apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-06 00:57:33 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

Add paddle quantize model support for ORT, TRT and MKLDNN deploy backend (#257)

Browse Source 
* add quantize model support for trt and paddle

* fix bugs

* fix

* update paddle2onnx version

* update version

* add quantize test

Co-authored-by: Jason <jiangjiajun@baidu.com>
This commit is contained in:
yeliang2258
2022-10-09 20:00:05 +08:00
committed by GitHub
parent ff5e798b7f
commit 2a68a23baf
10 changed files with 187 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
fastdeploy/backends/paddle/paddle_backend.cc
View File

@@ -16,13 +16,23 @@
namespace fastdeploy {
void PaddleBackend::BuildOption(const PaddleBackendOption& option) {
void PaddleBackend::BuildOption(const PaddleBackendOption& option,
const std::string& model_file) {
if (option.use_gpu) {
config_.EnableUseGpu(option.gpu_mem_init_size, option.gpu_id);
} else {
config_.DisableGpu();
if (option.enable_mkldnn) {
config_.EnableMKLDNN();
std::string contents;
if (!ReadBinaryFromFile(model_file, &contents)) {
return;
}
auto reader =
paddle2onnx::PaddleReader(contents.c_str(), contents.size());
if (reader.is_quantize_model) {
config_.EnableMkldnnInt8();
}
config_.SetMkldnnCacheCapacity(option.mkldnn_cache_size);
}
}
@@ -52,7 +62,7 @@ bool PaddleBackend::InitFromPaddle(const std::string& model_file,
return false;
}
config_.SetModel(model_file, params_file);
BuildOption(option);
BuildOption(option, model_file);
predictor_ = paddle_infer::CreatePredictor(config_);
std::vector<std::string> input_names = predictor_->GetInputNames();
std::vector<std::string> output_names = predictor_->GetOutputNames();