chore: bump llama.cpp

Signed-off-by: thxCode <[email protected]>
gpustack · Feb 20, 2025 · 736132b · 736132b
1 parent dc55da8
commit 736132b
Show file tree

Hide file tree

Showing 3 changed files with 6 additions and 6 deletions.
diff --git a/llama-box/patches/llama.cpp/ggml-cpu.patch b/llama-box/patches/llama.cpp/ggml-cpu.patch
@@ -1,8 +1,8 @@
 diff --git a/ggml/src/ggml-cpu/ggml-cpu.c b/ggml/src/ggml-cpu/ggml-cpu.c
-index dbef5df2..bb116240 100644
+index f27b9817..01b5aa32 100644
 --- a/ggml/src/ggml-cpu/ggml-cpu.c
 +++ b/ggml/src/ggml-cpu/ggml-cpu.c
-@@ -9023,7 +9023,9 @@ static void ggml_compute_forward_soft_max_f32(
+@@ -9040,7 +9040,9 @@ static void ggml_compute_forward_soft_max_f32(
          ggml_vec_max_f32(nc, &max, wp);
 
          ggml_float sum = ggml_vec_soft_max_f32(nc, dp, wp, max);

diff --git a/llama-box/patches/llama.cpp/mrope.patch b/llama-box/patches/llama.cpp/mrope.patch
@@ -15,10 +15,10 @@ index d2b0d50e..1382498a 100644
      if (lctx == NULL) {
          LOG_ERR("%s: failed to create context with model '%s'\n", __func__, params.model.c_str());
 diff --git a/ggml/src/ggml-cpu/ggml-cpu.c b/ggml/src/ggml-cpu/ggml-cpu.c
-index dbef5df2..d514e1c7 100644
+index f27b9817..27fb32d2 100644
 --- a/ggml/src/ggml-cpu/ggml-cpu.c
 +++ b/ggml/src/ggml-cpu/ggml-cpu.c
-@@ -9406,10 +9406,6 @@ static void ggml_compute_forward_rope_f32(
+@@ -9423,10 +9423,6 @@ static void ggml_compute_forward_rope_f32(
      const bool is_mrope = mode & GGML_ROPE_TYPE_MROPE;  // ggml_rope_multi, multimodal rotary position embedding
      const bool is_vision = mode == GGML_ROPE_TYPE_VISION;
 
@@ -29,7 +29,7 @@ index dbef5df2..d514e1c7 100644
      if (is_vision) {
          GGML_ASSERT(n_dims == ne0/2);
      }
-@@ -9592,10 +9588,6 @@ static void ggml_compute_forward_rope_f16(
+@@ -9609,10 +9605,6 @@ static void ggml_compute_forward_rope_f16(
      const bool is_mrope = mode & GGML_ROPE_TYPE_MROPE;
      const bool is_vision = mode == GGML_ROPE_TYPE_VISION;
 

diff --git a/llama.cpp b/llama.cpp
+70 −5		examples/run/run.cpp
+1 −0		ggml/CMakeLists.txt
+1 −0		ggml/include/ggml-cpu.h
+96 −6		ggml/src/ggml-cpu/CMakeLists.txt
+176 −1		ggml/src/ggml-cpu/ggml-cpu-quants.c
+29 −4		ggml/src/ggml-cpu/ggml-cpu.c
+16 −0		ggml/src/ggml-cpu/ggml-cpu.cpp
+259 −0		ggml/src/ggml-cpu/kleidiai/kernels.cpp
+61 −0		ggml/src/ggml-cpu/kleidiai/kernels.h
+287 −0		ggml/src/ggml-cpu/kleidiai/kleidiai.cpp
+17 −0		ggml/src/ggml-cpu/kleidiai/kleidiai.h