Clay Mullis

Out of curiosity, is this something that can be enabled on Windows? I'm receiving the same error.

@mehdidc is using the captions I scraped from the Open AI dalle blog post. They definitely seem to encourage compositional diversity and zero-shot style transfer. Could be useful to you...

@rom1504 Haven't had a chance to look into this but do let me know how your experiments go as I think it's a great solution for people who need data...

Made some progress on this for a discord bot i'm working on for a personal discord. Still a work in progress and isn't very fast on spinning disk (where my...

The "zpadded_key" would just the laion400m "image_path" - but I wasn't sure that was uniform and I noticed the metadata also has an equivalent albeit unpadded "key" field. In case...

@rom1504 neat! Indeed, it's very slow. It may not matter as much with a prefetch to SSD but I'm all ears for faster alternatives. I was going to see what...

Let's see so I believe the saturation loss is from @nshepperd. Still need to track down the attribution for the remaining methods. (Thanks btw, it helps stabilize the 64x64 checkpoint...

Here is a link to the original notebook - which may include updates while the method is refined: https://colab.research.google.com/drive/12a_Wrfi2_gwwAuN3VvMTwVMz9TfqctNj

> Another thing is, diffusion models are already iterative, so the "feed forward" aspect wouldn't apply here, we still need to go through all the timesteps to generate the image....

@johndpope have you considered re-embedding the outputs from the trained `vitgan` as clip image-embeds; and then using those as prompts to a "normal" VQGAN-CLIP optimization with a much higher learning...