TensorBloom

A graphical editor for PyTorch neural networks. See your model, train it, export it.

GitHub Features
TensorBloom

Visual Graph Editor

Build networks as node graphs. 70+ PyTorch modules you can drag, connect, and configure. Real-time shape inference catches dimension errors as you build.

One-Click Training

Train from the editor with live loss and accuracy charts. 13 built-in datasets, 4 optimizers, learning rate schedulers, mixed precision, and early stopping.

Smart Auto-Fix

Click Start Training and the auto-fix system corrects shape mismatches, adjusts dimensions, and validates loss-target compatibility before the first epoch.

Download

Grab the latest installer for your platform. Requires Python 3.9+ (PyTorch is installed automatically on first launch).

Download for Windows / Linux

Or build from source: see README

Documentation