when I use tf 1.8 to train the code,the gpu util is 0 always

solivr / tf-crnn

TensorFlow convolutional recurrent neural network (CRNN) for text recognition

GNU General Public License v3.0

292 stars 98 forks source link

when I use tf 1.8 to train the code,the gpu util is 0 always #43

Closed WenmuZhou closed 6 years ago

WenmuZhou commented 6 years ago

solivr commented 6 years ago

I changed this line so that the device name is composed programmatically. Can you give it a new try ?

WenmuZhou commented 6 years ago

There will be fluctuations at the beginning, and it will be 0 after that and it will stop here.

solivr commented 6 years ago

Does the usage stay at zero once the training starts ? Your GPU should start being used after the buffer (Dataset) is filled, i.e after the log message:

I tensorflow/core/kernels/data/shuffle_dataset_op.cc:94] Filling up shuffle buffer (this may take a while): 1885 of 2048
I tensorflow/core/kernels/data/shuffle_dataset_op.cc:129] Shuffle buffer filled.

Otherwise, I am not able to reproduce your error...