Rosanne Liu

@[email protected]
166 Followers
33 Following
5 Posts

Co-founded & running ML Collective (https://mlcollective.org). Doing #MachineLearning #DeepLearning research at Google Brain->DeepMind. ICLR 2022-2024 DEI Chair. NeurIPS 2023 DIA Chair. Writing at https://deardeuce.com

Full bio here: https://rosanneliu.com/bio/

Websitehttps://rosanneliu.com
Twitterhttps://twitter.com/savvyRL
Rosanne LiuOct 26, 2023
Marcus LewisOct 24, 2023

New blog post. I tried something new, got good results, and learned some surprising nuances of GPU programming.

https://probablymarcus.com/blocks/2023/10/19/vectorizing-wide-pytorch-expressions.html

What happens when you vectorize wide PyTorch expressions?

Marcus Lewis

Marcus Lewis
Show threadRosanne LiuJan 4, 2023
@ids1024 😂
Rosanne LiuDec 22, 2022
Popular ML resourcesDec 22, 2022

The most popular Arxiv link yesterday (via noahconst@twitter):

Want your image generation model to stop misspelling everything? Try giving it access to character-level input features! https://t.co/AH6pI8VN47 https://t.co/YTOUq95F5x

https://twitter.com/noahconst/status/1605395202398507008

Character-Aware Models Improve Visual Text Rendering

Current image generation models struggle to reliably produce well-formed visual text. In this paper, we investigate a key contributing factor: popular text-to-image models lack character-level input features, making it much harder to predict a word's visual makeup as a series of glyphs. To quantify this effect, we conduct a series of experiments comparing character-aware vs. character-blind text encoders. In the text-only domain, we find that character-aware models provide large gains on a novel spelling task (WikiSpell). Applying our learnings to the visual domain, we train a suite of image generation models, and show that character-aware variants outperform their character-blind counterparts across a range of novel text rendering tasks (our DrawText benchmark). Our models set a much higher state-of-the-art on visual spelling, with 30+ point accuracy gains over competitors on rare words, despite training on far fewer examples.

arXiv.org
Rosanne LiuNov 23, 2022
I have a feeling ~6/10 "NeurIPS party" invites in my inbox are spam... for one they all started with "Dear X, thank you for organizing a NeurIPS Workshop on..." was my email scrapped from the NeurIPS site?
Rosanne LiuNov 21, 2022

For those who know me, the 2020 Uber layoff that affected me had, in retrospect, changed my life for the better. But I wouldn't have known it then. And it takes daily reminders to see it as an "opening" other than "failure."

Sharing my friend's post:
https://www.linkedin.com/posts/arpithahanumanth_amazonlayoffs-opentowork-tech-activity-7000484010917302272-0_nZ

Arpitha Hanumanth on LinkedIn: #amazonlayoffs #opentowork #tech

Hi friends, Last week I lost my job along with over 10,000 others during #amazonlayoffs. It's been a surreal experience, and my feelings towards it change...