Superminaren/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-09-27 21:46:57 +02:00
Code Issues Packages Projects Releases Wiki Activity

server: Dedup the draft HF model via dedup-cache-models (#27934)

Browse Source 
* server: Dedup the draft HF model via dedup-cache-models
Fixes #27846

* server: avoid capturing structured binding in lambda
This commit is contained in:
Xie Wenxiang
2026-09-23 12:57:53 +02:00
committed by GitHub
parent 26758d38f9
commit bcbc936a87
2 changed files with 26 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tools/server/server-models.cpp
+14 -10
View File
@@ -734,21 +734,25 @@ void server_models::load_models() {
std::set<std::string> hidden_models;
{
std::set<std::string> preset_paths;
auto add_hf_path = [&preset_paths](const common_preset & preset, const char * repo_key, const char * file_key) {
std::string hf_repo;
if (!preset.get_option(repo_key, hf_repo) || hf_repo.empty()) {
return;
}
std::string hf_file;
preset.get_option(file_key, hf_file);
std::string path = common_download_resolve_path(hf_repo, hf_file);
if (!path.empty()) {
preset_paths.insert(path);
}
};
for (const auto & [name, preset] : custom_presets) {
std::string val;
if (!preset.get_option(COMMON_ARG_PRESET_DEDUP_CACHE_MODELS, val) || !common_arg_utils::is_truthy(val)) {
continue;
}
std::string hf_repo;
if (!preset.get_option("LLAMA_ARG_HF_REPO", hf_repo) || hf_repo.empty()) {
continue;
}
std::string hf_file;
preset.get_option("LLAMA_ARG_HF_FILE", hf_file);
std::string path = common_download_resolve_path(hf_repo, hf_file);
if (!path.empty()) {
preset_paths.insert(path);
}
add_hf_path(preset, "LLAMA_ARG_HF_REPO", "LLAMA_ARG_HF_FILE");
add_hf_path(preset, "LLAMA_ARG_SPEC_DRAFT_HF_REPO", "LLAMA_ARG_SPEC_DRAFT_MODEL");
}
if (!preset_paths.empty()) {
for (const auto & [name, preset] : cached_models) {