llama_cpp 0.28.0 → 0.28.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 002371b1977caa3b67be951da5098f478c540b0b85c02e49e0263579626e983b
-  data.tar.gz: da1b91e34d7c494f3a4661b85a3c6bda909389b9c40a223aac9675330619132e
+  metadata.gz: 53c9355352c9ef5fe64fd761f0ac33ff4966a92e753fa1a60a8c477b2e0f7194
+  data.tar.gz: 391826211e106323b684a38ea67bd7c446f8fec5eb1b3f0af02ff110298baf76
 SHA512:
-  metadata.gz: 53a2f89d9dbc2e3fe73de60dcd8c09218f9c4b7e49c89916619d74df3b7c245cd0252a3aead4f5caf918e86b21da5eb7a2630a56fe40a2fb6f01567b2a11bc92
-  data.tar.gz: 3f784cd9cfd6fa24d26ceafd643013b8b70e5c5610366719bea147f62d3febfee24b54ad1099027f279454f8ced4d0b8d0fd0859b54175a2222e311786898f00
+  metadata.gz: acd2678b5e508a86909561ae9f7440710e516fa9d241bb746771c990d034dbac31a3de95813b9876861a4459f824c6f4ea89399cbe7c7dc6e6bf8d9ad3a20fac
+  data.tar.gz: 05717ebcd77750c112ce1f77edfb2906846f12e0e8b321488e271737c858f3fa7c576da6675bd851cc2ec5af24fd26f35596fe24bb7aae6303e3d2cf05c0cb8d

data/CHANGELOG.md

@@ -1,8 +1,16 @@
+## [[0.28.1](https://github.com/yoshoku/llama_cpp.rb/compare/v0.28.0...v0.28.1)] - 2026-09-05
+
+- Change supported llama.cpp version to v0.4.0.
+  - Add `LLAMA_LAZY_MODE_OFF` constant value.
+  - Add `LLAMA_LAZY_MODE_AUTO` constant value.
+  - Add `LLAMA_LAZY_MODE_ON` constant value.
+  - Add `max_buf_size` accessor to `LlamaModelQuantizeParams`.
+
 ## [[0.28.0](https://github.com/yoshoku/llama_cpp.rb/compare/v0.27.0...v0.28.0)] - 2026-08-23
 
 - Change supported llama.cpp version to v0.2.0.
   - Add `LLAMA_LOAD_MODE_AUTO` constant value.
-  - Add `n_outputs_max_per_seq accessor` to `LlamaContextParams`.
+  - Add `n_outputs_max_per_seq` accessor to `LlamaContextParams`.
   - Add `llama_version` module function to `LlamaCpp`.
   - Add `llama_sampler_copy` module function to `LlamaCpp`.
 

data/ext/llama_cpp/llama_cpp.c

@@ -1153,6 +1153,17 @@ static VALUE llama_model_quantize_params_set_dry_run(VALUE self, VALUE dry_run)
   return dry_run;
 }
 
+static VALUE llama_model_quantize_params_get_max_buf_size(VALUE self) {
+  llama_model_quantize_params* data = get_llama_model_quantize_params(self);
+  return SIZET2NUM(data->max_buf_size);
+}
+
+static VALUE llama_model_quantize_params_set_max_buf_size(VALUE self, VALUE max_buf_size) {
+  llama_model_quantize_params* data = get_llama_model_quantize_params(self);
+  data->max_buf_size = NUM2SIZET(max_buf_size);
+  return max_buf_size;
+}
+
 /* llama_logit_bias */
 static void llama_logit_bias_free(void *ptr) {
   if (ptr) {
@@ -4583,6 +4594,11 @@ void Init_llama_cpp(void) {
   rb_define_const(rb_mLlamaCpp, "LLAMA_LOAD_MODE_MLOCK", INT2NUM(LLAMA_LOAD_MODE_MLOCK));
   rb_define_const(rb_mLlamaCpp, "LLAMA_LOAD_MODE_MMAP_MLOCK", INT2NUM(LLAMA_LOAD_MODE_MMAP_MLOCK));
   rb_define_const(rb_mLlamaCpp, "LLAMA_LOAD_MODE_DIRECT_IO", INT2NUM(LLAMA_LOAD_MODE_DIRECT_IO));
+  /* llama_lazy_mode */
+  /* Document-const: LlamaCpp::LLAMA_LAZY_MODE_OFF */
+  rb_define_const(rb_mLlamaCpp, "LLAMA_LAZY_MODE_OFF", INT2NUM(LLAMA_LAZY_MODE_OFF));
+  rb_define_const(rb_mLlamaCpp, "LLAMA_LAZY_MODE_AUTO", INT2NUM(LLAMA_LAZY_MODE_AUTO));
+  rb_define_const(rb_mLlamaCpp, "LLAMA_LAZY_MODE_ON", INT2NUM(LLAMA_LAZY_MODE_ON));
   /* llama_context_type */
   /* Document-const: LlamaCpp::LLAMA_CONTEXT_TYPE_DEFAULT */
   rb_define_const(rb_mLlamaCpp, "LLAMA_CONTEXT_TYPE_DEFAULT", INT2NUM(LLAMA_CONTEXT_TYPE_DEFAULT));
@@ -5352,6 +5368,17 @@ void Init_llama_cpp(void) {
   /* TODO: const struct llama_model_kv_override* kv_overrides */
   /* TODO: const struct llama_model_tensor_override* tt_overrides */
   /* TODO: const int32_t* prune_layers */
+  /**
+   * Document-method: max_buf_size
+   * @return [Integer]
+   */
+  rb_define_method(rb_cLlamaModelQuantizeParams, "max_buf_size", RUBY_METHOD_FUNC(llama_model_quantize_params_get_max_buf_size), 0);
+  /**
+   * Document-method: max_buf_size=
+   * @param [Integer] max_buf_size
+   * @return [Integer]
+   */
+  rb_define_method(rb_cLlamaModelQuantizeParams, "max_buf_size=", RUBY_METHOD_FUNC(llama_model_quantize_params_set_max_buf_size), 1);
 
   /**
    * Document-class: LlamaCpp::LlamaLogitBias

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 = '0.28.0'
+  VERSION = '0.28.1'
 
   # The supported version of llama.cpp.
-  LLAMA_CPP_VERSION = 'v0.2.0'
+  LLAMA_CPP_VERSION = 'v0.4.0'
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: llama_cpp
 version: !ruby/object:Gem::Version
-  version: 0.28.0
+  version: 0.28.1
 platform: ruby
 authors:
 - yoshoku
@@ -33,7 +33,7 @@ metadata:
   homepage_uri: https://github.com/yoshoku/llama_cpp.rb
   source_code_uri: https://github.com/yoshoku/llama_cpp.rb
   changelog_uri: https://github.com/yoshoku/llama_cpp.rb/blob/main/CHANGELOG.md
-  documentation_uri: https://gemdocs.org/gems/llama_cpp/0.28.0/
+  documentation_uri: https://gemdocs.org/gems/llama_cpp/0.28.1/
   rubygems_mfa_required: 'true'
 rdoc_options: []
 require_paths: