Tensorflow benchmark does not work

[xumukns]

Could you help me please to start tesnsorflow benchmark? The python script calls a bash script "t.sh" from home folder. What is it? As I commented it out, I've got:

File "fcn5_mnist.py", line 2, in import models1 as models ImportError: No module named models1 finished with execute time: .942914696 ./t.sh: Zeile 10: deactivate: Command not found.

What do I wrong?

Thanks a lot in advance

[shyhuai]

"t.sh" is the script to run the benchmark, and it receives the arguments of batch_size, epochs and deviceId. There is a typos in the file of fcn5_minst.py. I have fixed it. Please pull the newest code and try again. You can run with the command like: batch_size=1024 epochs=40 deviceId=0 ./t.sh

[xumukns]

Thank you for the fast response and help. It seems to work. Could you say please, the lines in the t.sh

... source ~/tf11/bin/activate ... deactivate ...

Can I comment it out? I get always in the log file "./t.sh: line 10: deactivate: command not found..."

And other question. if I use the option "-post True" I get:

Traceback (most recent call last): File "post_record.py", line 3, in import settings ImportError: No module named settings

where can find this part "settings"? Is it critical ?

Thanks

best regards Sergey

[shyhuai]

The deactivate command is used for virtual python environment of TensorFlow, so if you don't use the virtual environment, you can comment it out. Please use the option of "-post False" for experiments. It's a neglectable option.

[xumukns]

Thank you very much. It works.