Very slow model to render

[justinchuby]

This model is very slow to render

phi-4-mm-speech.textproto.json

[aravindmurali22]

Hi @justinchuby,

My apologies for a late response and thank you for bringing this to our attention and for sharing the model file.

We realize that rendering large and complex models (such as LLMs or multi-modal models like Phi-4) can be slow with our current visualization engine. This is primarily due to limitations in the graph layout algorithm (dagre) and the necessary client-side processing.

We are actively tracking improvements to the rendering algorithm in issue #530 to address these performance limitations.

In the meantime, would you mind confirming if the browser tab eventually renders the graph after a wait, or if it crashes or freezes completely? This information would be very helpful for distinguishing between processing delays and out-of-memory scenarios.

[justinchuby]

On a mac I think it eventually renders. On a Windows machine with lower spec I have it took longer than I was able to wait.

[justinchuby]

One important issue is if I clicked the ignored the layers (flatten) in model explorer, I would get a stack overflow error, when the namespaces are deep.

[aravindmurali22]

@justinchuby , thanks for the information. We will investigate this and get back. Thanks.