llama_cpp 0.15.1 → 0.15.2
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/lib/llama_cpp/version.rb +2 -2
- data/vendor/tmp/llama.cpp/Makefile +3 -3
- data/vendor/tmp/llama.cpp/ggml-backend.c +2 -3
- data/vendor/tmp/llama.cpp/ggml-cuda.cu +15 -7
- data/vendor/tmp/llama.cpp/ggml-impl.h +7 -0
- data/vendor/tmp/llama.cpp/ggml-kompute.cpp +9 -3
- data/vendor/tmp/llama.cpp/ggml-metal.m +114 -125
- data/vendor/tmp/llama.cpp/ggml-metal.metal +86 -109
- data/vendor/tmp/llama.cpp/ggml-quants.c +2202 -28
- data/vendor/tmp/llama.cpp/ggml-rpc.cpp +1032 -0
- data/vendor/tmp/llama.cpp/ggml-rpc.h +24 -0
- data/vendor/tmp/llama.cpp/ggml-sycl.cpp +24 -143
- data/vendor/tmp/llama.cpp/ggml-vulkan.cpp +4 -2
- data/vendor/tmp/llama.cpp/ggml.c +726 -646
- data/vendor/tmp/llama.cpp/ggml.h +28 -17
- data/vendor/tmp/llama.cpp/llama.cpp +478 -281
- data/vendor/tmp/llama.cpp/llama.h +3 -0
- data/vendor/tmp/llama.cpp/unicode-data.cpp +6969 -2169
- data/vendor/tmp/llama.cpp/unicode-data.h +15 -12
- data/vendor/tmp/llama.cpp/unicode.cpp +89 -111
- data/vendor/tmp/llama.cpp/unicode.h +44 -12
- metadata +4 -2
|
@@ -242,6 +242,9 @@ extern "C" {
|
|
|
242
242
|
// proportion of the model (layers or rows) to offload to each GPU, size: llama_max_devices()
|
|
243
243
|
const float * tensor_split;
|
|
244
244
|
|
|
245
|
+
// comma separated list of RPC servers to use for offloading
|
|
246
|
+
const char * rpc_servers;
|
|
247
|
+
|
|
245
248
|
// Called with a progress value between 0.0 and 1.0. Pass NULL to disable.
|
|
246
249
|
// If the provided progress_callback returns true, model loading continues.
|
|
247
250
|
// If it returns false, model loading is immediately aborted.
|