tensorflow
Using tensorboard
I am trying to store information for"loss" and accuracy for tensorboard for each iteration. However, could not figure out which file to look into for this. Any comments.
Can you clarify? In tf_cnn_benchmarks, loss is already recorded and accessible in TensorBoard. Training accuracy is also recorded IIRC if --print_training_accuracy is specified.
@reedwm I am interested in tracing the loss function and would like to visualize it. I am getting it now. However, I would like to use tf_profile. I downloaded TF using virtual environment. However, I can't activate it? In the example, it seems that you need to download TF source code.
This seems like a general TensorFlow issue, not a benchmarks issue. You do not need to install TF from source, but you do need the latest nightly pip package which you can get by running pip install tf-nightly-gpu in a virtual environment. I recommend asking on StackOverflow if you are having trouble activating a virtual environment.