llama_cpp 0.5.2 → 0.6.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- checksums.yaml +4 -4
- data/CHANGELOG.md +13 -0
- data/README.md +6 -5
- data/examples/chat.rb +13 -13
- data/examples/embedding.rb +9 -9
- data/ext/llama_cpp/llama_cpp.cpp +547 -272
- data/ext/llama_cpp/src/ggml-alloc.c +14 -8
- data/ext/llama_cpp/src/ggml-alloc.h +1 -0
- data/ext/llama_cpp/src/ggml-cuda.cu +307 -127
- data/ext/llama_cpp/src/ggml-cuda.h +1 -0
- data/ext/llama_cpp/src/ggml-metal.h +4 -0
- data/ext/llama_cpp/src/ggml-metal.m +200 -94
- data/ext/llama_cpp/src/ggml-metal.metal +264 -82
- data/ext/llama_cpp/src/ggml-opencl.cpp +3 -3
- data/ext/llama_cpp/src/ggml.c +1647 -865
- data/ext/llama_cpp/src/ggml.h +143 -52
- data/ext/llama_cpp/src/llama.cpp +1427 -635
- data/ext/llama_cpp/src/llama.h +308 -119
- data/lib/llama_cpp/version.rb +2 -2
- data/lib/llama_cpp.rb +5 -9
- data/sig/llama_cpp.rbs +65 -34
- metadata +3 -3
|
@@ -31,6 +31,7 @@ GGML_API void ggml_cuda_assign_buffers_force_inplace(struct ggml_tensor * tens
|
|
|
31
31
|
|
|
32
32
|
GGML_API void ggml_cuda_assign_buffers_no_alloc(struct ggml_tensor * tensor);
|
|
33
33
|
GGML_API void ggml_cuda_assign_scratch_offset(struct ggml_tensor * tensor, size_t offset);
|
|
34
|
+
GGML_API void ggml_cuda_copy_to_device(struct ggml_tensor * tensor);
|
|
34
35
|
|
|
35
36
|
GGML_API void ggml_cuda_set_main_device(int main_device);
|
|
36
37
|
GGML_API void ggml_cuda_set_mul_mat_q(bool mul_mat_q);
|
|
@@ -19,6 +19,8 @@
|
|
|
19
19
|
|
|
20
20
|
#pragma once
|
|
21
21
|
|
|
22
|
+
#include "ggml.h"
|
|
23
|
+
|
|
22
24
|
#include <stddef.h>
|
|
23
25
|
#include <stdbool.h>
|
|
24
26
|
|
|
@@ -33,6 +35,8 @@ struct ggml_cgraph;
|
|
|
33
35
|
extern "C" {
|
|
34
36
|
#endif
|
|
35
37
|
|
|
38
|
+
void ggml_metal_log_set_callback(ggml_log_callback log_callback, void * user_data);
|
|
39
|
+
|
|
36
40
|
struct ggml_metal_context;
|
|
37
41
|
|
|
38
42
|
// number of command buffers to use
|