Python Debug Plotter

Visualize NumPy, PyTorch, and Graphs while debugging — no extra dependencies required.

Last Commit

Features

View 2D single and three channel arrays as images, Nx2 and Nx3 arrays as point clouds, networkx graphs and 1D arrays as histograms.

Histogram viewer

You can explore any ND array or tensor by first creating one of the primitives above in the debug console. A Bx3xHxW tensor can be shown by first typing img = batch[4] and then plotting the image.

Batch viewer

Compound data that cannot be plotted is printed with rich information. For instance, a list with a torch tensor on the GPU, a dictionary with an array, a tensor and a pillow image and a list of 10 arrays is shown as

Data viewer

Requirements

VS Code 1.96.0 or higher
Python debugger session active

Usage

Start a Python debugging session
Set a breakpoint and pause execution
In the Variables panel, right-click on a PyTorch tensor, NumPy array, or NetworkX graph
Select "Plot variable" from the context menu
The data will be visualized in a new panel

Supported Variable Types

The extension uses a smart heuristic to automatically detect the best visualization for your data.

TYPE	DETECTION LOGIC	SPECIAL NOTES
Histogram	`ndim == 1`	NaNs and Infs are converted to `0`.
Point Cloud	`ndim == 2` AND `shape[1]` is `2` or `3`	Supports 2D (XY) and 3D (XYZ) coordinates.
Image	`ndim == 2` (width ≥ 4) OR `ndim == 3`	Supports Channels First/Last and 1, 3, or 4 channels.
Graph	Has `.nodes` and `.edges`	Looks for `pos`, `coord`, or `xy` keys for node layout.
PyTorch	Checks for `detach` and `cpu` attributes	Automatically calls `.detach().cpu().numpy()`.
Generic	Supports `np.asarray()` or `.numpy()`	Seamlessly handles Pillow images and Tensorflow tensors.
Text	Anything else	Fallback to a rich string representation.

Note on Graphs: If no coordinate information is found in the node data, the extension will attempt to use NetworkX for automatic layout calculation if available in the environment.

Change Log

See CHANGELOG.md for full release history.

Python Debug Plotter: Arrays, Tensors & Graphs

inag-ufscar