llama_cpp 0.12.7 → 0.14.0
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- data/CHANGELOG.md +24 -0
- data/ext/llama_cpp/llama_cpp.cpp +131 -288
- data/lib/llama_cpp/version.rb +2 -2
- data/sig/llama_cpp.rbs +29 -29
- data/vendor/tmp/llama.cpp/Makefile +10 -6
- data/vendor/tmp/llama.cpp/ggml-backend-impl.h +6 -3
- data/vendor/tmp/llama.cpp/ggml-backend.c +32 -23
- data/vendor/tmp/llama.cpp/ggml-backend.h +17 -16
- data/vendor/tmp/llama.cpp/ggml-cuda.cu +949 -168
- data/vendor/tmp/llama.cpp/ggml-kompute.cpp +9 -3
- data/vendor/tmp/llama.cpp/ggml-metal.m +159 -22
- data/vendor/tmp/llama.cpp/ggml-metal.metal +1195 -139
- data/vendor/tmp/llama.cpp/ggml-opencl.cpp +27 -27
- data/vendor/tmp/llama.cpp/ggml-quants.c +1971 -271
- data/vendor/tmp/llama.cpp/ggml-quants.h +52 -0
- data/vendor/tmp/llama.cpp/ggml-sycl.cpp +3586 -1201
- data/vendor/tmp/llama.cpp/ggml-sycl.h +5 -0
- data/vendor/tmp/llama.cpp/ggml-vulkan-shaders.hpp +39336 -43461
- data/vendor/tmp/llama.cpp/ggml-vulkan.cpp +1391 -825
- data/vendor/tmp/llama.cpp/ggml-vulkan.h +1 -0
- data/vendor/tmp/llama.cpp/ggml.c +545 -210
- data/vendor/tmp/llama.cpp/ggml.h +65 -23
- data/vendor/tmp/llama.cpp/llama.cpp +1458 -763
- data/vendor/tmp/llama.cpp/llama.h +81 -75
- data/vendor/tmp/llama.cpp/unicode.h +310 -1
- 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: c7d855ccd32ae097f26a671751d6a2178361cf8d8a6c1b99af37859f2c47ca03
|
|
4
|
+
data.tar.gz: 3b17318424d08c65ad34da3fa14956c86db0a2ea05ac174323a9b8d2b9e69d59
|
|
5
5
|
SHA512:
|
|
6
|
-
metadata.gz:
|
|
7
|
-
data.tar.gz:
|
|
6
|
+
metadata.gz: 2d90bf9fdd8dbaf5e67b7fb8797a9412168ae6ce5fcfc4c6aca34e194d5beb5204184b5bb36d65dc507a7a618ac9e938987e8d8bf5871e4eb6304b5e6de06020
|
|
7
|
+
data.tar.gz: eab524367ace146eb6e20786bd530cead145e1651bcdb726afbb5364609d04b22ca8a515016bb0c2d154ea97fb62f19222c122bc9bb5efe7fc389a6f259da6f0
|
data/CHANGELOG.md
CHANGED
|
@@ -1,3 +1,27 @@
|
|
|
1
|
+
## [[0.14.0](https://github.com/yoshoku/llama_cpp.rb/compare/v0.13.0...v0.14.0)] - 2024-03-09
|
|
2
|
+
|
|
3
|
+
**Breaking Changes**
|
|
4
|
+
|
|
5
|
+
- Bump bundled llama.cpp from b2303 to b2361.
|
|
6
|
+
- Rename embedding accessor to `embeddings` in `ContextParams`.
|
|
7
|
+
- Remove `do_pooling` accessor from `ContextParams`.
|
|
8
|
+
- Add `pooling_type` accessor to `ContextParams`.
|
|
9
|
+
- Fix the size of array returned by `embedding` method in `Context` from `n_embd` to `n_tokens * n_embd`.
|
|
10
|
+
- Add `embeddings_seq` method to `Context`.
|
|
11
|
+
|
|
12
|
+
## [[0.13.0](https://github.com/yoshoku/llama_cpp.rb/compare/v0.12.7...v0.13.0)] - 2024-03-02
|
|
13
|
+
|
|
14
|
+
**Breaking Changes**
|
|
15
|
+
|
|
16
|
+
- Bump bundled llama.cpp from b2143 to b2303.
|
|
17
|
+
- Remove deprecated methods:
|
|
18
|
+
- `map_supported?`, `mlock_supported?`, `apply_lora_from_file`, `eval`, `eval_embd`, `sample_classifier_free_guidance`, `sample_temperature`, and `mul_mat_q`.
|
|
19
|
+
- Rename some constants.
|
|
20
|
+
- Rename `kv_cache_seq_shift` method to `kv_cache_seq_add`.
|
|
21
|
+
- Add `defrag_thold` accessor to `ContextParams`.
|
|
22
|
+
- Add `vocab_type` and `rope_type` methods to `Model`.
|
|
23
|
+
- Add `kv_cache_seq_pos_max`, `kv_cache_defrag`, and `kv_cache_update` methods to `Context`.
|
|
24
|
+
|
|
1
25
|
## [[0.12.7](https://github.com/yoshoku/llama_cpp.rb/compare/v0.12.6...v0.12.7)] - 2024-02-24
|
|
2
26
|
|
|
3
27
|
- Bump bundled llama.cpp from b2106 to b2143.
|