Error converting TensorFlow model to TensorRT on Windows 10

Hello I’m trying to convert a TensorFlow model to a TensorRT model, but I have this error:
ModuleNotFoundError: No module named ‘tensorflow.python.compiler.tensorrt’

I have tried all this and has fail:
import tensorflow.contrib.tensorrt as trt
from tensorflow.python.compiler.tensorrt import trt_convert as trt
from tensorflow.python.compiler.tensorrt import trt
from tensorflow.python.compiler.tensorrt import trt_convert

I am using TensorFlow-gpu 14, CUDA 10.0 and Python 3.7.4 on Windows 10.

2 Likes

Hi juanhauara,

Unfortunately, TF-TRT isn’t yet supported for Windows.

For now, you’ll have to use Linux - Ubuntu 16.04/18.04 are probably best supported at the moment.

1 Like

ok, thanks for the answer.
Does TF-TRT work in Nvidia Jetson Nano?

Yes it does, there are some examples here: GitHub - NVIDIA-AI-IOT/tf_trt_models: TensorFlow models accelerated with NVIDIA TensorRT

And also the samples that come with the TensorRT installation (.deb/.tar) or container, are tested on Jetson as well.

Thanks,
NVIDIA Enterprise Support

1 Like

Wow i’ve been trying to run this on windows for hours :/

2 Likes

But I can’t do it when I’m in Ubuntu20.4. Can you help me? My tensorflow-gpu is 1.4.1.

Hello, I’m wondering if TF-TRT is still not supported for Windows 10?

Hi,
We recommend you to check the below samples links in case of tf-trt integration issues.

If issue persist, We recommend you to reach out to Tensorflow forum.
Thanks!