Results?

[mindplay-dk]

The project has been 1.x for a while now - are there any updated results to show?

The README still shows the same early sample image with some flowers.

Is the issue still with training and compute resources? As mentioned, you can apply with Google for free cloud TPU resources - I think you'd have to agree to share the results and trained weights, but I would guess you're willing to do that either way?

I've been following this project from early onset - it sounds very promising, and I'm very curious to see what this project can do, but I don't guess we're going to see anything (beyond code and metrics mainly of interest to ML engineers) until compute resources are allocated at scale?

Chewing through 5 billion images isn't really within reach for ML engineers or end users - probably not even to AI image companies, unless maybe they're keeping the results to themselves and trying to exploit it commercially.

How far are you from being able to train the algorithm and show something? Do you have plans? ☺