vrr/koboldcpp
2
0
Fork 0
mirror of https://github.com/LostRuins/koboldcpp.git synced 2025-09-10 17:14:36 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

llama : add llama_model_load_from_splits (#11255)

Browse source 
* llama : add `llama_model_load_from_splits`

* update
This commit is contained in:
Xuan Son Nguyen 2025-01-16 13:54:08 +01:00 • committed by GitHub
parent c67cc9837d
commit 681149ced2
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
5 changed files with 116 additions and 24 deletions
Download patch file Download diff file Expand all files Collapse all files

46
src/llama.cpp
View file

@ -31,7 +31,7 @@
#endif
// Returns 0 on success, -1 on error, and -2 on cancellation via llama_progress_callback
static int llama_model_load(const std::string & fname, llama_model & model, llama_model_params & params) {
static int llama_model_load(const std::string & fname, std::vector<std::string> & splits, llama_model & model, llama_model_params & params) {
// loading time will be recalculated after the first eval, so
// we take page faults deferred by mmap() into consideration
model.t_load_us = 0;
@ -40,7 +40,7 @@ static int llama_model_load(const std::string & fname, llama_model & model, llam
model.t_start_us = tm.t_start_us;
try {
llama_model_loader ml(fname, params.use_mmap, params.check_tensors, params.kv_overrides);
llama_model_loader ml(fname, splits, params.use_mmap, params.check_tensors, params.kv_overrides);
ml.print_info();
@ -9374,14 +9374,9 @@ int64_t llama_time_us(void) {
return ggml_time_us();
}
struct llama_model * llama_load_model_from_file(
const char * path_model,
struct llama_model_params params) {
return llama_model_load_from_file(path_model, params);
}
struct llama_model * llama_model_load_from_file(
const char * path_model,
static struct llama_model * llama_model_load_from_file_impl(
const std::string & path_model,
std::vector<std::string> & splits,
struct llama_model_params params) {
ggml_time_init();
@ -9485,7 +9480,7 @@ struct llama_model * llama_model_load_from_file(
LLAMA_LOG_INFO("%s: using device %s (%s) - %zu MiB free\n", __func__, ggml_backend_dev_name(dev), ggml_backend_dev_description(dev), free/1024/1024);
}
const int status = llama_model_load(path_model, *model, params);
const int status = llama_model_load(path_model, splits, *model, params);
GGML_ASSERT(status <= 0);
if (status < 0) {
if (status == -1) {
@ -9501,6 +9496,35 @@ struct llama_model * llama_model_load_from_file(
return model;
}
// deprecated
struct llama_model * llama_load_model_from_file(
const char * path_model,
struct llama_model_params params) {
return llama_model_load_from_file(path_model, params);
}
struct llama_model * llama_model_load_from_file(
const char * path_model,
struct llama_model_params params) {
std::vector<std::string> splits = {};
return llama_model_load_from_file_impl(path_model, splits, params);
}
struct llama_model * llama_model_load_from_splits(
const char ** paths,
size_t n_paths,
struct llama_model_params params) {
std::vector<std::string> splits;
if (n_paths == 0) {
LLAMA_LOG_ERROR("%s: list of splits is empty\n", __func__);
return nullptr;
}
for (size_t i = 0; i < n_paths; ++i) {
splits.push_back(paths[i]);
}
return llama_model_load_from_file_impl(splits.front(), splits, params);
}
struct llama_context * llama_init_from_model(
struct llama_model * model,
struct llama_context_params params) {