llama_cpp 0.10.0 → 0.10.1
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 +6 -0
- data/ext/llama_cpp/llama_cpp.cpp +2 -0
- data/ext/llama_cpp/src/ggml-alloc.h +1 -1
- data/ext/llama_cpp/src/ggml-cuda.cu +691 -93
- data/ext/llama_cpp/src/ggml-metal.m +535 -54
- data/ext/llama_cpp/src/ggml-metal.metal +1497 -169
- data/ext/llama_cpp/src/ggml-quants.c +2 -2
- data/ext/llama_cpp/src/ggml.c +325 -159
- data/ext/llama_cpp/src/ggml.h +34 -13
- data/ext/llama_cpp/src/llama.cpp +195 -35
- data/ext/llama_cpp/src/llama.h +1 -1
- data/lib/llama_cpp/version.rb +2 -2
- data/sig/llama_cpp.rbs +2 -0
- metadata +2 -2
checksums.yaml
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
---
|
|
2
2
|
SHA256:
|
|
3
|
-
metadata.gz:
|
|
4
|
-
data.tar.gz:
|
|
3
|
+
metadata.gz: 9bd5dbea8695fdb41ea6d97e372c2cea452ee8ed070e26bd558a720d6c24fe27
|
|
4
|
+
data.tar.gz: 8d7bfd02445df81644eebb3a6db49bb1ddf241a344fef76a0b949f1c12d0639d
|
|
5
5
|
SHA512:
|
|
6
|
-
metadata.gz:
|
|
7
|
-
data.tar.gz:
|
|
6
|
+
metadata.gz: 116423f4581f605ee379bcd690299e152087c03665be6171bc137205f0824be5d5e0ce6d3c0b548fc6d28193c601679c3cc27a101b6f5b58968c69388a70cbfc
|
|
7
|
+
data.tar.gz: fd162b1c4e26732573d32ba7439fc44dd2ff09a9024c64ff5adce997e154a934d9556f0ff3cfd6b83dd7dcbfb8366cc6ce394b8324f39ef349d4cb5834ea43f1
|
data/CHANGELOG.md
CHANGED
|
@@ -1,3 +1,9 @@
|
|
|
1
|
+
## [[0.10.1](https://github.com/yoshoku/llama_cpp.rb/compare/v0.10.0...v0.10.1)] - 2023-12-16
|
|
2
|
+
|
|
3
|
+
- Bump bundled llama.cpp from b1620 to b1641.
|
|
4
|
+
- Add attribute reader for `params` to `Model`.
|
|
5
|
+
- Add `Batch` class, this class was not published because the author forgot to write `rb_define_class`.
|
|
6
|
+
|
|
1
7
|
## [[0.10.0](https://github.com/yoshoku/llama_cpp.rb/compare/v0.9.5...v0.10.0)] - 2023-12-09
|
|
2
8
|
|
|
3
9
|
- Bump bundled llama.cpp from b1593 to b1620.
|
data/ext/llama_cpp/llama_cpp.cpp
CHANGED
|
@@ -1333,6 +1333,7 @@ public:
|
|
|
1333
1333
|
static void define_class(VALUE outer) {
|
|
1334
1334
|
rb_cLLaMAModel = rb_define_class_under(outer, "Model", rb_cObject);
|
|
1335
1335
|
rb_define_alloc_func(rb_cLLaMAModel, llama_model_alloc);
|
|
1336
|
+
rb_define_attr(rb_cLLaMAModel, "params", 1, 0);
|
|
1336
1337
|
rb_define_method(rb_cLLaMAModel, "initialize", RUBY_METHOD_FUNC(_llama_model_initialize), -1);
|
|
1337
1338
|
rb_define_method(rb_cLLaMAModel, "empty?", RUBY_METHOD_FUNC(_llama_model_empty), 0);
|
|
1338
1339
|
rb_define_method(rb_cLLaMAModel, "free", RUBY_METHOD_FUNC(_llama_model_free), 0);
|
|
@@ -3071,6 +3072,7 @@ static VALUE rb_llama_max_devices(VALUE self) {
|
|
|
3071
3072
|
extern "C" void Init_llama_cpp(void) {
|
|
3072
3073
|
rb_mLLaMACpp = rb_define_module("LLaMACpp");
|
|
3073
3074
|
|
|
3075
|
+
RbLLaMABatch::define_class(rb_mLLaMACpp);
|
|
3074
3076
|
RbLLaMATokenData::define_class(rb_mLLaMACpp);
|
|
3075
3077
|
RbLLaMATokenDataArray::define_class(rb_mLLaMACpp);
|
|
3076
3078
|
RbLLaMAModel::define_class(rb_mLLaMACpp);
|
|
@@ -43,7 +43,7 @@ GGML_API size_t ggml_allocr_alloc_graph(ggml_allocr_t alloc, struct ggml_cgraph
|
|
|
43
43
|
// ggml-backend v2 API
|
|
44
44
|
//
|
|
45
45
|
|
|
46
|
-
//
|
|
46
|
+
// Separate tensor and graph allocator objects
|
|
47
47
|
// This is necessary for multi-backend allocation because the graph allocator needs to use multiple tensor allocators
|
|
48
48
|
// The original API is kept as a wrapper around the new API
|
|
49
49
|
|