Skip to content

ml: don't try to convert weights if CPU backend is not available #15

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Acly merged 1 commit into from
Dec 10, 2025
Merged

ml: don't try to convert weights if CPU backend is not available #15

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 10 additions & 2 deletions src/visp/ml.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -359,6 +359,16 @@ struct tensor_converter {
dst_type = src_type;
}

backend.reset(ggml_backend_init_by_type(GGML_BACKEND_DEVICE_TYPE_CPU, nullptr));
if (!backend && whcn_to_cwhn) {
// cpu instruction set not supported, but we must do layout conversion
throw except("Cannot use CPU for tensor conversion: hardware not supported");
} else if (!backend) {
// don't convert, hope the backend can deal with the model as-is (might be slow)
dst_type = GGML_TYPE_COUNT;
return;
}

ggml_init_params ctx_params{
.mem_size = ggml_tensor_overhead() + ggml_graph_overhead(),
.mem_buffer = nullptr,
Expand All @@ -380,8 +390,6 @@ struct tensor_converter {

gallocr.reset(ggml_gallocr_new(ggml_backend_cpu_buffer_type()));
ggml_gallocr_reserve(gallocr.get(), graph);

backend.reset(ggml_backend_init_by_type(GGML_BACKEND_DEVICE_TYPE_CPU, nullptr));
}

ggml_type target_type(ggml_tensor const* t) const {
Expand Down