diff --git a/scripts/sync-llama.last b/scripts/sync-llama.last
index 50fbf773..fe0dce4f 100644
--- a/scripts/sync-llama.last
+++ b/scripts/sync-llama.last
@@ -1 +1 @@
-e82f9e2b833d88cd2b30123ef57346c2cb8abd99
+4d17d0e240a87661f0873facec7d60df46130589
diff --git a/src/ggml-cuda/common.cuh b/src/ggml-cuda/common.cuh
index 79e1d542..b98d7cbd 100644
--- a/src/ggml-cuda/common.cuh
+++ b/src/ggml-cuda/common.cuh
@@ -231,12 +231,6 @@ typedef float dfloat; // dequantize float
 typedef float2 dfloat2;
 #endif //GGML_CUDA_F16
 
-// dmmv = dequantize_mul_mat_vec
-// TODO: remove this?
-#ifndef GGML_CUDA_DMMV_X
-#define GGML_CUDA_DMMV_X 32
-#endif
-
 [[noreturn]]
 static __device__ void no_device_code(
     const char * file_name, const int line, const char * function_name, const int arch, const char * arch_list) {
diff --git a/src/ggml-cuda/dmmv.cu b/src/ggml-cuda/dmmv.cu
index ea2950f3..0b17e3cb 100644
--- a/src/ggml-cuda/dmmv.cu
+++ b/src/ggml-cuda/dmmv.cu
@@ -2,10 +2,6 @@
 #include "dequantize.cuh"
 #include "convert.cuh"
 
-#ifndef GGML_CUDA_MMV_Y
-#define GGML_CUDA_MMV_Y 1
-#endif
-
 #ifndef K_QUANTS_PER_ITERATION
 #define K_QUANTS_PER_ITERATION 2
 #else
diff --git a/src/ggml-cuda/dmmv.cuh b/src/ggml-cuda/dmmv.cuh
index 3802678f..4c5ebd47 100644
--- a/src/ggml-cuda/dmmv.cuh
+++ b/src/ggml-cuda/dmmv.cuh
@@ -1,5 +1,16 @@
 #include "common.cuh"
 
+// dmmv = dequantize_mul_mat_vec
+
+// TODO: remove this?
+#ifndef GGML_CUDA_DMMV_X
+#define GGML_CUDA_DMMV_X 32
+#endif
+
+#ifndef GGML_CUDA_MMV_Y
+#define GGML_CUDA_MMV_Y 1
+#endif
+
 void ggml_cuda_op_dequantize_mul_mat_vec(
     ggml_backend_cuda_context & ctx,
     const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst, const char * src0_dd_i, const float * src1_ddf_i,