llama_cpp 0.9.2 → 0.9.3
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 +7 -1
- data/ext/llama_cpp/llama_cpp.cpp +12 -0
- data/ext/llama_cpp/src/ggml-alloc.c +378 -208
- data/ext/llama_cpp/src/ggml-alloc.h +68 -16
- data/ext/llama_cpp/src/ggml-backend-impl.h +87 -0
- data/ext/llama_cpp/src/ggml-backend.c +578 -13
- data/ext/llama_cpp/src/ggml-backend.h +70 -77
- data/ext/llama_cpp/src/ggml-cuda.cu +194 -8
- data/ext/llama_cpp/src/ggml-impl.h +13 -7
- data/ext/llama_cpp/src/ggml-metal.h +1 -1
- data/ext/llama_cpp/src/ggml-metal.m +113 -32
- data/ext/llama_cpp/src/ggml-metal.metal +107 -1
- data/ext/llama_cpp/src/ggml-quants.c +173 -73
- data/ext/llama_cpp/src/ggml.c +826 -1482
- data/ext/llama_cpp/src/ggml.h +63 -45
- data/ext/llama_cpp/src/llama.cpp +364 -38
- data/ext/llama_cpp/src/llama.h +6 -0
- data/lib/llama_cpp/version.rb +2 -2
- data/sig/llama_cpp.rbs +2 -0
- metadata +3 -2
checksums.yaml
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
---
|
|
2
2
|
SHA256:
|
|
3
|
-
metadata.gz:
|
|
4
|
-
data.tar.gz:
|
|
3
|
+
metadata.gz: 2491ee80a5e822375f140b7d465a6783be62ef9f98aa510495723bd2d80b3f81
|
|
4
|
+
data.tar.gz: ad9ddbda1470602b976231edae030efd1ef0d854b41e0ce509e9b07ec78113e9
|
|
5
5
|
SHA512:
|
|
6
|
-
metadata.gz:
|
|
7
|
-
data.tar.gz:
|
|
6
|
+
metadata.gz: 8a82ed440ae2bbe20f2c3818f22f88f1c5cab659060ad085a43ee657d1e60919acb74b9aac9b1d027fe84ddb30d170efc0e3799d33deddc59b4d34300332a798
|
|
7
|
+
data.tar.gz: 164b4356580f0d2f17582fb84d59f0fbb9f816ac18921ea67d7cdda7f484620b605fdb88111ee32c1a42400c0770c520841304f7c2230ba577f4df1e5db453a0
|
data/CHANGELOG.md
CHANGED
|
@@ -1,3 +1,9 @@
|
|
|
1
|
+
## [[0.9.3](https://github.com/yoshoku/llama_cpp.rb/compare/v0.9.2...v0.9.3)] - 2023-11-18
|
|
2
|
+
|
|
3
|
+
- Bump bundled llama.cpp from b1500 to b1523.
|
|
4
|
+
- Add `add_bos_token?` method to Model.
|
|
5
|
+
- Add `add_eos_token?` method to Model.
|
|
6
|
+
|
|
1
7
|
## [[0.9.2](https://github.com/yoshoku/llama_cpp.rb/compare/v0.9.1...v0.9.2)] - 2023-11-11
|
|
2
8
|
|
|
3
9
|
- Bump bundled llama.cpp from b1472 to b1500.
|
|
@@ -6,7 +12,7 @@
|
|
|
6
12
|
|
|
7
13
|
- Bump bundled llama.cpp from b1429 to b1472
|
|
8
14
|
- Rename `kv_cahe_tokens_rm` method to `kv_cahce_clear` in Context.
|
|
9
|
-
- Add `sample_min_p method
|
|
15
|
+
- Add `sample_min_p` method to Context.
|
|
10
16
|
- Add `rope_scaling_type`, `rope_freq_base`, `rope_freq_scale`, `yarn_ext_factor`, `yarn_attn_factor`, `yarn_beta_fast`, `yarn_beta_slow`, and `yarn_orig_ctx` to ContextParams.
|
|
11
17
|
- Add `pure` to ModelQuantizeParams.
|
|
12
18
|
- Add contstants for RoPE scaling type.
|
data/ext/llama_cpp/llama_cpp.cpp
CHANGED
|
@@ -1252,6 +1252,8 @@ public:
|
|
|
1252
1252
|
rb_define_method(rb_cLLaMAModel, "token_bos", RUBY_METHOD_FUNC(_llama_model_token_bos), 0);
|
|
1253
1253
|
rb_define_method(rb_cLLaMAModel, "token_eos", RUBY_METHOD_FUNC(_llama_model_token_eos), 0);
|
|
1254
1254
|
rb_define_method(rb_cLLaMAModel, "token_nl", RUBY_METHOD_FUNC(_llama_model_token_nl), 0);
|
|
1255
|
+
rb_define_method(rb_cLLaMAModel, "add_bos_token?", RUBY_METHOD_FUNC(_llama_model_add_bos_token), 0);
|
|
1256
|
+
rb_define_method(rb_cLLaMAModel, "add_eos_token?", RUBY_METHOD_FUNC(_llama_model_add_eos_token), 0);
|
|
1255
1257
|
rb_define_method(rb_cLLaMAModel, "token_prefix", RUBY_METHOD_FUNC(_llama_model_token_prefix), 0);
|
|
1256
1258
|
rb_define_method(rb_cLLaMAModel, "token_middle", RUBY_METHOD_FUNC(_llama_model_token_middle), 0);
|
|
1257
1259
|
rb_define_method(rb_cLLaMAModel, "token_suffix", RUBY_METHOD_FUNC(_llama_model_token_suffix), 0);
|
|
@@ -1541,6 +1543,16 @@ private:
|
|
|
1541
1543
|
return INT2NUM(llama_token_nl(ptr->model));
|
|
1542
1544
|
}
|
|
1543
1545
|
|
|
1546
|
+
static VALUE _llama_model_add_bos_token(VALUE self) {
|
|
1547
|
+
LLaMAModelWrapper* ptr = get_llama_model(self);
|
|
1548
|
+
return llama_add_bos_token(ptr->model) ? Qtrue : Qfalse;
|
|
1549
|
+
}
|
|
1550
|
+
|
|
1551
|
+
static VALUE _llama_model_add_eos_token(VALUE self) {
|
|
1552
|
+
LLaMAModelWrapper* ptr = get_llama_model(self);
|
|
1553
|
+
return llama_add_eos_token(ptr->model) ? Qtrue : Qfalse;
|
|
1554
|
+
}
|
|
1555
|
+
|
|
1544
1556
|
static VALUE _llama_model_token_prefix(VALUE self) {
|
|
1545
1557
|
LLaMAModelWrapper* ptr = get_llama_model(self);
|
|
1546
1558
|
return INT2NUM(llama_token_prefix(ptr->model));
|