vizy 1.4.0

Tiny tensor visualiser: vizy.plot(t) / vizy.save(t)

vizy: One-line tensor visualization for PyTorch, JAX and NumPy

Stop juggling tensor formats. Just visualize.

Display or save any NumPy array, PyTorch tensor, or JAX array (supports 2D, 3D, 4D shapes) with a single line:

import vizy

vizy.plot(tensor)               # shows image or grid
vizy.save("image.png", tensor)  # saves to file
vizy.save(tensor)               # saves to temp file and prints path
vizy.summary(tensor)            # prints info like res, dtype, device, range, etc.

Let's say you have a PyTorch tensor or JAX array with shape (BS, 3, H, W). Instead of

plt.imshow(tensor.cpu().numpy()[0].transpose(1, 2, 0))
plt.imshow(tensor.cpu().numpy()[1].transpose(1, 2, 0))
...

You can just do:

vizy.plot(tensor)

Example output:

Or if you are in an ssh session, you can just do:

vizy.save(tensor)

It will automatically save the tensor to a temporary file and print the path, so you can scp it to your local machine and visualize it.

Technical Details

Note: This library uses PIL/Pillow for image visualization and saving. This library handles tensor format detection, conversion, and grid layout, but the actual plotting is done via PIL.

Installation

pip install vizy