llama_cpp 0.24.3 → 0.25.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: e4ec06dde3c07e65bccca17e900a37042082fc706c8cddd34e447affe1198ba5
-  data.tar.gz: a4ed3057e3da8fc8be59ab578faf7fdd89773c17a6ee99d03cb25bb5a9af282d
+  metadata.gz: 953b205d7cedadd2f1db35fc301c6b94b1db87e0121317f6c154c204e09e9d56
+  data.tar.gz: bbbc1eef7f7312e667fd238b1e5ef353861beaa445c348595470879eccd12280
 SHA512:
-  metadata.gz: 71803b7ed68c4c9911bf1a91a63e5a5e8d588e297f8ba8d9a28665e06cc367c2eecf6b644033b57ab50358aacc4cb3c6c69f44c54a121dc521fd843e2d4ac727
-  data.tar.gz: f35964446f5d954531da4f62862007b97123e158211b01817ffba01409569a8f949b30cf2dc8b8915eebffcf90793978fa5d9d25abf3fbacee991ccfbef48ea8
+  metadata.gz: a38097de3f8e5a8acf862bf28db94140efc8f5e31d2be8e9fcbbf461ad79c77614a65acb873d25a706380056295b09cc14f94d90661e4c05e81a7c84a7cc461d
+  data.tar.gz: 2225bfdc2526274a6c3b60ec7c6ad0d88c62a291c847d511dd43772b7cc4c978cecbf69dd0fa50d10fafc3a915b95c44d8eebfdb14815777ace34318f1b28066

data/CHANGELOG.md

@@ -1,3 +1,13 @@
+## [[0.25.0](https://github.com/yoshoku/llama_cpp.rb/compare/v0.24.3...v0.25.0)] - 2026-04-25
+
+- Change supported llama.cpp version to b8920.
+  - Remove `LLAMA_PARAMS_FIT_STATUS_SUCCESS` constant value.
+  - Remove `LLAMA_PARAMS_FIT_STATUS_FAILURE` constant value.
+  - Remove `LLAMA_PARAMS_FIT_STATUS_ERROR` constant value.
+  - Remove `llama_memory_breakdown_print` module function.
+  - Add `LLAMA_FTYPE_MOSTLY_Q1_0` constant value.
+  - Add `LLAMA_SPLIT_MODE_TENSOR` constant value.
+
 ## [[0.24.2](https://github.com/yoshoku/llama_cpp.rb/compare/v0.24.2...v0.24.3)] - 2026-04-06
 
 - Change supported llama.cpp version to b8640.

data/ext/llama_cpp/llama_cpp.c

@@ -3375,22 +3375,6 @@ static VALUE rb_llama_sampler_reset(VALUE self, VALUE sampler) {
   return Qnil;
 }
 
-/**
- * @overload llama_memory_breakdown_print(context)
- *  @param [LlamaContext] context
- *  @return [NilClass]
- */
-static VALUE rb_llama_memory_breakdown_print(VALUE self, VALUE ctx) {
-  if (!rb_obj_is_kind_of(ctx, rb_cLlamaContext)) {
-    rb_raise(rb_eArgError, "ctx must be a LlamaContext");
-    return Qnil;
-  }
-  llama_context_wrapper* context_wrapper = get_llama_context_wrapper(ctx);
-  llama_memory_breakdown_print(context_wrapper->context);
-  RB_GC_GUARD(ctx);
-  return Qnil;
-}
-
 /**
  * @overload llama_sampler_clone(sampler)
  *  @param [LlamaSampler] sampler
@@ -4292,6 +4276,7 @@ void Init_llama_cpp(void) {
   rb_define_const(rb_mLlamaCpp, "LLAMA_FTYPE_MOSTLY_TQ2_0", INT2NUM(LLAMA_FTYPE_MOSTLY_TQ2_0));
   rb_define_const(rb_mLlamaCpp, "LLAMA_FTYPE_MOSTLY_MXFP4_MOE", INT2NUM(LLAMA_FTYPE_MOSTLY_MXFP4_MOE));
   rb_define_const(rb_mLlamaCpp, "LLAMA_FTYPE_MOSTLY_NVFP4", INT2NUM(LLAMA_FTYPE_MOSTLY_NVFP4));
+  rb_define_const(rb_mLlamaCpp, "LLAMA_FTYPE_MOSTLY_Q1_0", INT2NUM(LLAMA_FTYPE_MOSTLY_Q1_0));
   rb_define_const(rb_mLlamaCpp, "LLAMA_FTYPE_GUESSED", INT2NUM(LLAMA_FTYPE_GUESSED));
   /* llama_rope_scaling_type */
   /* Document-const: LlamaCpp::LLAMA_ROPE_SCALING_TYPE_UNSPECIFIED */
@@ -4324,6 +4309,7 @@ void Init_llama_cpp(void) {
   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));
   rb_define_const(rb_mLlamaCpp, "LLAMA_SPLIT_MODE_ROW", INT2NUM(LLAMA_SPLIT_MODE_ROW));
+  rb_define_const(rb_mLlamaCpp, "LLAMA_SPLIT_MODE_TENSOR", INT2NUM(LLAMA_SPLIT_MODE_TENSOR));
 
   rb_define_module_function(rb_mLlamaCpp, "llama_flash_attn_type_name", rb_llama_flash_attn_type_name, 1);
 
@@ -5157,12 +5143,6 @@ void Init_llama_cpp(void) {
   /* llama_free */
   rb_define_module_function(rb_mLlamaCpp, "llama_free", rb_llama_free, 1);
 
-  /* llama_params_fit_status */
-  /* Document-const: LlamaCpp::LLAMA_PARAMS_FIT_STATUS_SUCCESS */
-  rb_define_const(rb_mLlamaCpp, "LLAMA_PARAMS_FIT_STATUS_SUCCESS", INT2NUM(LLAMA_PARAMS_FIT_STATUS_SUCCESS));
-  rb_define_const(rb_mLlamaCpp, "LLAMA_PARAMS_FIT_STATUS_FAILURE", INT2NUM(LLAMA_PARAMS_FIT_STATUS_FAILURE));
-  rb_define_const(rb_mLlamaCpp, "LLAMA_PARAMS_FIT_STATUS_ERROR", INT2NUM(LLAMA_PARAMS_FIT_STATUS_ERROR));
-
   /* TODO: llama_params_fit */
 
   /* llama_time_us */
@@ -5699,9 +5679,6 @@ void Init_llama_cpp(void) {
   /* llama_perf_sampler_reset */
   rb_define_module_function(rb_mLlamaCpp, "llama_perf_sampler_reset", rb_llama_perf_sampler_reset, 1);
 
-  /* llama_memory_breakdown_print */
-  rb_define_module_function(rb_mLlamaCpp, "llama_memory_breakdown_print", rb_llama_memory_breakdown_print, 1);
-
   /* TODO: typedef bool (*llama_opt_param_filter) */
   /* TODO: bool llama_opt_param_filter_all */
   /* TODO: struct llama_opt_params */

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 = '0.24.3'
+  VERSION = '0.25.0'
 
   # The supported version of llama.cpp.
-  LLAMA_CPP_VERSION = 'b8640'
+  LLAMA_CPP_VERSION = 'b8920'
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: llama_cpp
 version: !ruby/object:Gem::Version
-  version: 0.24.3
+  version: 0.25.0
 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/0.24.3/
+  documentation_uri: https://gemdocs.org/gems/llama_cpp/0.25.0/
   rubygems_mfa_required: 'true'
 rdoc_options: []
 require_paths: