AttributeError: 'NoneType' object has no attribute 'serialize'

gpqls7669 · April 14, 2020, 12:07pm

Description

[TensorRT] ERROR: UFFParser: Validator error: FeatureExtractor/MobilenetV2/layer_19_2_Conv2d_5_3x3_s2_128_depthwise/BatchNorm/FusedBatchNormV3: Unsupported operation _FusedBatchNormV3
[TensorRT] ERROR: Network must have at least one output
Traceback (most recent call last):
File “main.py”, line 44, in
buf = engine.serialize()
AttributeError: ‘NoneType’ object has no attribute ‘serialize’

Environment

TensorRT Version: 5.0.6.3
GPU Type: Jetson TX2
Nvidia Driver Version: x
CUDA Version: 10.0
CUDNN Version: 7.3.1
Operating System + Version: Jetpack 4.2 + ubuntu 18.04
Python Version (if applicable): 3.6.9
TensorFlow Version (if applicable): 1.14.0
PyTorch Version (if applicable):
Baremetal or Container (if container which image + tag):

How to solve it?

SunilJB · April 14, 2020, 1:07pm

Hi,
FusedBatchNormV3 operation is currently not supported by UFF parser.

You can try to convert your model to ONNX instead of UFF using tf2onnx:

tf2onnx supports converting this op to BatchNormalization op in ONNX:

github.com/onnx/tensorflow-onnx

tf2onnx/onnx_opset/nn.py

main


      
          if output_shape is not None:
              if is_channels_last(node):
                  new_output_shape = [output_shape[1], output_shape[2]]
                  if spatial == 3:
                      new_output_shape.append(output_shape[3])
              else:
                  new_output_shape = [output_shape[2], output_shape[3]]
                  if spatial == 3:
                      new_output_shape.append(output_shape[4])
          
              utils.make_sure(new_output_shape.count(-1) <= 0, "output dims need to be known")
              utils.make_sure(all(new_output_shape[i] >= input_dims[i] for i in range(spatial)),
                              "output dims cannot be smaller than input dims.")
              if -1 in input_dims:
                  node.set_attr("output_shape", new_output_shape)
              else:
                  if "strides" in node.attr:
                      strides = parse_dims_attr(node, node.get_attr("strides").ints, spatial)
                  else:
                      strides = [1] * spatial
                  if "dilations" in node.attr:

And BatchNormalization op is supported by the TensorRT ONNX parser:
https://github.com/onnx/onnx-tensorrt/blob/master/operators.md

I will also recommend you to use latest TRT version.

Thanks

gpqls7669 · April 19, 2020, 10:46am

Is it working on jetson tx2 and jetpack 4.2?

Topic		Replies	Views
About TRT-object-detecction TensorRT	7	466	April 27, 2020
unsupported operation _FusedBatchNormV3 TensorRT	1	1888	November 21, 2019
TensorRT 6.0.1 + TensorFlow 1.14 - No conversion function registered for layer: FusedBatchNormV3 yet TensorRT	4	2771	February 16, 2020
About TRT-object-detecction Jetson TX2 tensorrt	1	359	April 21, 2020
Error about tensorrt in jetson tx2 Jetson TX2 tensorrt	1	749	April 21, 2020
TensorRT engine - Impossible to parse UFF file Jetson Nano	5	1195	February 19, 2020
Get an Error when Convert frozen graph to TensorRT engine Jetson Nano	2	751	January 13, 2020
uff_to_trt does not convert BatchNorm op TensorRT	0	831	August 5, 2018
TF to UFF issue TensorRT	5	2702	November 27, 2018
How to run custom model on Jteson nano Jetson Nano	5	1043	February 21, 2020

AttributeError: 'NoneType' object has no attribute 'serialize'

Description

Environment

Related topics