TensorBoard metrics

Save TensorBoard logs

You can use TensorBoard in Onepanel by placing your log file into the following folder:

/onepanel/output

PyTorch

from tensorboardX import SummaryWriter

# Save logs to `/onepanel/output`
writer = SummaryWriter('/onepanel/output')
...
writer.add_scalar('Accuracy', accuracy, iteration)

TensorFlow

import tensorflow as tf

# Save logs to `/onepanel/output`
tensorboard = tf.keras.callbacks.TensorBoard(log_dir='/onepanel/output', batch_size=32, write_images=True)
...
model.fit(x_train, y_train, epochs=10, callbacks=[tensorboard])

View TensorBoard

You can view TensorBoard for a running job by clicking TensorBoard in Job details: