Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

Cannot find any scalar summaries in TensorBoard

I am doing mnist tutorial, and fully_connected_feed.py works and saves events.out.tfevents.1447186888 file to ~..\data\ when I trying to open TensorBoard like this

python ~/tensorflow/tensorflow/tensorboard/tensorboard.py --logdir=~/tensorflow/tensorflow/g3doc/tutorials/mnist/data 

or like this

tensorboard --logdir=~/tensorflow/tensorflow/g3doc/tutorials/mnist/data

It opens, but then I see "No scalar summary tags were found."

like image 915
Andrii Zadaianchuk Avatar asked Nov 10 '15 20:11

Andrii Zadaianchuk


People also ask

How do I view the unscaled image in a tensor?

The image is scaled to a default size for easier viewing. If you want to view the unscaled original image, check "Show actual image size" at the upper left. Play with the brightness and contrast sliders to see how they affect the image pixels. Logging one tensor is great, but what if you wanted to log multiple training examples?

How can I use tensorboard with TensorFlow to classify images?

# Load the TensorBoard notebook extension. TensorFlow 2.x selected. "This notebook requires TensorFlow 2.0 or above." You're going to construct a simple neural network to classify images in the the Fashion-MNIST dataset. This dataset consist of 70,000 28x28 grayscale images of fashion products from 10 categories, with 7,000 images per category.

How do I view the size of an image on tensorboard?

Now, use TensorBoard to examine the image. Wait a few seconds for the UI to spin up. The "Images" tab displays the image you just logged. It's an "ankle boot". The image is scaled to a default size for easier viewing. If you want to view the unscaled original image, check "Show actual image size" at the upper left.

How do I use the TensorFlow image summary API?

Using the TensorFlow Image Summary API, you can easily log tensors and arbitrary images and view them in TensorBoard. This can be extremely helpful to sample and examine your input data, or to visualize layer weights and generated tensors. You can also log diagnostic data as images that can be helpful in the course of your model development.


1 Answers

Try to use

tensorboard --logdir=home/$USER/tensorflow/tensorflow/g3doc/tutorials/mnist/data

or

tensorboard --logdir=${PWD} in that directory

Because tensorboard checks path existence by using os.path.exists()

=

Regarding that, I would like to set alias tensorboard='tensorboard --logdir=${PWD}' for convenient

like image 51
吳彥諶 Avatar answered Sep 18 '22 08:09

吳彥諶