[tfjs-models/deeplab] why do we not have a save method in class SemanticSegmentation (to save the model in localstorage/indexdb)

[cu8code]

System Information

• TensorFlow.js version: 4.22.0
• Are you willing to contribute: Yes

Feature Request Description
I propose adding a save method to the SemanticSegmentation here class. Currently, I use a workaround where I save the model to IndexedDB after loading it. Here's my implementation:

let graphModel = null;
try {
    console.log("loading from memory");
    graphModel = await tfjs.loadGraphModel('indexeddb://deep');
    console.log("loaded from memory");
} catch (e) {
    console.log("failed to load from memory, loading from network");
    graphModel = await tfconv.loadGraphModel(
        modelConfig.modelUrl ||
        getURL(modelConfig.base!, modelConfig.quantizationBytes!)
    );
    console.log("saving to memory");
    graphModel.save("indexeddb://deep");
}

Impact on Current API
This addition would simplify workflows for browser-based applications.

Target Audience
Users running TensorFlow.js in the browser.

[shmishra99]

Hi @cu8code ,

Thanks for the feature request. tfjs-models provide pre-trained models and direct APIs for making predictions. The reason why not all TFJS APIs are available in the model repository is because this repository focuses on simplifying prediction for users. However, you can still leverage TFJS APIs (like the .save method) in conjunction with tfjs-models.

Could you please elaborate further on how this would benefit the community?

Let me know if i have missed anything. Thank You!!

[github-actions]

This issue has been marked stale because it has no recent activity since 7 days. It will be closed if no further activity occurs. Thank you.