I stumbled across an interesting bug while porting my StyleGAN2 stuff to flax/TPU VMs. When parallelizing using pmap, if I set my batch size wrong, parts of the output images just *vanish*. I love when neural networks fail in such visually interesting ways.
2
13
128
Here's a notebook for reproducing the effect. colab.research.google.com/dr… I'm still at a bit of a loss as to what's actually going on here. It's almost like the network weights are being corrupted by pmap somehow, but it's strange that it modifies the image in such a striking way.
1
1
8
My current guess is that it's somehow related to this (Rank promotion warning), which throws an exception if I set the error level to 'raise'. jax.readthedocs.io/en/latest… But what tensors are being reshaped incorrectly, and where? 🤔
2
Transfer learning the pretrained GNR checkpoint on my furry dataset, before the heat wave forced me to stop training because my GPU was heating up my apartment too much.
GANs N’ Roses: Stable, Controllable, Diverse Image to Image Translation (works for videos too!) pdf: arxiv.org/pdf/2106.06561.pdf abs: arxiv.org/abs/2106.06561 github: github.com/mchong6/GANsNRose…
3
3
45
Cursed samples (selfie to furry and furry to selfie).
1
13
Finally, @Kettukaa has launched a store for his foxy AI-generated art! I'm amazed at how quickly he's managed to learn and apply all the different ML skills involved in creating these. Check it out and consider supporting him. 🦊
I opened up an art store! To support a college student and get your hands on some beautiful fox art, please check out my store at Kettukaa.redbubble.com !
2
12
Teaching ponies to speak (client-side playback for github.com/yzhou359/MakeItTa…, depth estimation via github.com/intel-isl/MiDaS/, Three.js rendering)
1
4
27
I've been building tools for exploring datasets and models. Here's Danbooru + ThisAnimeDoesNotExist. The top is a UMAP clustering of the CLIP embeddings of the images. The hexagons indicate common tags for images in that region. Clicking an image lets you edit it on the bottom.
5
15
78
AI Generated Furries?? I had a blast making this video and @arfafax I absolutely love the AI you've made!! piped.video/watch?v=EiV-IFEf…
1
6
TADNE -> e621s after 2.1 million steps of transfer learning.
2
4
42
18
GAN training shows an "aha!" moment, when fakes *suddenly* look like reals during training! Embeddings: 🔴 Reals. 🔵 Fakes, with image samples on the right. ▫️ Generator creates a fake that looks a bit real ▫️ Discriminator has learned enough to start providing useful feedback
5
75
505
I'm still amazed at how powerful @AydaoAI's ThisAnimeDoesNotExist model is. So much potential behind this technology.🤯
3
22
105
Here's the notebook for playing around with this. I believe @SkyLi0n is planning on releasing a more feature-complete version with some optimizations eventually. colab.research.google.com/dr…
1
3
20
Latent directions using Ganspace.
1
10
59
1
7
1
3
Some newer interpolations.
Some heavily cherrypicked samples from transfer learning using @AydaoAI's enhanced StyleGAN2 anime model after 2 days.
1
2
10
2
6
Anime to Furry to Real to Anime
2
2
12