echodict/llama.cpp

1#pragma once2 3#include "llama.h"4 5#include <cstdint>6 7#define LLAMA_MAX_SEQ 2568 9struct llama_cparams {10    uint32_t n_ctx;           // context size used during inference11    uint32_t n_ctx_seq;       // context for a single sequence12    uint32_t n_batch;13    uint32_t n_ubatch;14    uint32_t n_seq_max;15    int32_t  n_threads;       // number of threads to use for generation16    int32_t  n_threads_batch; // number of threads to use for batch processing17 18    float rope_freq_base;19    float rope_freq_scale;20 21    uint32_t n_ctx_orig_yarn;22    // These hyperparameters are not exposed in GGUF, because all23    // existing YaRN models use the same values for them.24    float yarn_ext_factor;25    float yarn_attn_factor;26    float yarn_beta_fast;27    float yarn_beta_slow;28 29    bool embeddings;30    bool causal_attn;31    bool offload_kqv;32    bool flash_attn;33    bool auto_fa;34    bool fused_gdn_ar;       // use fused gated delta net (autoregressive)35    bool fused_gdn_ch;       // use fused gated delta net (chunked)36    bool auto_fgdn;37    bool no_perf;38    bool warmup;39    bool op_offload;40    bool kv_unified;41    bool pipeline_parallel;42 43    enum llama_pooling_type pooling_type;44 45    ggml_backend_sched_eval_callback cb_eval;46    void * cb_eval_user_data;47};48