TensorBoard provides great suite of visualization tools to help understand, debug and optimize your TensorFlow or PyTorch programs.  

In Onepanel, you can use the built-in TensorBoard by saving your TensorFlow and PyTorch logs (using tensorboardx ) in the /onepanel/output  directory.

You can then view your visualizations by clicking the "TensorBoard" from the services drop-down menu:

TensorFlow and Keras Example

import tensorflow as tf

# Log TensorBoard event files into `/onepanel/output`
tensorboard = tf.keras.callbacks.TensorBoard(log_dir='/onepanel/output', batch_size=32, write_images=True)

# Model code here...

Fork this example project in Onepanel to get started.

PyTorch Example with TensorboardX

from tensorboardX import SummaryWriter

writer = SummaryWriter('/onepanel/output')

# Write Scalar to TensorBoard
writer.add_scalar('Train/Loss', loss, iteration)

Fork this example project in Onepanel to get started.

Did this answer your question?