Quick Look visualizer for neural networks/machine learning models based on Lutz Roeder's Netron.
Supports most neural network and machine learning models, configured for twenty frameworks such as
CoreML,torch,keras,tensorflow,onnxand more 1.
Does everything the official Netron2 does - except telemetry, disabled here.
Familiar interface - it's Netron, adapted for Quick Look.
Lightweight - 45MB, 10 times less than electron-based.
No web-server, no python, nothing external nor internet needed - made in Swift for macOS.
Download from Releases
Launch the app once. This registers the extension as the extension itself stays within the app bundle.
Delete the app.
Replace the older app with the newer app. MacOS looks for the extension in the app: if app is moved, a relaunch might be needed.
Works on all macOSX versions (Ventura to latest Sequoia) and all Silicon or Intel mac architectures.
Direct download (1.0 Universal Binary)
opt + spacequicklooks in full-screen.
Once the app is launched, the Quicklook Netron extension appears in Extensions from General in macOS' System Settings, and disapears if app is deleted.
Notarized and gatekeeper-approved.
Screen.Recording.2025-03-26.at.1.05.20.AM.webm
Does need some colors, but light and fast.
Made in Swift.
a mac is a mac and it needs to stay comfortable, but it's still your mac.
Footnotes
-
"onnx" "pb" "tflite" "mlmodel" "mlpackage" "pt" "pth" "h5" "keras" "safetensors" "weights" "bin" "torch" "pkl" "npy" "params" "caffemodel" "prototxt" "pbtxt" "dlc" ↩
-
Quicklook-Netron is not affiliated with, but uses code from Lutz Roeder's Netron. ↩