2024-07-23 12:10:17 +02:00
|
|
|
#pragma once
|
|
|
|
|
2024-09-07 14:16:19 +02:00
|
|
|
// TODO: rename llama-sampling.h/.cpp to llama-sampler.h/.cpp ?
|
2024-07-23 12:10:17 +02:00
|
|
|
|
2024-09-07 14:16:19 +02:00
|
|
|
#include "llama-grammar.h"
|
2024-07-23 12:10:17 +02:00
|
|
|
|
2024-09-07 14:16:19 +02:00
|
|
|
#include <unordered_map>
|
2024-07-23 12:10:17 +02:00
|
|
|
|
2024-09-07 14:16:19 +02:00
|
|
|
struct llama_vocab;
|
|
|
|
struct llama_grammar;
|
2024-07-23 12:10:17 +02:00
|
|
|
|
2024-09-07 14:16:19 +02:00
|
|
|
// sampler chain
|
2024-07-23 12:10:17 +02:00
|
|
|
|
2024-09-07 14:16:19 +02:00
|
|
|
struct llama_sampler_chain {
|
|
|
|
llama_sampler_chain_params params;
|
|
|
|
|
|
|
|
std::vector<struct llama_sampler *> samplers;
|
|
|
|
|
|
|
|
// timing
|
|
|
|
|
|
|
|
mutable int64_t t_sample_us;
|
|
|
|
|
|
|
|
mutable int32_t n_sample;
|
2024-07-23 12:10:17 +02:00
|
|
|
};
|
|
|
|
|
2024-09-07 14:16:19 +02:00
|
|
|
using llama_token_cnt = std::unordered_map<llama_token, int>;
|
|
|
|
|
|
|
|
// TODO: tmp exposed until test-sampling is fixed
|
|
|
|
void llama_sampler_penalties_impl(
|
|
|
|
llama_token_data_array * cur_p,
|
|
|
|
const llama_token_cnt & token_count,
|
2024-07-23 12:10:17 +02:00
|
|
|
float penalty_repeat,
|
|
|
|
float penalty_freq,
|
|
|
|
float penalty_present);
|
|
|
|
|
2024-09-07 14:16:19 +02:00
|
|
|
struct llama_sampler * llama_sampler_init_grammar_impl(
|
|
|
|
const struct llama_vocab & vocab,
|
|
|
|
const char * grammar_str,
|
|
|
|
const char * grammar_root);
|