GPU usage is zero? about east HOT 9 CLOSED

I was facing the same issue. Turned out that the memory of my GPU was not sufficient for the task.

from east.

I thinks its a tensorflow problem, you can check your tensorflow.

from east.

There are some logs about reading images. Please check whether you have encounter some thing like:

0 training images in icdar15_training_data
0 training images in icdar15_training_data
0 training images in icdar15_training_data
0 training images in icdar15_training_data
...

from east.

@zxytim @argman
my log like this:
1000 training images in icdar15_training_data
1000 training images in icdar15_training_data
1000 training images in icdar15_training_data
1000 training images in icdar15_training_data
1000 training images in icdar15_training_data
1000 training images in icdar15_training_data
1000 training images in icdar15_training_data
1000 training images in icdar15_training_data
1000 training images in icdar15_training_data
1000 training images in icdar15_training_data
After print out 10 lines of such log row, the the model seems suspend, no longer give new output one the command line. And I locate the the function get_batch() of icdar.py, I have question about the queue manager codes, not sure whether the problem caused by that.

from east.

@argman Could you share the version of tensorflow you used?

from east.

from east.

My current tensorflow version is also 1.2.1, but it has the above problem...

from east.

@argman @zxytim What python version that you used in this project? python2.7 or python3.0+?

from east.

from east.