Skip to content
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

disable int8 if there is no quant info #36900

Merged
merged 2 commits into from
Nov 1, 2021
Merged

disable int8 if there is no quant info #36900

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
2eb8403
disable int8
b3602sss Nov 1, 2021
a971c1f
size_t to int
b3602sss Nov 1, 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
9 changes: 9 additions & 0 deletions paddle/fluid/inference/tensorrt/engine.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -148,12 +148,21 @@ void TensorRTEngine::FreezeNetwork() {
// and outputs have scales,
// this layer's precision and output type are set to float32.
// This step has no effect if this layer is fused during TRT optimization.
int layers_no_int8 = 0;
for (int i = 0; i < network()->getNbLayers(); i++) {
auto layer = network()->getLayer(i);
if (!is_layer_int8(layer)) {
layer->setPrecision(nvinfer1::DataType::kFLOAT);
++layers_no_int8;
}
}
// Disable int8 or build engine failed if all layers aren't int8
if (layers_no_int8 == network()->getNbLayers()) {
nvinfer1::BuilderFlags flags = infer_builder_config_->getFlags();
flags = flags & ~(1U << static_cast<int>(nvinfer1::BuilderFlag::kINT8));
// reset flags
infer_builder_config_->setFlags(flags);
}
#else
LOG(WARNING) << "If your TensorRT version is lower than 5.1.2.2, you "
"must provide quantization scales for all tensors using "
Expand Down