vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2026-05-19 08:00:25 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

split utils.cpp into 2 files to support sd.cpp

Browse source
This commit is contained in:
Concedo 2026-05-04 15:04:12 +08:00
parent 276c651a12
commit 950676fdb7
9 changed files with 231 additions and 225 deletions
Download patch file Download diff file Expand all files Collapse all files

57
otherarch/utils.h
View file

@ -8,16 +8,6 @@
#include <random>
#include <thread>
#include "ggml_v3.h"
#include "llama.h"
//
// CLI argument parsing
//
//
// Vocab utils
//
struct gpt_vocab {
using id = int32_t;
@ -73,6 +63,7 @@ std::vector<std::string> split_string(const std::string& input, const std::strin
bool kcpp_decode_audio_from_buf(const unsigned char * buf_in, size_t len, int target_sampler_rate, std::vector<float> & pcmf32_mono);
bool kcpp_decode_audio_to_f32_stereo_48k(const uint8_t * data, size_t data_size, std::vector<float> & pcm, int & T_audio);
typedef struct ggml_backend_device * ggml_backend_dev_t;
std::vector<ggml_backend_dev_t> kcpp_parse_device_list(const std::string & value);
bool kcpp_string_ends_with(const std::string& str, const std::string& suffix);
@ -81,52 +72,6 @@ int ComputeSharedPrefixLength(const std::vector<int> &tokens_a,const std::vector
float ComputePrefixMatchPercent(const std::vector<int> &tokens_a,const std::vector<int> &tokens_b);
bool FullyContainedPrefix(std::vector<int> &sequence1, std::vector<int> &sequence2);
//duplcated and modified from llava_embd_batch
struct kcpp_embd_batch {
std::vector<llama_pos> pos;
std::vector<llama_pos> pos_view;
std::vector<int32_t> n_seq_id;
std::vector<llama_seq_id> seq_id_0;
std::vector<llama_seq_id*> seq_ids;
std::vector<int8_t> logits;
llama_batch batch;
llama_batch get_view(int offset, int n_tokens, int n_embd_mmproj);
// Embedding constructor
kcpp_embd_batch(
float * embd,
int32_t n_tokens,
int32_t npast,
bool use_mrope,
bool mrope_is_image = false,
int img_nx = 0,
int img_ny = 0
);
// Token constructor
kcpp_embd_batch(
std::vector<llama_token> & tokens,
int32_t npast,
bool use_mrope,
bool return_all_logits,
bool mrope_is_image = false,
int img_nx = 0,
int img_ny = 0
);
private:
void init_kcpp_batch(
int32_t n_tokens,
int32_t npast,
bool use_mrope,
bool return_all_logits,
bool mrope_is_image,
int img_nx,
int img_ny
);
};
#pragma pack(push, 1)
struct wav16_header {
char riff[4] = {'R', 'I', 'F', 'F'};