llama_cpp 0.17.10 → 0.18.1

data/sig/llama_cpp.rbs

@@ -1,425 +0,0 @@
-module LLaMACpp
-  VERSION: String
-  LLAMA_CPP_VERSION: String
-  LLAMA_DEFALUT_SEED: String
-
-  LLAMA_FILE_MAGIC_GGLA: String
-  LLAMA_FILE_MAGIC_GGSN: String
-  LLAMA_FILE_MAGIC_GGSQ: String
-  LLAMA_SESSION_MAGIC: String
-  LLAMA_SESSION_VERSION: String
-  LLAMA_STATE_SEQ_MAGIC: String
-  LLAMA_STATE_SEQ_VERSION: String
-
-  LLAMA_VOCAB_TYPE_NONE: Integer
-  LLAMA_VOCAB_TYPE_SPM: Integer
-  LLAMA_VOCAB_TYPE_BPE: Integer
-  LLAMA_VOCAB_TYPE_WPM: Integer
-  LLAMA_VOCAB_TYPE_UGM: Integer
-  LLAMA_VOCAB_TYPE_RWKV: Integer
-
-  LLAMA_VOCAB_PRE_TYPE_DEFAULT: Integer
-  LLAMA_VOCAB_PRE_TYPE_LLAMA3: Integer
-  LLAMA_VOCAB_PRE_TYPE_DEEPSEEK_LLM: Integer
-  LLAMA_VOCAB_PRE_TYPE_DEEPSEEK_CODER: Integer
-  LLAMA_VOCAB_PRE_TYPE_FALCON: Integer
-  LLAMA_VOCAB_PRE_TYPE_MPT: Integer
-  LLAMA_VOCAB_PRE_TYPE_STARCODER: Integer
-  LLAMA_VOCAB_PRE_TYPE_GPT2: Integer
-  LLAMA_VOCAB_PRE_TYPE_REFACT: Integer
-  LLAMA_VOCAB_PRE_TYPE_COMMAND_R: Integer
-  LLAMA_VOCAB_PRE_TYPE_STABLELM2: Integer
-  LLAMA_VOCAB_PRE_TYPE_QWEN2: Integer
-  LLAMA_VOCAB_PRE_TYPE_OLMO: Integer
-  LLAMA_VOCAB_PRE_TYPE_DBRX: Integer
-  LLAMA_VOCAB_PRE_TYPE_SMAUG: Integer
-  LLAMA_VOCAB_PRE_TYPE_PORO: Integer
-  LLAMA_VOCAB_PRE_TYPE_CHATGLM3: Integer
-  LLAMA_VOCAB_PRE_TYPE_CHATGLM4: Integer
-  LLAMA_VOCAB_PRE_TYPE_VIKING: Integer
-  LLAMA_VOCAB_PRE_TYPE_JAIS: Integer
-  LLAMA_VOCAB_PRE_TYPE_TEKKEN: Integer
-  LLAMA_VOCAB_PRE_TYPE_SMOLLM: Integer
-  LLAMA_VOCAB_PRE_TYPE_CODESHELL: Integer
-  LLAMA_VOCAB_PRE_TYPE_BLOOM: Integer
-  LLAMA_VOCAB_PRE_TYPE_GPT3_FINNISH: Integer
-  LLAMA_VOCAB_PRE_TYPE_EXAONE: Integer
-
-  LLAMA_TOKEN_ATTR_UNDEFINED: Integer
-  LLAMA_TOKEN_ATTR_UNKNOWN: Integer
-  LLAMA_TOKEN_ATTR_UNUSED: Integer
-  LLAMA_TOKEN_ATTR_NORMAL: Integer
-  LLAMA_TOKEN_ATTR_CONTROL: Integer
-  LLAMA_TOKEN_ATTR_USER_DEFINED: Integer
-  LLAMA_TOKEN_ATTR_BYTE: Integer
-  LLAMA_TOKEN_ATTR_NORMALIZED: Integer
-  LLAMA_TOKEN_ATTR_LSTRIP: Integer
-  LLAMA_TOKEN_ATTR_RSTRIP: Integer
-  LLAMA_TOKEN_ATTR_SINGLE_WORD: Integer
-
-  LLAMA_FTYPE_ALL_F32: Integer
-  LLAMA_FTYPE_MOSTLY_F16: Integer
-  LLAMA_FTYPE_MOSTLY_Q4_0: Integer
-  LLAMA_FTYPE_MOSTLY_Q4_1: Integer
-  LLAMA_FTYPE_MOSTLY_Q8_0: Integer
-  LLAMA_FTYPE_MOSTLY_Q5_0: Integer
-  LLAMA_FTYPE_MOSTLY_Q5_1: Integer
-  LLAMA_FTYPE_MOSTLY_Q2_K: Integer
-  LLAMA_FTYPE_MOSTLY_Q3_K_S: Integer
-  LLAMA_FTYPE_MOSTLY_Q3_K_M: Integer
-  LLAMA_FTYPE_MOSTLY_Q3_K_L: Integer
-  LLAMA_FTYPE_MOSTLY_Q4_K_S: Integer
-  LLAMA_FTYPE_MOSTLY_Q4_K_M: Integer
-  LLAMA_FTYPE_MOSTLY_Q5_K_S: Integer
-  LLAMA_FTYPE_MOSTLY_Q5_K_M: Integer
-  LLAMA_FTYPE_MOSTLY_Q6_K: Integer
-  LLAMA_FTYPE_MOSTLY_IQ2_XXS: Integer
-  LLAMA_FTYPE_MOSTLY_IQ2_XS: Integer
-  LLAMA_FTYPE_MOSTLY_Q2_K_S: Integer
-  LLAMA_FTYPE_MOSTLY_IQ3_XS: Integer
-  LLAMA_FTYPE_MOSTLY_IQ3_XXS: Integer
-  LLAMA_FTYPE_MOSTLY_IQ1_S: Integer
-  LLAMA_FTYPE_MOSTLY_IQ4_NL: Integer
-  LLAMA_FTYPE_MOSTLY_IQ3_S: Integer
-  LLAMA_FTYPE_MOSTLY_IQ3_M: Integer
-  LLAMA_FTYPE_MOSTLY_IQ4_XS: Integer
-  LLAMA_FTYPE_MOSTLY_IQ1_M: Integer
-  LLAMA_FTYPE_MOSTLY_BF16: Integer
-  LLAMA_FTYPE_MOSTLY_Q4_0_4_4: Integer
-  LLAMA_FTYPE_MOSTLY_Q4_0_4_8: Integer
-  LLAMA_FTYPE_MOSTLY_Q4_0_8_8: Integer
-  LLAMA_FTYPE_MOSTLY_TQ1_0: Integer
-  LLAMA_FTYPE_MOSTLY_TQ2_0: Integer
-  LLAMA_FTYPE_GUESSED: Integer
-
-  LLAMA_KV_OVERRIDE_TYPE_INT: Integer
-  LLAMA_KV_OVERRIDE_TYPE_FLOAT: Integer
-  LLAMA_KV_OVERRIDE_TYPE_BOOL: Integer
-  LLAMA_KV_OVERRIDE_TYPE_STR: Integer
-
-  LLAMA_GRETYPE_END: Integer
-  LLAMA_GRETYPE_ALT: Integer
-  LLAMA_GRETYPE_RULE_REF: Integer
-  LLAMA_GRETYPE_CHAR: Integer
-  LLAMA_GRETYPE_CHAR_NOT: Integer
-  LLAMA_GRETYPE_CHAR_RNG_UPPER: Integer
-  LLAMA_GRETYPE_CHAR_ALT: Integer
-  LLAMA_GRETYPE_CHAR_ANY: Integer
-
-  LLAMA_ROPE_SCALING_TYPE_UNSPECIFIED: Integer
-  LLAMA_ROPE_SCALING_TYPE_NONE: Integer
-  LLAMA_ROPE_SCALING_TYPE_LINEAR: Integer
-  LLAMA_ROPE_SCALING_TYPE_YARN: Integer
-  LLAMA_ROPE_SCALING_TYPE_MAX_VALUE: Integer
-
-  LLAMA_POOLING_TYPE_UNSPECIFIED: Integer
-  LLAMA_POOLING_TYPE_NONE: Integer
-  LLAMA_POOLING_TYPE_MEAN: Integer
-  LLAMA_POOLING_TYPE_CLS: Integer
-  LLAMA_POOLING_TYPE_LAST: Integer
-
-  LLAMA_ATTENTION_TYPE_UNSPECIFIED: Integer
-  LLAMA_ATTENTION_TYPE_CAUSAL: Integer
-  LLAMA_ATTENTION_TYPE_NON_CAUSAL: Integer
-
-  LLAMA_SPLIT_MODE_NONE: Integer
-  LLAMA_SPLIT_MODE_LAYER: Integer
-  LLAMA_SPLIT_MODE_ROW: Integer
-
-  def self?.backend_init: () -> void
-  def self?.backend_free: () -> void
-  def self?.numa_init: (Integer) -> void
-  def self?.model_quantize: (input_path: String, output_path: String, params: ModelQuantizeParams) -> void
-  def self?.generate: (::LLaMACpp::Context, String,
-    ?n_predict: Integer, ?n_threads: Integer, ?n_keep: Integer, ?n_batch: Integer,
-    ?repeat_last_n: Integer, ?repeat_penalty: Float, ?frequency: Float, ?presence: Float,
-    ?top_k: Integer, ?top_p: Float, ?tfs_z: Float, ?typical_p: Float, ?temperature: Float) -> String
-  def self?.print_system_info: () -> void
-  def self?.time_us: () -> Integer
-  def self?.max_devices: () -> Integer
-  def self?.supports_mmap?: () -> bool
-  def self?.supports_mlock?: () -> bool
-  def self?.supports_gpu_offload?: () -> bool
-
-  class TokenData
-    public
-
-    def initialize: (id: Integer, logit: Float, p: Float) -> void
-    def id: () -> Integer
-    def id=: (Integer) -> Integer
-    def logit: () -> Float
-    def logit=: (Float) -> Float
-    def p: () -> Float
-    def p=: (Float) -> Float
-  end
-
-  class TokenDataArray
-    public
-
-    def initialize: (Array[::LLaMACpp::TokenData], ?sorted: bool) -> void
-    def size: () -> Integer
-    def sorted: () -> bool
-  end
-
-  class Model
-    public
-
-    def initialize: (model_path: String, params: ::LLaMACpp::ModelParams) -> void
-                  | () -> void
-    def empty?: () -> bool
-    def free: () -> void
-    def load: (model_path: String, params: ::LLaMACpp::ModelParams) -> void
-    def vocab_type: () -> Integer
-    def rope_type: () -> Integer
-    def n_vocab: () -> Integer
-    def n_ctx_train: () -> Integer
-    def n_embd: () -> Integer
-    def n_layer: () -> Integer
-    def rope_freq_scale_train: () -> Float
-    def token_to_piece: (Integer, ?lstrip: Integer, ?special: bool) -> String
-    def tokenize: (text: String, ?n_max_tokens: Integer, ?add_bos: bool, ?special: bool) -> Array[Integer]
-    def desc: () -> String
-    def size: () -> Integer
-    def n_params: () -> Integer
-    def text: (Integer) -> String
-    def score: (Integer) -> Float
-    def token_attr: (Integer) -> Integer
-    def token_bos: () -> Integer
-    def token_eos: () -> Integer
-    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
-    def token_middle: () -> Integer
-    def token_suffix: () -> Integer
-    def token_eot: () -> Integer
-    def token_is_eog?: (Integer) -> bool
-    def token_is_control?: (Integer) -> bool
-    def has_encoder?: () -> bool
-    def has_decoder?: () -> bool
-    def decoder_start_token: () -> Integer
-    def is_recurrent?: () -> bool
-    def detokenize: (Array[Integer], ?remove_special: bool, ?unparse_special: bool) -> String
-  end
-
-  class Timings
-    public
-
-    def t_start_ms: () -> Float
-    def t_end_ms: () -> Float
-    def t_load_ms: () -> Float
-    def t_sample_ms: () -> Float
-    def t_p_eval_ms: () -> Float
-    def t_eval_ms: () -> Float
-    def n_sample: () -> Integer
-    def n_p_eval: () -> Integer
-    def n_eval: () -> Integer
-  end
-
-  class ModelKVOverride
-    public
-
-    def key: () -> String
-    def tag: () -> Integer
-    def val_i64: () -> Integer
-    def val_f64: () -> Float
-    def val_bool: () -> bool
-    def val_str: () -> String
-  end
-
-  class ModelParams
-    public
-
-    def n_gpu_layers: () -> Integer
-    def n_gpu_layers=: (Integer) -> Integer
-    def split_mode: () -> Integer
-    def split_mode=: (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
-    def check_tensors: () -> bool
-    def check_tensors=: (bool) -> bool
-  end
-
-  class Batch
-    public
-
-    def self.get_one: (tokens: Array[Integer], n_tokens: Integer, pos_zero: Integer, seq_id: Integer) -> ::LLaMACpp::Batch
-    def initialize: (max_n_token: Integer, n_embd: Integer, max_n_seq: 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_n_seq_id: (Integer, Integer) -> Integer
-    def get_n_seq_id: (Integer) -> Integer
-    def set_seq_id: (Integer, Integer, Integer) -> Integer
-    def get_seq_id: (Integer, Integer) -> Integer
-    def set_logit: (Integer, bool) -> bool
-    def get_logit: (Integer) -> bool
-  end
-
-  class Context
-    public
-
-    attr_reader model: ::LLaMACpp::Model
-
-    def initialize: (model: ::LLaMACpp::Model, params: ::LLaMACpp::ContextParams) -> void
-    def embeddings: () -> Array[Float]
-    def embeddings_ith: (Integer) -> Array[Float]
-    def embeddings_seq: (Integer) -> Array[Float]
-    def encode: (::LLaMACpp::Batch) -> void
-    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
-    def n_ubatch: () -> Integer
-    def n_seq_max: () -> Integer
-    def n_threads: () -> Integer
-    def n_threads_batch: () -> Integer
-    def timings: () -> ::LLaMACpp::Timings
-    def print_timings: () -> void
-    def reset_timings: () -> void
-    def kv_cache_token_count: () -> Integer
-    def kv_cache_clear: () -> 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_add: (Integer, Integer, Integer, Integer) -> void
-    def kv_cache_seq_div: (Integer, Integer, Integer, Integer) -> void
-    def kv_cache_seq_pos_max: (Integer) -> Integer
-    def kv_cache_defrag: () -> void
-    def kv_cache_update: () -> void
-    def set_rng_seed: (Integer) -> void
-    def set_causal_attn: (bool) -> void
-    def synchronize: () -> void
-    def load_session_file: (session_path: String) -> void
-    def save_session_file: (session_path: String, session_tokens: Array[Integer]) -> void
-    def sample_repetition_penalties: (::LLaMACpp::TokenDataArray, Array[Integer], penalty_repeat: Float, penalty_freq: Float, penalty_present: Float) -> void
-    def sample_apply_guidance: (logits: Array[Float], logits_guidance: Array[Float], scale: Float) -> void
-    def sample_softmax: (::LLaMACpp::TokenDataArray) -> void
-    def sample_top_k: (::LLaMACpp::TokenDataArray, k: Integer, ?min_keep: Integer) -> void
-    def sample_top_p: (::LLaMACpp::TokenDataArray, prob: Float, ?min_keep: Integer) -> void
-    def sample_min_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_entropy: (::LLaMACpp::TokenDataArray, min_temp: Float, max_temp: Float, exponent_val: 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]
-    def sample_token_greedy: (::LLaMACpp::TokenDataArray) -> Integer
-    def sample_token: (::LLaMACpp::TokenDataArray) -> Integer
-    def sample_grammar: (::LLaMACpp::TokenDataArray, grammar: ::LLaMACpp::Grammar) -> void
-    def grammar_accept_token: (grammar: ::LLaMACpp::Grammar, token: Integer) -> void
-    def apply_control_vector: (data: Array[Float], n_embd: Integer, il_start: Integer, il_end: Integer) -> void
-    def pooling_type: () -> Integer
-  end
-
-  class ContextParams
-    public
-
-    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_ubatch: () -> Integer
-    def n_ubatch=: (Integer) -> Integer
-    def n_seq_max: () -> Integer
-    def n_seq_max=: (Integer) -> Integer
-    def n_threads: () -> Integer
-    def n_threads=: (Integer) -> Integer
-    def n_threads_batch: () -> Integer
-    def n_threads_batch=: (Integer) -> Integer
-    def rope_scaling_type=: (Integer) -> Integer
-    def rope_scaling_type: () -> Integer
-    def pooling_type=: (Integer) -> Integer
-    def pooling_type: () -> Integer
-    def attention_type=: (Integer) -> Integer
-    def attention_type: () -> Integer
-    def rope_freq_base=: (Float) -> Float
-    def rope_freq_base: () -> Float
-    def rope_freq_scale=: (Float) -> Float
-    def rope_freq_scale: () -> Float
-    def yarn_ext_factor=: (Float) -> Float
-    def yarn_ext_factor: () -> Float
-    def yarn_attn_factor=: (Float) -> Float
-    def yarn_attn_factor: () -> Float
-    def yarn_beta_fast=: (Float) -> Float
-    def yarn_beta_fast: () -> Float
-    def yarn_beta_slow=: (Float) -> Float
-    def yarn_beta_slow: () -> Float
-    def yarn_orig_ctx=: (Integer) -> Integer
-    def yarn_orig_ctx: () -> Integer
-    def defrag_thold=: (Float) -> Float
-    def defrag_thold: () -> Float
-    def type_k=: (Integer) -> Integer
-    def type_k: () -> Integer
-    def type_v=: (Integer) -> Integer
-    def type_v: () -> Integer
-    def logits_all: () -> bool
-    def logits_all=: (bool) -> bool
-    def embeddings: () -> bool
-    def embeddings=: (bool) -> bool
-    def offload_kqv: () -> bool
-    def offload_kqv=: (bool) -> bool
-    def flash_attn: () -> bool
-    def flash_attn=: (bool) -> bool
-  end
-
-  class ModelQuantizeParams
-    public
-
-    attr_reader params: ::LLaMACpp::ModelParams
-
-    def n_thread: () -> Integer
-    def n_thread=: (Integer) -> Integer
-    def ftype: () -> Integer
-    def ftype=: (Integer) -> Integer
-    def allow_quantization: () -> bool
-    def allow_quantization=: (bool) -> bool
-    def quantize_output_tensor: () -> bool
-    def quantize_output_tensor=: (bool) -> bool
-    def only_copy: () -> bool
-    def only_copy=: (bool) -> bool
-    def pure: () -> bool
-    def pure=: (bool) -> bool
-    def keep_split: () -> bool
-    def keep_split=: (bool) -> bool
-  end
-
-  class Params = ContextParams
-
-  class GrammarElement
-    public
-
-    def initialize: (?type: Integer, ?value: Integer) -> void
-    def type: () -> Integer
-    def type=: (Integer) -> Integer
-    def value: () -> Integer
-    def value=: (Integer) -> Integer
-  end
-
-  class Grammar
-    def initialize: (rules: Array[Array[LLaMACpp::GrammarElement]], start_rule_index: Integer) -> void
-  end
-end