I participated in the previous hackathon and started a new GitHub repo where I improved my project submission. I’m now getting ready to publish the extension to the Chrome Web Store.
However, I ran into a major issue today: when testing on another computer, the model simply doesn’t download. I had the same problem earlier when a friend tried to help me develop — for days he struggled to get the model to download properly, and only eventually it worked by itself.
I’ve set both flags (device optimization and run_on_prompt_api), and the “language model availability” check reports that the model is downloadable.
But both my implementation and the Chrome sample (https://developer.chrome.com/docs/ai/inform-users-of-model-download) remain stuck at 0% forever.
This is really frustrating.
Does anyone have any idea what might be going on or how to fix it?
--
You received this message because you are subscribed to the Google Groups "Chrome Built-in AI Early Preview Program Discussions" group.
To unsubscribe from this group and stop receiving emails from it, send an email to chrome-ai-dev-previe...@chromium.org.
To view this discussion visit https://groups.google.com/a/chromium.org/d/msgid/chrome-ai-dev-preview-discuss/9a7526f8-39bb-40d3-b9c7-00aeebc7fdd4n%40chromium.org.
I tested it with other friends, and they were able to use it without enabling any flags(I only use prompt api, not multimodal), it downloaded successfully and everything. For others, it correctly showed the warning that the feature wasn’t supported.
It’s only this one computer that claims the model is available and downloadable, but it never actually downloads. This is the machine I was trying to install it on myself. Honestly, I don’t really want to dive into debugging this right now. I’m okay with the fact that it works as intended on most machines, and that on this one it silently fails without any errors.