Tweet

This Thread may be Removed Anytime!

Twitter may remove this content at anytime! Save it as PDF for later use!

More from @ak92501

AK

@ak92501

11 Oct

Using DeepSpeed and Megatron to Train Megatron-Turing NLG 530B
blog: microsoft.com/en-us/research…

selected the subset of datasets (the top 11 rows in Figure 2, below) from The Pile that we found to be of the highest relative quality. Then, following a similar approach as that used to generate Pile-CC, we downloaded and filtered two recent Common Crawl (CC) snapshots.

based evaluation setting on the open-source project lm-evaluation-harness and made task-specific changes as appropriate to align settings more closely with prior work. evaluated MT-NLG in zero-, one-, and few-shot settings without performing search for the optimal number of shots

Read 4 tweets

AK

@ak92501

3 Jul

VQGAN + CLIP "matte painting of a city built on top of a giant turtle walking slowly towards the viewer with clear blue skies and a lush green landscape | trending on artstation" + 3D photo inpainting

Read 6 tweets

AK

@ak92501

24 Jun

Fine-Tuning StyleGAN2 For Cartoon Face Generation
pdf: arxiv.org/pdf/2106.12445…
abs: arxiv.org/abs/2106.12445
github: github.com/happy-jihye/Ca…

Read 4 tweets

AK

@ak92501

23 Jun

Alias-Free GAN
pdf: nvlabs-fi-cdn.nvidia.com/alias-free-gan…
project page: nvlabs.github.io/alias-free-gan/

networks match the FID of StyleGAN2 but differ dramatically in their internal representations, and
they are fully equivariant to translation and rotation even at subpixel scales

Read 9 tweets

AK

@ak92501

8 Jun

Hierarchical Video Generation for Complex Data
pdf: arxiv.org/pdf/2106.02719…
abs: arxiv.org/abs/2106.02719

model generates a low resolution video, establishing the
global scene structure, that is then refined by subsequent levels in the hierarchy