mikhail/llama.cpp-mtp-turboquant
1
0
Fork 0
Code Issues Pull Requests Actions 28 Packages Projects Releases Wiki Activity

common : refactor common_sampler + grammar logic changes (#17937)

Browse Source 
* common : refactor common_sampler + grammar logic changes

* tests : increase max_tokens to get needed response

* batched : fix uninitialized samplers
This commit is contained in:
Georgi Gerganov
2025-12-14 10:11:13 +02:00
committed by GitHub
parent 3238b1400c
commit 254098a279
27 changed files with 372 additions and 293 deletions
Download Patch File Download Diff File Expand all files Collapse all files
common/common.h
+24 -7
View File
@@ -195,7 +195,6 @@ struct common_params_sampling {
std::vector<std::string> dry_sequence_breakers = {"\n", ":", "\"", "*"}; // default sequence breakers for DRY
std::vector<enum common_sampler_type> samplers = {
COMMON_SAMPLER_TYPE_PENALTIES,
COMMON_SAMPLER_TYPE_DRY,
@@ -216,6 +215,10 @@ struct common_params_sampling {
std::vector<llama_logit_bias> logit_bias; // logit biases to apply
std::vector<llama_logit_bias> logit_bias_eog; // pre-calculated logit biases for EOG tokens
bool has_logit_bias() const {
return !logit_bias.empty();
}
// print the parameters into a string
std::string print() const;
};
@@ -669,15 +672,29 @@ bool tty_can_use_colors();
// Model utils
//
// note: defines object's lifetime
struct common_init_result {
llama_model_ptr model;
llama_context_ptr context;
struct common_sampler;
std::vector<llama_adapter_lora_ptr> lora;
// note: defines the model, context, samplers, ets. lifetimes
struct common_init_result {
common_init_result(common_params & params);
~common_init_result();
llama_model * model();
llama_context * context();
common_sampler * sampler(llama_seq_id seq_id);
std::vector<llama_adapter_lora_ptr> & lora();
void free_context();
private:
struct impl;
std::unique_ptr<impl> pimpl;
};
struct common_init_result common_init_from_params(common_params & params);
using common_init_result_ptr = std::unique_ptr<common_init_result>;
common_init_result_ptr common_init_from_params(common_params & params);
struct llama_model_params common_model_params_to_llama ( common_params & params);
struct llama_context_params common_context_params_to_llama(const common_params & params);