llama_cpp 0.24.3 → 0.25.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: e4ec06dde3c07e65bccca17e900a37042082fc706c8cddd34e447affe1198ba5
-  data.tar.gz: a4ed3057e3da8fc8be59ab578faf7fdd89773c17a6ee99d03cb25bb5a9af282d
+  metadata.gz: 5ed500b166eb331a19679b390da8d9647902fcc1d528d6f3933e4bfbcc6fdf0c
+  data.tar.gz: '02973a0de0e2ab05f2c27516aab9b21e22aef42a257c82858ec3e621e8ded374'
 SHA512:
-  metadata.gz: 71803b7ed68c4c9911bf1a91a63e5a5e8d588e297f8ba8d9a28665e06cc367c2eecf6b644033b57ab50358aacc4cb3c6c69f44c54a121dc521fd843e2d4ac727
-  data.tar.gz: f35964446f5d954531da4f62862007b97123e158211b01817ffba01409569a8f949b30cf2dc8b8915eebffcf90793978fa5d9d25abf3fbacee991ccfbef48ea8
+  metadata.gz: 255742254176964515b90f4b6109ea9a9488c73f2c323af4fdeced1b4f5bbaa5111df0e699f2b069dd88605616dec4a5f385c8099f8347cffcb7a1c1020883c1
+  data.tar.gz: 8eba6b3bbe403ba591570e10f2290c18223ffa48801906e2f1f292ca4925cbb38ef92b253d30826154d18a129c4d4d1ca072a6aa05f399ba87e4d36351c9176f

data/CHANGELOG.md

@@ -1,3 +1,18 @@
+## [[0.25.1](https://github.com/yoshoku/llama_cpp.rb/compare/v0.25.0...v0.25.1)] - 2026-05-09
+
+- Change supported llama.cpp version to b9070.
+  - ADD `LLAMA_STATE_SEQ_FLAGS_ON_DEVICE` constant value.
+
+## [[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
@@ -4200,6 +4184,7 @@ void Init_llama_cpp(void) {
   rb_define_const(rb_mLlamaCpp, "LLAMA_TOKEN_NULL", INT2NUM(LLAMA_TOKEN_NULL));
   rb_define_const(rb_mLlamaCpp, "LLAMA_STATE_SEQ_FLAGS_SWA_ONLY", INT2NUM(LLAMA_STATE_SEQ_FLAGS_SWA_ONLY));
   rb_define_const(rb_mLlamaCpp, "LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY", INT2NUM(LLAMA_STATE_SEQ_FLAGS_PARTIAL_ONLY));
+  rb_define_const(rb_mLlamaCpp, "LLAMA_STATE_SEQ_FLAGS_ON_DEVICE", INT2NUM(LLAMA_STATE_SEQ_FLAGS_ON_DEVICE));
 
   sprintf(tmp, "0x%x", LLAMA_FILE_MAGIC_GGLA);
   rb_define_const(rb_mLlamaCpp, "LLAMA_FILE_MAGIC_GGLA", rb_str_new2(tmp));
@@ -4292,6 +4277,7 @@ 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 +4310,7 @@ 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 +5144,6 @@ 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 +5680,6 @@ 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 = '0.25.1'
 
   # The supported version of llama.cpp.
-  LLAMA_CPP_VERSION = 'b8640'
+  LLAMA_CPP_VERSION = 'b9070'
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: llama_cpp
 version: !ruby/object:Gem::Version
-  version: 0.24.3
+  version: 0.25.1
 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/0.24.3/
+  documentation_uri: https://gemdocs.org/gems/llama_cpp/0.25.1/
   rubygems_mfa_required: 'true'
 rdoc_options: []
 require_paths: