subgraph conversion error

AI & Data Science Deep Learning (Training & Inference) TensorRT
1

[Platform details]
Linux distro and version: Ubuntu 16.04
GPU type: P$
nvidia driver version: 384.111
CUDA version: 9.0.176
CUDNN version: 7.1.3
Python version:3.4
Tensorflow version: 1.18

when i use tensorrt, there is an error. what does it mean?

import tensorflow.contrib.tensorrt as trt

trt_frozen_graph = trt.create_inference_graph(
input_graph_def=frozen_graph,
outputs=[‘feature/dense/BiasAdd’],
max_batch_size=100,
max_workspace_size_bytes=1 << 26,
#precision_mode=‘FP16’,
precision_mode=‘FP32’,
minimum_segment_size=2
)

2018-11-13 12:01:44.779864: W tensorflow/contrib/tensorrt/convert/convert_graph.cc:418] subgraph conversion error for subgraph_index:19 due to: “Invalid argument: Failed to create Input layer” SKIPPING…( 7 nodes)
2018-11-13 12:01:44.780510: E tensorflow/contrib/tensorrt/log/trt_logger.cc:38] DefaultLogger Parameter check failed at: …/builder/Network.cpp::addInput::368, condition: inName != knownInputs->name

2

Hello,

what version of TensorRT are you using?

3

I installed the TensorRT with TensorRT-4.0.1.6.Ubuntu-14.04.5.x86_64-gnu.cuda-9.0.cudnn7.1.tar.gz.

4

Hello,

Per engineering, this issue should be fixed in 18.11 or TF 1.12.

regards,
NVIDIA Enterprise Support