'TensorRT_model.pb' model's size is twice as large as the original pb model

ardianumam / Tensorflow-TensorRT

This repository is for my YT video series about optimizing a Tensorflow deep learning model using TensorRT. We demonstrate optimizing LeNet-like model and YOLOv3 model, and get 3.7x and 1.5x faster for the former and the latter, respectively, compared to the original models.

303 stars 110 forks source link

'TensorRT_model.pb' model's size is twice as large as the original pb model #15

Open Julyliying opened 5 years ago

Julyliying commented 5 years ago

Hi, when i use tensorrt to generate trt model, i found the FP16 or INT8 trt model's size is twice as large as the original tensorflow pb model, i think it's wired, and do you know why? my running environment is: GPU 1080Ti, CUDA10.0, cudann7.5, Tensorflow1.13, TensorRT5.0 thanks

Julyliying commented 5 years ago

Does anyone know what is wrong?

ruyijidan commented 5 years ago

my FP16 trt model is 653.5M ， I guess it‘s OK，because at GPU 970M the PFS from 10 to 20

ruyijidan commented 5 years ago

can you see the trt model fps？if you can please tell me the detail version ty!!
cudann7.5.?, Tensorflow1.13.?, TensorRT5.0.?

Julyliying commented 5 years ago

GPU 1080Ti, CUDA10.0, cudann7.5, Tensorflow1.13, TensorRT5.0. @ruyijidan

ruyijidan commented 5 years ago

@Julyliying ty！but little obvious effect at 1080ti, cudnn7.5.1.10 Tensorflow1.13,.1 TensorRT5.1.5.0