Skip to content

fix: Bugfix in TRT Engine deserialization indexing #1646

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
merged 1 commit into from
Feb 3, 2023
Merged

fix: Bugfix in TRT Engine deserialization indexing #1646

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
9 changes: 4 additions & 5 deletions core/runtime/TRTEngine.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -128,14 +128,13 @@ TRTEngine::TRTEngine(
for (size_t pyt_idx = 0; pyt_idx < outputs; pyt_idx++) {
auto binding_name = _out_binding_names[pyt_idx];
auto trt_idx = cuda_engine->getBindingIndex(binding_name.c_str());
std::string engine_binded_name = cuda_engine->getIOTensorName(inputs_size + pyt_idx);
Comment on lines 130 to -131
Copy link
Collaborator Author

@gs-olive gs-olive Feb 2, 2023 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Original issue occurs here, since trt_idx != inputs_size + pyt_idx in general (for example, when there are multiple output tensors from the TRT Engine)

TORCHTRT_CHECK(
(binding_name == engine_binded_name),
"Could not find a TensorRT engine binding for output named " << binding_name);
TORCHTRT_CHECK((trt_idx != -1), "Could not find a TensorRT engine binding for output named " << binding_name);
TORCHTRT_CHECK(
!(cuda_engine->getTensorIOMode(binding_name.c_str()) == nvinfer1::TensorIOMode::kINPUT),
"Binding " << binding_name << " specified as output but found as input in TensorRT engine");
LOG_DEBUG("Output binding name: " << binding_name << "pyt return idx: " << inputs_size + pyt_idx << ")");
LOG_DEBUG(
"Output binding name: " << binding_name << " has TensorRT binding index: " << trt_idx
<< ", Torch binding index: " << inputs_size + pyt_idx);
out_binding_map[trt_idx] = pyt_idx;
out_binding_names[pyt_idx] = binding_name;
}
Expand Down