Skip to content

fix: support c10::List type of output in MarkOutputs #419

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

Closed
wants to merge 2 commits into from
Closed

fix: support c10::List type of output in MarkOutputs #419

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
150643d
fix: support c10::List type of output in MarkOutputs
inocsin Apr 1, 2021
053567e
chore: update error message and function name of MarkOutputs
inocsin Apr 2, 2021
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
31 changes: 21 additions & 10 deletions core/conversion/conversion.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -174,23 +174,34 @@ void AddInputs(ConversionCtx* ctx, at::ArrayRef<const torch::jit::Value*> inputs
#endif
}

void MarkIValueOutputs(ConversionCtx* ctx, c10::IValue out_ivalue, const torch::jit::Value* out) {
if (out_ivalue.isCustomClass()) {
std::string name = std::string("output_") + std::to_string(ctx->num_outputs);
auto output_container = out_ivalue.toCustomClass<TensorContainer>();
nvinfer1::ITensor* out_tensor = output_container.get()->tensor();
out_tensor->setName(name.c_str());
ctx->net->markOutput(*out_tensor);
LOG_INFO(
ctx->logger, "Marking Output " << out->debugName() << " named " << name << " in engine (ctx.MarkOutput)");
ctx->num_outputs += 1;
} else {
TRTORCH_THROW_ERROR("Unsupported output type, only Tensors or unwrapped collections of Tensors can be marked as engine outputs but found type: " << out_ivalue.tagKind());
}
}

void MarkOutputs(ConversionCtx* ctx, at::ArrayRef<const torch::jit::Value*> outputs) {
for (auto out : outputs) {
auto it = ctx->value_tensor_map.find(out);
if (it == ctx->value_tensor_map.end()) {
if (ctx->evaluated_value_map.find(out) != ctx->evaluated_value_map.end()) {
auto out_ivalue = ctx->evaluated_value_map[out];
if (out_ivalue.isCustomClass()) {
std::string name = std::string("output_") + std::to_string(ctx->num_outputs);
auto output_container = out_ivalue.toCustomClass<TensorContainer>();
nvinfer1::ITensor* out_tensor = output_container.get()->tensor();
out_tensor->setName(name.c_str());
ctx->net->markOutput(*out_tensor);
LOG_INFO(
ctx->logger, "Marking Output " << out->debugName() << " named " << name << " in engine (ctx.MarkOutput)");
ctx->num_outputs += 1;
if (out_ivalue.isList()) {
c10::List<c10::IValue> value_list = out_ivalue.toList();
for(auto it = value_list.begin(); it != value_list.end(); it++) {
MarkIValueOutputs(ctx, *it, out);
}
} else {
TRTORCH_THROW_ERROR("Unknown output type. Only a single tensor or a TensorList type is supported.");
MarkIValueOutputs(ctx, out_ivalue, out);
}
}
} else {
Expand Down