Update llama.cpp

LahiRumesh · Aug 15, 2023 · b77388e · b77388e
1 parent a8398a6
commit b77388e
Show file tree

Hide file tree

Showing 3 changed files with 16 additions and 1 deletion.
diff --git a/crates/ggml/sys/llama-cpp b/crates/ggml/sys/llama-cpp
diff --git a/crates/ggml/sys/src/cuda.rs b/crates/ggml/sys/src/cuda.rs
@@ -60,6 +60,9 @@ extern "C" {
 extern "C" {
     pub fn ggml_cuda_set_main_device(main_device: ::std::os::raw::c_int);
 }
+extern "C" {
+    pub fn ggml_cuda_set_mul_mat_q(mul_mat_q: bool);
+}
 extern "C" {
     pub fn ggml_cuda_set_scratch_size(scratch_size: usize);
 }

diff --git a/crates/ggml/sys/src/lib.rs b/crates/ggml/sys/src/lib.rs
@@ -1567,6 +1567,18 @@ extern "C" {
         n_ctx: ::std::os::raw::c_int,
     ) -> *mut ggml_tensor;
 }
+extern "C" {
+    pub fn ggml_rope_custom(
+        ctx: *mut ggml_context,
+        a: *mut ggml_tensor,
+        n_past: ::std::os::raw::c_int,
+        n_dims: ::std::os::raw::c_int,
+        mode: ::std::os::raw::c_int,
+        n_ctx: ::std::os::raw::c_int,
+        freq_base: f32,
+        freq_scale: f32,
+    ) -> *mut ggml_tensor;
+}
 extern "C" {
     pub fn ggml_rope_custom_inplace(
         ctx: *mut ggml_context,
+19 −7		CMakeLists.txt
+23 −7		Makefile
+28 −3		README.md
+52 −44		convert.py
+24 −7		examples/common.cpp
+5 −2		examples/common.h
+132 −0		examples/json-schema-to-grammar.py
+2 −2		examples/main/README.md
+140 −39		examples/perplexity/perplexity.cpp
+1 −0		examples/save-load-state/save-load-state.cpp
+26 −0		examples/server-llama2-13B.sh
+1 −1		examples/server/README.md
+109 −0		examples/server/chat-llama2.sh
+1,063 −1,063		examples/server/index.html.hpp
+1 −2		examples/server/public/index.html
+13 −2		examples/server/server.cpp
+541 −0		ggml-alloc.c
+22 −0		ggml-alloc.h
+2,090 −438		ggml-cuda.cu
+1 −0		ggml-cuda.h
+17 −16		ggml-metal.m
+4 −1		ggml-metal.metal
+49 −26		ggml.c
+12 −1		ggml.h
+6 −10		grammars/json.gbnf
+32 −30		k_quants.c
+199 −68		llama.cpp
+1 −0		llama.h
+2 −2		scripts/sync-ggml.sh
+3 −3		tests/CMakeLists.txt
+7 −5		tests/test-double-float.cpp
+16 −16		tests/test-grad0.cpp
+8 −7		tests/test-opt.cpp