Show HN: Easily visualize torch, Jax, tf, NumPy, etc. tensors (github.com)

A new Python library, tensordiagram, makes it easy to visualize tensors from PyTorch, JAX, TensorFlow, NumPy and other array libraries inside notebooks and REPLs. Built on the chalk graphics backend, it turns n‑dimensional arrays into annotated diagrams so you can see shapes, slices and layout at a glance—much more intuitive than printing shapes or raw values. The project targets debugging and education: it helps you reason about complex ops (e.g., unsqueeze/transpose/view patterns), inspect slices, and communicate tensor layouts in tutorials or bug reports. Technically, tensordiagram exposes td.to_diagram(...) for converting arrays, rendering to PNG or SVG (render_png/render_svg), and a small fluent API for styling and annotation (fill_color, fill_values, annotate_dim_size). It supports coloring by index/value (via lambdas), composing multiple tensor views into a single chalk diagram (to_chalk_diagram and chalk.hcat), and handles 2D/3D tensors cleanly. Install with pip install tensordiagram and choose extras for cairo (PNG) or SVG backends; libcairo2-dev may be required on Debian for PNG. The repo includes visual regression tests (generate_references.py, pytest), developer install options, and ships under the MIT license—making it a lightweight, practical tool for everyday ML debugging and teaching.