search

google-ai-edge / model-explorer

A modern model graph visualizer and debugger
Apache License 2.0
1k stars 75 forks source link

[Feature Request] Shaper text #55

Open justinchuby opened 4 months ago

justinchuby commented 4 months ago

Shown in https://github.com/google-ai-edge/model-explorer/issues/54, netron is able to display the graph more clearly using the same screen real estate. Things are more legible.

1v1 comparison

image image
justinchuby commented 4 months ago

Model: https://github.com/onnx/models/blob/main/validated/text/machine_comprehension/gpt-2/model/gpt2-10.onnx

pkgoogle commented 4 months ago

Hi @justinchuby, I think I agree in this case conciseness helps as most people don't think about the constant ops/initializer ops -- wondering if this should be an option -- like "hyper-detail" option or simply make those nodes have another level of double-clickability. Either way we'll look into it.

justinchuby commented 4 months ago

I think displaying constants are ok. Even zooming out more Netron was able to render text more clearly. Maybe something to do with antialiasing in OpenGL? I am by no means an expert. A heavier font weight may help?