Provide details on the platforms you are using:
Linux distro and version: Ubuntu 18.04
GPU type: NVIDIA GeForce GTX 1080 Ti Founder’s Edition
nvidia driver version: 418.56
CUDA version: 10.0
CUDNN version: 7.5.0
Python version [if using python]: 3.6.7
TensorRT version: 5.1.5
Following the instructions per the TensorRT manual, I used the frozen model to implement Tensorflow-RT:
import tensorflow.contrib.tensorrt as trt
trt_graph = trt.create_inference_graph(input_graph_def=frozen_graph, outputs=[out.op.name for out in model.outputs], max_batch_size=1,max_workspace_size_bytes=2 << 20, precision_mode=“fp16”)
tf.train.write_graph(trt_graph, “model”, “tfrt_model.pb”, as_text=False)
However, then I get the error: WARNING:tensorflow:TensorRT mismatch. Compiled against version 5.0.2, but loaded 5.1.5. Things may not work.
What exactly is not lining up? Does CUDA need to be upgraded to 10.1? Or should I downgrade my TensorRT? Am I missing something? Any help would be greatly appreciated.