Skip to content

LLM inference in C/C++

C++ ggml

arg: Skip mmproj download when user supplied mmproj (#24239)

konradmb465b1f0e75c5
diff --git a/common/arg.cpp b/common/arg.cpp
index 1ffaf70485..a859aac4fe 100644
--- a/common/arg.cpp
+++ b/common/arg.cpp
@@ -444,7 +444,7 @@ bool common_params_handle_models(common_params & params, llama_example curr_ex)
opts.offline = params.offline;
opts.skip_download = params.skip_download;
opts.download_mtp = spec_type_draft_mtp;
- opts.download_mmproj = !params.no_mmproj;
+ opts.download_mmproj = !params.no_mmproj && params.mmproj.path.empty() && params.mmproj.url.empty();
// sub-models (draft, mmproj, vocoder) are explicitly specified by the user,
// so we should not auto-discover mtp/mmproj siblings for them