llama.cpp/common/speculative.h

34 lines
920 B
C
Raw Normal View History

#pragma once
#include "llama.h"
#include "common.h"
struct common_speculative;
struct common_speculative_params {
int n_draft = 16;
int n_min = 5; // do not add drafts smaller than this, TODO: leave this to user?
int n_reuse = 256;
float p_min = 0.9f;
};
2024-11-22 09:05:49 +00:00
struct common_speculative * common_speculative_init(
struct common_speculative_params params,
struct llama_context * ctx_dft);
void common_speculative_free(struct common_speculative * spec);
2024-11-22 09:05:49 +00:00
bool common_speculative_are_compatible(
const struct llama_context * ctx_tgt,
const struct llama_context * ctx_dft);
// sample up to n_draft tokens and add them to the batch using the draft model
//
void common_speculative_add_draft(
struct common_speculative * spec,
struct llama_batch & batch_tgt,
const llama_tokens & prompt,
llama_token id_last,
llama_token n_past_tgt);