llama_cpp 0.5.3 → 0.7.0

data/sig/llama_cpp.rbs

@@ -67,16 +67,16 @@ module LLaMACpp
   class Model
     public
 
-    def initialize: (model_path: String, params: ::LLaMACpp::ContextParams) -> void
+    def initialize: (model_path: String, params: ::LLaMACpp::ModelParams) -> void
                   | () -> void
     def empty?: () -> bool
     def free: () -> void
-    def load: (model_path: String, params: ::LLaMACpp::ContextParams) -> void
-    def apply_lora_from_file: (lora_path: String, ?base_model_path: String, ?n_threads: Integer) -> void
+    def load: (model_path: String, params: ::LLaMACpp::ModelParams) -> void
+    def apply_lora_from_file: (lora_path: String, ?scale: Float, ?base_model_path: String, ?n_threads: Integer) -> void
     def n_vocab: () -> Integer
-    def n_ctx: () -> Integer
     def n_ctx_train: () -> Integer
     def n_embd: () -> Integer
+    def rope_freq_scale_train: () -> Float
     def token_to_piece: (Integer) -> String
     def tokenize: (text: String, ?n_max_tokens: Integer, ?add_bos: bool) -> Array[Integer]
     def desc: () -> String
@@ -98,10 +98,50 @@ module LLaMACpp
     def n_eval: () -> Integer
   end
 
+  class ModelParams
+    public
+
+    def n_gpu_layers: () -> Integer
+    def n_gpu_layers=: (Integer) -> Integer
+    def main_gpu: () -> Integer
+    def main_gpu=: (Integer) -> Integer
+    def tensor_split: () -> Array[Float]
+    def vocab_only: () -> bool
+    def vocab_only=: (bool) -> bool
+    def use_mmap: () -> bool
+    def use_mmap=: (bool) -> bool
+    def use_mlock: () -> bool
+    def use_mlock=: (bool) -> bool
+  end
+
+  class Batch
+    public
+
+    def initialize: (n_tokens: Integer, embd: Integer) -> void
+    def n_tokens=: (Integer) -> Integer
+    def n_tokens: () -> Integer
+    def all_pos_zero=: (Integer) -> Integer
+    def all_pos_zero: () -> Integer
+    def all_pos_one=: (Integer) -> Integer
+    def all_pos_one: () -> Integer
+    def all_seq_id=: (Integer) -> Integer
+    def all_seq_id: () -> Integer
+    def set_token: (Integer, Integer) -> Integer
+    def get_token: (Integer) -> Integer
+    def set_pos: (Integer, Integer) -> Integer
+    def get_pos: (Integer) -> Integer
+    def set_seq_id: (Integer, Integer) -> Integer
+    def get_seq_id: (Integer) -> Integer
+    def set_logit: (Integer, bool) -> bool
+    def get_logit: (Integer) -> bool
+  end
+
   class Context
     public
 
-    def initialize: (model: ::LLaMACpp::Model) -> void
+    attr_reader model: ::LLaMACpp::Model
+
+    def initialize: (model: ::LLaMACpp::Model, params: ::LLaMACpp::ContextParams) -> void
     def embeddings: () -> Array[Float]
     def text: (Integer) -> String
     def score: (Integer) -> Float
@@ -109,20 +149,24 @@ module LLaMACpp
     def token_bos: () -> Integer
     def token_eos: () -> Integer
     def token_nl: () -> Integer
-    def eval: (tokens: Array[Integer], n_past: Integer, ?n_tokens: Integer, ?n_threads: Integer) -> void
-    def eval_embd: (tokens: Array[Float], n_past: Integer, ?n_tokens: Integer, ?n_threads: Integer) -> void
-    def eval_export: (String) -> bool
+    def token_prefix: () -> Integer
+    def token_middle: () -> Integer
+    def token_suffix: () -> Integer
+    def token_eot: () -> Integer
+    def eval: (tokens: Array[Integer], n_past: Integer, ?n_tokens: Integer) -> void
+    def eval_embd: (tokens: Array[Float], n_past: Integer, ?n_tokens: Integer) -> void
+    def decode: (::LLaMACpp::Batch) -> void
     def logits: () -> Array[Float]
     def n_ctx: () -> Integer
-    def n_ctx_train: () -> Integer
-    def n_embd: () -> Integer
-    def n_vocab: () -> Integer
     def timings: () -> ::LLaMACpp::Timings
     def print_timings: () -> void
     def reset_timings: () -> void
-    def token_to_piece: (Integer) -> String
-    def tokenize: (text: String, ?n_max_tokens: Integer, ?add_bos: bool) -> Array[Integer]
     def kv_cache_token_count: () -> Integer
+    def kv_cache_tokens_rm: (Integer, Integer) -> void
+    def kv_cache_seq_rm: (Integer, Integer,Integer) -> void
+    def kv_cache_seq_cp: (Integer, Integer,Integer, Integer) -> void
+    def kv_cache_seq_keep: (Integer) -> void
+    def kv_cache_seq_shift: (Integer, Integer, Ingteger, Integer) -> void
     def set_rng_seed: (Integer) -> void
     def load_session_file: (session_path: String) -> void
     def save_session_file: (session_path: String, session_tokens: Array[Integer]) -> void
@@ -134,6 +178,7 @@ module LLaMACpp
     def sample_top_p: (::LLaMACpp::TokenDataArray, prob: Float, ?min_keep: Integer) -> void
     def sample_tail_free: (::LLaMACpp::TokenDataArray, z: Float, ?min_keep: Integer) -> void
     def sample_typical: (::LLaMACpp::TokenDataArray, prob: Float, ?min_keep: Integer) -> void
+    def sample_temp: (::LLaMACpp::TokenDataArray, temp: Float) -> void
     def sample_temperature: (::LLaMACpp::TokenDataArray, temperature: Float) -> void
     def sample_token_mirostat: (::LLaMACpp::TokenDataArray, tau: Float, eta: Float, m: Integer, mu: Float) -> [Integer, Float]
     def sample_token_mirostat_v2: (::LLaMACpp::TokenDataArray, tau: Float, eta: Float, mu: Float) -> [Integer, Float]
@@ -146,37 +191,28 @@ module LLaMACpp
   class ContextParams
     public
 
-    def embedding: () -> bool
-    def embedding=: (bool) -> bool
-    def f16_kv: () -> bool
-    def f16_kv=: (bool) -> bool
-    def logits_all: () -> bool
-    def logits_all=: (bool) -> bool
+    def seed: () -> Integer
+    def seed=: (Integer) -> Integer
     def n_ctx: () -> Integer
     def n_ctx=: (Integer) -> Integer
     def n_batch: () -> Integer
     def n_batch=: (Integer) -> Integer
-    def n_gpu_layers: () -> Integer
-    def n_gpu_layers=: (Integer) -> Integer
-    def main_gpu: () -> Integer
-    def main_gpu=: (Integer) -> Integer
-    def tensor_split: () -> Array[Float]
+    def n_threads: () -> Integer
+    def n_threads=: (Integer) -> Integer
+    def n_threads_batch: () -> Integer
+    def n_threads_batch=: (Integer) -> Integer
     def rope_freq_base=: (Float) -> Float
     def rope_freq_base: () -> Float
     def rope_freq_scale=: (Float) -> Float
     def rope_freq_scale: () -> Float
-    def low_vram: () -> bool
-    def low_vram=: (bool) -> bool
     def mul_mat_q: () -> bool
     def mul_mat_q=: (bool) -> bool
-    def seed: () -> Integer
-    def seed=: (Integer) -> Integer
-    def use_mlock: () -> bool
-    def use_mlock=: (bool) -> bool
-    def use_mmap: () -> bool
-    def use_mmap=: (bool) -> bool
-    def vocab_only: () -> bool
-    def vocab_only=: (bool) -> bool
+    def f16_kv: () -> bool
+    def f16_kv=: (bool) -> bool
+    def logits_all: () -> bool
+    def logits_all=: (bool) -> bool
+    def embedding: () -> bool
+    def embedding=: (bool) -> bool
   end
 
   class ModelQuantizeParams

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: llama_cpp
 version: !ruby/object:Gem::Version
-  version: 0.5.3
+  version: 0.7.0
 platform: ruby
 authors:
 - yoshoku
 autorequire:
 bindir: exe
 cert_chain: []
-date: 2023-09-22 00:00:00.000000000 Z
+date: 2023-10-07 00:00:00.000000000 Z
 dependencies: []
 description: llama_cpp.rb provides Ruby bindings for the llama.cpp.
 email:
@@ -48,6 +48,7 @@ files:
 - ext/llama_cpp/src/llama-util.h
 - ext/llama_cpp/src/llama.cpp
 - ext/llama_cpp/src/llama.h
+- ext/llama_cpp/src/unicode.h
 - lib/llama_cpp.rb
 - lib/llama_cpp/version.rb
 - sig/llama_cpp.rbs
@@ -75,7 +76,7 @@ required_rubygems_version: !ruby/object:Gem::Requirement
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 3.3.26
+rubygems_version: 3.4.19
 signing_key:
 specification_version: 4
 summary: Ruby bindings for the llama.cpp.