llama_cpp 0.16.1 → 0.17.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: '0408c7d579262d0ba34013671a446a76513f6f4564270ef8ba471203fba75c59'
-  data.tar.gz: a8085d9678999bb79ba788f7ce856c3f4fe1c6e131af569eaa54aa71fc9ae773
+  metadata.gz: a63238d7d4a852e4a57667ba3e144364db201a691b9460c62fc8aa783677593d
+  data.tar.gz: 7a879c04eebc5a308ae3f937f35972b11c5d15edd5000885416e3c57cfe21648
 SHA512:
-  metadata.gz: 32e8f294a7f88db05abba3d1e11c951a38f366cac83712f89aa68ed95c581d8eaa4df3d5473f1af3cee965d7a66ea2bf5ccb00222337c59d97fca32ba5e9cade
-  data.tar.gz: 2c5f66e2902eb1d72e45261e80f1a2599534e02d0e54e83ff432414d355fab67f6bda9eee095f17904a4fb2d3bf5cb4a1509a346fadff13d999c2f057db972c6
+  metadata.gz: a76006fc44d8a7b4295c4d10bcee87a2f161868b9c119ddfae1c2aecd0a5d7989bd33134dc64d8f1994b41732a64e2ca91472a8245ee58e3fb4fdcb01a1b24f2
+  data.tar.gz: 63160f285f7fdb89e6d03e9cb83b064acbe8869ae384f9b3d32f0a822d7fc63354cf0fb6b6da39758140d885493baff716d31c42a956e3437c47adaf74172783

data/CHANGELOG.md

@@ -1,3 +1,28 @@
+## [[0.17.0](https://github.com/yoshoku/llama_cpp.rb/compare/v0.16.2...v0.17.0)] - 2024-06-29
+
+**Breaking Changes**
+
+I stopped including the llama.cpp source code in the gem,
+as it became difficult to keep up with changes in the llama.cpp file structure.
+You need to install the llama.cpp library separately.
+If you are using homebrew on macOS, the following command will install the library:
+
+```sh
+$ brew install llama.cpp
+$ gem install llama_cpp -- --with-opt-dir=/opt/homebrew
+```
+
+- Change supported llama.cpp version to b3265
+ - Add `LLAMA_VOCAB_TYPE_UGM` and `LLAMA_VOCAB_PRE_TYPE_VIKING` constants.
+ - Add `token_pad` method to `Model`.
+
+## [[0.16.2](https://github.com/yoshoku/llama_cpp.rb/compare/v0.16.1...v0.16.2)] - 2024-06-22
+
+- Bump llama.cpp from b3151 to b3197.
+  - Add `LLAMA_POOLING_TYPE_LAST` constant.
+  - Add `--with-vulkan-memory-debug` config option.
+  - Add `set_embeddings` method to `Context`.
+
 ## [[0.16.1](https://github.com/yoshoku/llama_cpp.rb/compare/v0.16.0...v0.16.1)] - 2024-06-15
 
 - Bump llama.cpp from b3091 to b3151.

data/README.md

@@ -10,30 +10,25 @@ This gem is still under development and may undergo many changes in the future.
 
 ## Installation
 
-Install the gem and add to the application's Gemfile by executing:
+Install the llama.cpp. If you use homebrew, install it by executing:
 
 ```sh
-$ bundle add llama_cpp
+$ brew install llama.cpp
 ```
 
-If bundler is not being used to manage dependencies, install the gem by executing:
+Install the gem and add to the application's Gemfile by executing:
 
 ```sh
-$ gem install llama_cpp
+$ bundle config --local build.llama_cpp "--with-opt-dir=/opt/homebrew/"
+$ bundle add llama_cpp
 ```
 
-There are several installation options:
+If bundler is not being used to manage dependencies, install the gem by executing:
 
 ```sh
-# use OpenBLAS
-$ gem install llama_cpp -- --with-openblas
-
-# use CUDA
-$ gem install llama_cpp -- --with-cuda
+$ gem install llama_cpp -- --with-opt-dir=/opt/homebrew
 ```
 
-Those options are defined in [extconf.rb](https://github.com/yoshoku/llama_cpp.rb/blob/main/ext/llama_cpp/extconf.rb) by with_config method.
-
 ## Usage
 
 Prepare the quantized model by refering to [the usage section on the llama.cpp README](https://github.com/ggerganov/llama.cpp#usage).

data/ext/llama_cpp/extconf.rb

@@ -1,50 +1,10 @@
 # frozen_string_literal: true
 
 require 'mkmf'
-require 'fileutils'
-require 'open3'
-
-VENDOR_DIR = File.expand_path("#{__dir__}/../../vendor")
-VENDOR_LIB_DIR = "#{VENDOR_DIR}/lib"
-VENDOR_INC_DIR = "#{VENDOR_DIR}/include"
-LLAMA_CPP_DIR = "#{VENDOR_DIR}/tmp/llama.cpp"
-
-make_envs = +''
-make_envs << ' LLAMA_DEBUG=1' if with_config('debug')
-make_envs << ' LLAMA_QKK_64=1' if with_config('qkk-64')
-make_envs << ' LLAMA_NO_ACCELERATE=1' if with_config('no-accelerate')
-make_envs << ' LLAMA_OPENBLAS=1' if with_config('openblas')
-make_envs << ' LLAMA_OPENBLAS64=1' if with_config('openblas64')
-make_envs << ' LLAMA_BLIS=1' if with_config('blis')
-make_envs << ' LLAMA_CUBLAS=1' if with_config('cublas') # Deprecated, use --with-cuda instead
-make_envs << ' LLAMA_CUDA=1' if with_config('cuda')
-make_envs << ' LLAMA_HIPBLAS=1' if with_config('hipblas')
-make_envs << ' LLAMA_VULKAN=1' if with_config('vulkan')
-make_envs << ' LLAMA_NO_OPENMP=1' if with_config('no-openmp')
-make_envs << ' LLAMA_NO_LLAMAFILE=1' if with_config('no-llamafile')
-
-make_envs << ' LLAMA_METAL_EMBED_LIBRARY=1' if RUBY_PLATFORM.match?(/darwin/)
-
-Dir.chdir(LLAMA_CPP_DIR) do
-  _mkstdout, _mkstderr, mkstatus = Open3.capture3("make lib #{make_envs}".strip)
-  abort('Failed to build llama.cpp.') unless mkstatus.success?
-
-  FileUtils.cp(Dir.glob('libllama.*'), VENDOR_LIB_DIR)
-  FileUtils.cp(Dir.glob('*.h'), "#{VENDOR_DIR}/include/")
-end
-
-if RUBY_PLATFORM.match?(/darwin/)
-  Dir.chdir(VENDOR_LIB_DIR) do
-    _mkstdout, _mkstderr, mkstatus = Open3.capture3("install_name_tool -id #{VENDOR_LIB_DIR}/libllama.dylib libllama.dylib")
-    abort('Failed to set installation path for libllama.dylib.') unless mkstatus.success?
-  end
-  FileUtils.cp("#{LLAMA_CPP_DIR}/ggml-metal-embed.metal", VENDOR_LIB_DIR)
-  FileUtils.cp("#{LLAMA_CPP_DIR}/ggml-metal.metal", VENDOR_LIB_DIR)
-end
 
 abort('libstdc++ is not found.') unless have_library('stdc++')
-abort('libllama is not found.') unless find_library('llama', nil, VENDOR_LIB_DIR)
-abort('llama.h is not found.') unless find_header('llama.h', nil, VENDOR_INC_DIR)
+abort('libllama is not found.') unless have_library('llama')
+abort('llama.h is not found.') unless have_header('llama.h')
 
 $CXXFLAGS << ' -std=c++11'
 

data/ext/llama_cpp/llama_cpp.cpp

@@ -1529,6 +1529,7 @@ public:
     rb_define_method(rb_cLLaMAModel, "token_cls", RUBY_METHOD_FUNC(_llama_model_token_cls), 0);
     rb_define_method(rb_cLLaMAModel, "token_sep", RUBY_METHOD_FUNC(_llama_model_token_sep), 0);
     rb_define_method(rb_cLLaMAModel, "token_nl", RUBY_METHOD_FUNC(_llama_model_token_nl), 0);
+    rb_define_method(rb_cLLaMAModel, "token_pad", RUBY_METHOD_FUNC(_llama_model_token_pad), 0);
     rb_define_method(rb_cLLaMAModel, "add_bos_token?", RUBY_METHOD_FUNC(_llama_model_add_bos_token), 0);
     rb_define_method(rb_cLLaMAModel, "add_eos_token?", RUBY_METHOD_FUNC(_llama_model_add_eos_token), 0);
     rb_define_method(rb_cLLaMAModel, "token_prefix", RUBY_METHOD_FUNC(_llama_model_token_prefix), 0);
@@ -1810,6 +1811,11 @@ private:
     return INT2NUM(llama_token_nl(ptr->model));
   }
 
+  static VALUE _llama_model_token_pad(VALUE self) {
+    LLaMAModelWrapper* ptr = get_llama_model(self);
+    return INT2NUM(llama_token_pad(ptr->model));
+  }
+
   static VALUE _llama_model_add_bos_token(VALUE self) {
     LLaMAModelWrapper* ptr = get_llama_model(self);
     return llama_add_bos_token(ptr->model) ? Qtrue : Qfalse;
@@ -2133,6 +2139,7 @@ public:
     rb_define_method(rb_cLLaMAContext, "embeddings", RUBY_METHOD_FUNC(_llama_context_embeddings), 0);
     rb_define_method(rb_cLLaMAContext, "embeddings_ith", RUBY_METHOD_FUNC(_llama_context_embeddings_ith), 1);
     rb_define_method(rb_cLLaMAContext, "embeddings_seq", RUBY_METHOD_FUNC(_llama_context_embeddings_seq), 1);
+    rb_define_method(rb_cLLaMAContext, "set_embeddings", RUBY_METHOD_FUNC(_llama_context_set_embeddings), 1);
     rb_define_method(rb_cLLaMAContext, "set_n_threads", RUBY_METHOD_FUNC(_llama_context_set_n_threads), -1);
     rb_define_method(rb_cLLaMAContext, "n_ctx", RUBY_METHOD_FUNC(_llama_context_n_ctx), 0);
     rb_define_method(rb_cLLaMAContext, "n_batch", RUBY_METHOD_FUNC(_llama_context_n_batch), 0);
@@ -2357,6 +2364,16 @@ private:
     return output;
   }
 
+  static VALUE _llama_context_set_embeddings(VALUE self, VALUE embs) {
+    LLaMAContextWrapper* ptr = get_llama_context(self);
+    if (ptr->ctx == NULL) {
+      rb_raise(rb_eArgError, "LLaMA context is not initialized");
+      return Qnil;
+    }
+    llama_set_embeddings(ptr->ctx, RTEST(embs) ? true : false);
+    return Qnil;
+  }
+
   static VALUE _llama_context_set_n_threads(int argc, VALUE* argv, VALUE self) {
     VALUE kw_args = Qnil;
     ID kw_table[2] = { rb_intern("n_threads"), rb_intern("n_threads_batch") };
@@ -3478,6 +3495,7 @@ extern "C" void Init_llama_cpp(void) {
   rb_define_const(rb_mLLaMACpp, "LLAMA_VOCAB_TYPE_SPM", INT2NUM(LLAMA_VOCAB_TYPE_SPM));
   rb_define_const(rb_mLLaMACpp, "LLAMA_VOCAB_TYPE_BPE", INT2NUM(LLAMA_VOCAB_TYPE_BPE));
   rb_define_const(rb_mLLaMACpp, "LLAMA_VOCAB_TYPE_WPM", INT2NUM(LLAMA_VOCAB_TYPE_WPM));
+  rb_define_const(rb_mLLaMACpp, "LLAMA_VOCAB_TYPE_UGM", INT2NUM(LLAMA_VOCAB_TYPE_UGM));
 
   rb_define_const(rb_mLLaMACpp, "LLAMA_VOCAB_PRE_TYPE_DEFAULT", INT2NUM(LLAMA_VOCAB_PRE_TYPE_DEFAULT));
   rb_define_const(rb_mLLaMACpp, "LLAMA_VOCAB_PRE_TYPE_LLAMA3", INT2NUM(LLAMA_VOCAB_PRE_TYPE_LLAMA3));
@@ -3495,6 +3513,7 @@ extern "C" void Init_llama_cpp(void) {
   rb_define_const(rb_mLLaMACpp, "LLAMA_VOCAB_PRE_TYPE_DBRX", INT2NUM(LLAMA_VOCAB_PRE_TYPE_DBRX));
   rb_define_const(rb_mLLaMACpp, "LLAMA_VOCAB_PRE_TYPE_SMAUG", INT2NUM(LLAMA_VOCAB_PRE_TYPE_SMAUG));
   rb_define_const(rb_mLLaMACpp, "LLAMA_VOCAB_PRE_TYPE_PORO", INT2NUM(LLAMA_VOCAB_PRE_TYPE_PORO));
+  rb_define_const(rb_mLLaMACpp, "LLAMA_VOCAB_PRE_TYPE_VIKING", INT2NUM(LLAMA_VOCAB_PRE_TYPE_VIKING));
 
   rb_define_const(rb_mLLaMACpp, "LLAMA_TOKEN_TYPE_UNDEFINED", INT2NUM(LLAMA_TOKEN_TYPE_UNDEFINED));
   rb_define_const(rb_mLLaMACpp, "LLAMA_TOKEN_TYPE_NORMAL", INT2NUM(LLAMA_TOKEN_TYPE_NORMAL));
@@ -3572,6 +3591,7 @@ extern "C" void Init_llama_cpp(void) {
   rb_define_const(rb_mLLaMACpp, "LLAMA_POOLING_TYPE_NONE", INT2NUM(LLAMA_POOLING_TYPE_NONE));
   rb_define_const(rb_mLLaMACpp, "LLAMA_POOLING_TYPE_MEAN", INT2NUM(LLAMA_POOLING_TYPE_MEAN));
   rb_define_const(rb_mLLaMACpp, "LLAMA_POOLING_TYPE_CLS", INT2NUM(LLAMA_POOLING_TYPE_CLS));
+  rb_define_const(rb_mLLaMACpp, "LLAMA_POOLING_TYPE_LAST", INT2NUM(LLAMA_POOLING_TYPE_LAST));
 
   rb_define_const(rb_mLLaMACpp, "LLAMA_SPLIT_MODE_NONE", INT2NUM(LLAMA_SPLIT_MODE_NONE));
   rb_define_const(rb_mLLaMACpp, "LLAMA_SPLIT_MODE_LAYER", INT2NUM(LLAMA_SPLIT_MODE_LAYER));

data/lib/llama_cpp/version.rb

@@ -3,8 +3,8 @@
 # llama_cpp.rb provides Ruby bindings for the llama.cpp.
 module LLaMACpp
   # The version of llama_cpp.rb you install.
-  VERSION = '0.16.1'
+  VERSION = '0.17.0'
 
-  # The version of llama.cpp bundled with llama_cpp.rb.
-  LLAMA_CPP_VERSION = 'b3151'
+  # The supported version of llama.cpp.
+  LLAMA_CPP_VERSION = 'b3265'
 end

data/sig/llama_cpp.rbs

@@ -15,6 +15,7 @@ module LLaMACpp
   LLAMA_VOCAB_TYPE_SPM: Integer
   LLAMA_VOCAB_TYPE_BPE: Integer
   LLAMA_VOCAB_TYPE_WPM: Integer
+  LLAMA_VOCAB_TYPE_UGM: Integer
 
   LLAMA_VOCAB_PRE_TYPE_DEFAULT: Integer
   LLAMA_VOCAB_PRE_TYPE_LLAMA3: Integer
@@ -32,6 +33,7 @@ module LLaMACpp
   LLAMA_VOCAB_PRE_TYPE_DBRX: Integer
   LLAMA_VOCAB_PRE_TYPE_SMAUG: Integer
   LLAMA_VOCAB_PRE_TYPE_PORO: Integer
+  LLAMA_VOCAB_PRE_TYPE_VIKING: Integer
 
   LLAMA_TOKEN_ATTR_UNDEFINED: Integer
   LLAMA_TOKEN_ATTR_UNKNOWN: Integer
@@ -99,6 +101,7 @@ module LLaMACpp
   LLAMA_POOLING_TYPE_NONE: Integer
   LLAMA_POOLING_TYPE_MEAN: Integer
   LLAMA_POOLING_TYPE_CLS: Integer
+  LLAMA_POOLING_TYPE_LAST: Integer
 
   LLAMA_SPLIT_MODE_NONE: Integer
   LLAMA_SPLIT_MODE_LAYER: Integer
@@ -167,6 +170,7 @@ module LLaMACpp
     def token_cls: () -> Integer
     def token_sep: () -> Integer
     def token_nl: () -> Integer
+    def token_pad: () -> Integer
     def add_bos_token?: () -> bool
     def add_eos_token?: () -> bool
     def token_prefix: () -> Integer
@@ -258,6 +262,7 @@ module LLaMACpp
     def embeddings_seq: (Integer) -> Array[Float]
     def decode: (::LLaMACpp::Batch) -> void
     def logits: () -> Array[Float]
+    def set_embeddings: (bool) -> void
     def set_n_threads: (n_threads: Integer, n_threads_batch: Integer) -> void
     def n_ctx: () -> Integer
     def n_batch: () -> Integer

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: llama_cpp
 version: !ruby/object:Gem::Version
-  version: 0.16.1
+  version: 0.17.0
 platform: ruby
 authors:
 - yoshoku
 autorequire:
 bindir: exe
 cert_chain: []
-date: 2024-06-15 00:00:00.000000000 Z
+date: 2024-06-29 00:00:00.000000000 Z
 dependencies: []
 description: llama_cpp.rb provides Ruby bindings for the llama.cpp.
 email:
@@ -33,175 +33,6 @@ files:
 - lib/llama_cpp.rb
 - lib/llama_cpp/version.rb
 - sig/llama_cpp.rbs
-- vendor/include/.gitkeep
-- vendor/lib/.gitkeep
-- vendor/tmp/llama.cpp/LICENSE
-- vendor/tmp/llama.cpp/Makefile
-- vendor/tmp/llama.cpp/ggml-alloc.c
-- vendor/tmp/llama.cpp/ggml-alloc.h
-- vendor/tmp/llama.cpp/ggml-backend-impl.h
-- vendor/tmp/llama.cpp/ggml-backend.c
-- vendor/tmp/llama.cpp/ggml-backend.h
-- vendor/tmp/llama.cpp/ggml-blas.cpp
-- vendor/tmp/llama.cpp/ggml-blas.h
-- vendor/tmp/llama.cpp/ggml-common.h
-- vendor/tmp/llama.cpp/ggml-cuda.cu
-- vendor/tmp/llama.cpp/ggml-cuda.h
-- vendor/tmp/llama.cpp/ggml-cuda/acc.cu
-- vendor/tmp/llama.cpp/ggml-cuda/arange.cu
-- vendor/tmp/llama.cpp/ggml-cuda/argsort.cu
-- vendor/tmp/llama.cpp/ggml-cuda/binbcast.cu
-- vendor/tmp/llama.cpp/ggml-cuda/clamp.cu
-- vendor/tmp/llama.cpp/ggml-cuda/concat.cu
-- vendor/tmp/llama.cpp/ggml-cuda/convert.cu
-- vendor/tmp/llama.cpp/ggml-cuda/cpy.cu
-- vendor/tmp/llama.cpp/ggml-cuda/diagmask.cu
-- vendor/tmp/llama.cpp/ggml-cuda/dmmv.cu
-- vendor/tmp/llama.cpp/ggml-cuda/fattn-tile-f16.cu
-- vendor/tmp/llama.cpp/ggml-cuda/fattn-tile-f32.cu
-- vendor/tmp/llama.cpp/ggml-cuda/fattn.cu
-- vendor/tmp/llama.cpp/ggml-cuda/getrows.cu
-- vendor/tmp/llama.cpp/ggml-cuda/im2col.cu
-- vendor/tmp/llama.cpp/ggml-cuda/mmq.cu
-- vendor/tmp/llama.cpp/ggml-cuda/mmvq.cu
-- vendor/tmp/llama.cpp/ggml-cuda/norm.cu
-- vendor/tmp/llama.cpp/ggml-cuda/pad.cu
-- vendor/tmp/llama.cpp/ggml-cuda/pool2d.cu
-- vendor/tmp/llama.cpp/ggml-cuda/quantize.cu
-- vendor/tmp/llama.cpp/ggml-cuda/rope.cu
-- vendor/tmp/llama.cpp/ggml-cuda/scale.cu
-- vendor/tmp/llama.cpp/ggml-cuda/softmax.cu
-- vendor/tmp/llama.cpp/ggml-cuda/sumrows.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-f16.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q4_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q5_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-f16-q8_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-f16.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q4_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q5_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_0-q8_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-f16.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q4_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q5_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q4_1-q8_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-f16.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q4_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q5_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_0-q8_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-f16.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q4_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q5_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q5_1-q8_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-f16.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q4_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q5_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs128-q8_0-q8_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs256-f16-f16.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-f16.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q4_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q5_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f16-instance-hs64-f16-q8_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-f16.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q4_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q5_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-f16-q8_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-f16.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q4_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q5_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_0-q8_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-f16.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q4_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q5_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q4_1-q8_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-f16.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q4_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q5_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_0-q8_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-f16.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q4_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q5_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q5_1-q8_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-f16.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q4_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q5_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs128-q8_0-q8_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs256-f16-f16.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-f16.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q4_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q5_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-vec-f32-instance-hs64-f16-q8_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqfloat-cpb16.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqfloat-cpb32.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqhalf-cpb16.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqhalf-cpb32.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/fattn-wmma-f16-instance-kqhalf-cpb8.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q2_k.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q3_k.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q4_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q4_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q4_k.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q5_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q5_1.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q5_k.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q6_k.cu
-- vendor/tmp/llama.cpp/ggml-cuda/template-instances/mmq-instance-q8_0.cu
-- vendor/tmp/llama.cpp/ggml-cuda/tsembd.cu
-- vendor/tmp/llama.cpp/ggml-cuda/unary.cu
-- vendor/tmp/llama.cpp/ggml-cuda/upscale.cu
-- vendor/tmp/llama.cpp/ggml-impl.h
-- vendor/tmp/llama.cpp/ggml-kompute.cpp
-- vendor/tmp/llama.cpp/ggml-kompute.h
-- vendor/tmp/llama.cpp/ggml-metal.h
-- vendor/tmp/llama.cpp/ggml-metal.m
-- vendor/tmp/llama.cpp/ggml-metal.metal
-- vendor/tmp/llama.cpp/ggml-quants.c
-- vendor/tmp/llama.cpp/ggml-quants.h
-- vendor/tmp/llama.cpp/ggml-rpc.cpp
-- vendor/tmp/llama.cpp/ggml-rpc.h
-- vendor/tmp/llama.cpp/ggml-sycl.cpp
-- vendor/tmp/llama.cpp/ggml-sycl.h
-- vendor/tmp/llama.cpp/ggml-vulkan-shaders.hpp
-- vendor/tmp/llama.cpp/ggml-vulkan.cpp
-- vendor/tmp/llama.cpp/ggml-vulkan.h
-- vendor/tmp/llama.cpp/ggml.c
-- vendor/tmp/llama.cpp/ggml.h
-- vendor/tmp/llama.cpp/llama.cpp
-- vendor/tmp/llama.cpp/llama.h
-- vendor/tmp/llama.cpp/scripts/get-flags.mk
-- vendor/tmp/llama.cpp/sgemm.cpp
-- vendor/tmp/llama.cpp/sgemm.h
-- vendor/tmp/llama.cpp/unicode-data.cpp
-- vendor/tmp/llama.cpp/unicode-data.h
-- vendor/tmp/llama.cpp/unicode.cpp
-- vendor/tmp/llama.cpp/unicode.h
 homepage: https://github.com/yoshoku/llama_cpp.rb
 licenses:
 - MIT

data/vendor/tmp/llama.cpp/LICENSE

@@ -1,21 +0,0 @@
-MIT License
-
-Copyright (c) 2023-2024 The ggml authors
-
-Permission is hereby granted, free of charge, to any person obtaining a copy
-of this software and associated documentation files (the "Software"), to deal
-in the Software without restriction, including without limitation the rights
-to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-copies of the Software, and to permit persons to whom the Software is
-furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice shall be included in all
-copies or substantial portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.