Merge pull request #1054 from yeliang2258/fix_tensorrt_quantize_dev

Fix tensorrt quantize bugs
PaddlePaddle · Mar 15, 2023 · ed9ed99 · ed9ed99
2 parents 0af1092 + 7918064
commit ed9ed99
Showing 1 changed file with 9 additions and 4 deletions.
diff --git a/paddle2onnx/converter.cc b/paddle2onnx/converter.cc
@@ -67,7 +67,8 @@ PADDLE2ONNX_DECL bool IsExportable(const char* model, const char* params,
  P2OLogger(verbose) << "The exported ONNX model is invalid!" << std::endl;
  return false;
  }
- if (deploy_backend == "tensorrt" && calibration_str.empty()) {
+ if (parser.is_quantized_model && "tensorrt" == std::string(deploy_backend) &&
+ calibration_str.empty()) {
  P2OLogger(verbose) << "Can not generate calibration cache for TensorRT "
  "deploy backend when export quantize model."
  << std::endl;
@@ -120,7 +121,8 @@ PADDLE2ONNX_DECL bool IsExportable(const void* model_buffer, int model_size,
  P2OLogger(verbose) << "The exported ONNX model is invalid!" << std::endl;
  return false;
  }
- if (deploy_backend == "tensorrt" && calibration_str.empty()) {
+ if (parser.is_quantized_model && "tensorrt" == std::string(deploy_backend) &&
+ calibration_str.empty()) {
  P2OLogger(verbose) << "Can not generate calibration cache for TensorRT "
  "deploy backend when export quantize model."
  << std::endl;
@@ -166,7 +168,8 @@ PADDLE2ONNX_DECL bool Export(
  P2OLogger(verbose) << "The exported ONNX model is invalid!" << std::endl;
  return false;
  }
- if (deploy_backend == "tensorrt" && calibration_str.empty()) {
+ if (parser.is_quantized_model && "tensorrt" == std::string(deploy_backend) &&
+ calibration_str.empty()) {
  P2OLogger(verbose) << "Can not generate calibration cache for TensorRT "
  "deploy backend when export quantize model."
  << std::endl;
@@ -220,7 +223,9 @@ PADDLE2ONNX_DECL bool Export(
  P2OLogger(verbose) << "The exported ONNX model is invalid!" << std::endl;
  return false;
  }
- if (deploy_backend == "tensorrt" && calibration_str.empty()) {
+
+ if (parser.is_quantized_model && "tensorrt" == std::string(deploy_backend) &&
+ calibration_str.empty()) {
  P2OLogger(verbose) << "Can not generate calibration cache for TensorRT "
  "deploy backend when export quantize model."
  << std::endl;