Use TensorBoard for visualization experiment result in machine learning

Prepare development environment: Python 3.9.6 ; pip 21.1.3 ; jupyter-notebook 6.4.0 ; tensorflow 2.5.0 .

If you run by command line

Create file tensorboard_pytorch.py

from torch.utils.tensorboard import SummaryWriter writer = SummaryWriter() for x in range(5): y = 100 * x writer.add_scalar('y', y, x) writer.close()

(c1)

Result is

Use your web browser, go to address http://localhost:6006 to see the result

If you run by jupyter notebook

Create a new jupyter notebook

pin install jupyter
jupyter notebook

Code block is the same, the content is

from torch.utils.tensorboard import SummaryWriter writer = SummaryWriter() for x in range(5): y = 100 * x writer.add_scalar('y', y, x) writer.close()

(c1)

Load extension TensorBoard by

%load_ext tensorboard

Visualize the result of code snippet (c1) by

%tensorboard --logdir ./runs/

Result is

SummaryWriter methods are:

• add_scalar
• add_scalars
• add_custom_scalars
• add_historgram
• add_image
• add_images
• add_figure
• add_video
• add_audio
• add_text