@curiosity_notes Profile picture

Ziyue Li

@curiosity_notes

🌈 Data Scientist @remax | sharing what I learn about software engineering, AI, data, and science | ignorant and curious

Similar User
Sam Ching photo

@samcwl

NOVA VISUALS photo

@nova_visualss

Mickmumpitz photo

@mickmumpitz

Tsinghua KEG (THUDM) photo

@thukeg

Yonglong Tian photo

@YonglongT

Maarten Bosma photo

@MaartenBosma

Oana-Maria Camburu photo

@oanacamb

Darek Kłeczek photo

@dk21

XRPoint. {X} photo

@CarloPunt

Ted Xiao photo

@xiao_ted

Brian Burns photo

@brian_a_burns

Belongie Lab photo

@BelongieLab

Jenia Jitsev 🏳️‍🌈 🇺🇦 photo

@JJitsev

Michael Zhang photo

@michaelrzhang

Justin Torre photo

@justinstorre

Pinned

LEFT: the default img2img “latent upscale” in Automatic1111 #stablediffusion RIGHT: custom “latent upscale” generated by my plugin github.com/feynlee/latent… People have long been asking for a “Hires Fix” equivalent for img2img, and my plugin fills that gap. More details in🧵…

Tweet Image 1
Tweet Image 2

Tried @runwayml's Camera Control with a photo of me as a baby👶 The model failed to orbit around objects when I gave it artistic paintings, but it deals with photos of people relatively well. #AI #2Dto3D


3 ultimate questions: 1. Existence: why something instead of nothing? (Nature of substance) 2. Events and dynamics: how does anything happen? (Nature of space-time and quantum probability) 3. Emergence: complexity, awareness/feeling/consciousness, possible illusion of will …

There are only 3 great scientific questions: 1. What's the universe made of? 2. What's life all about? 3. What is intelligence? There are interesting sub-questions: 1.1 What's dark matter and dark energy? 1.2 how do you get "it from bit" to paraphrase John Wheeler 1.3 what is…



Ziyue Li Reposted

These 94 lines of code are everything that is needed to train a neural network. Everything else is just efficiency. This is my earlier project Micrograd. It implements a scalar-valued auto-grad engine. You start with some numbers at the leafs (usually the input data and the…

Tweet Image 1

Ziyue Li Reposted

Marketing speak to terms you already know: semantic index -> embeddings app intents -> function calling on device language model -> 3B fine tuned LLM w/ included LoRA adapters on device image model -> diffusion model w/ included LoRA adapters orchestration -> Siri Neural Engine…

Tweet Image 1

This is from Apple's State of the Union The local model is a 3B parameter SLM that uses adapters trained for each specific feature. Diffusion model does the same thing, adapter for each style. Anything running locally or Apple's Secure Cloud is an Apple model, not OpenAI.



Finally got the flashcard animations to work 😅 The app is still in early development... #SwiftUI #iosdev


Just found out about this hidden feature today. Did you know that you can now share passwords with people on Mac? It works with iPhone with #iOS17, iPad with #iPadOS17, or a Mac with macOS #Sonoma. support.apple.com/guide/mac-help…

Tweet Image 1
Tweet Image 2

Don't be afraid of what you love. Cherish anything that can motivate you. Lean into anything that tugs at your heartstrings, no matter how unconventional and bizarre. Channel the desire, energy, and passion to shape yourself into the best version of who you can be.


In #SwiftUI, Text() can parse Markdown text directly. However, if the text is passed through a variable: - It must first be converted to AttributedString - Use .inlineOnly or .inlineOnlyPreservingWhitespace for interpretedSyntax to preserve line breaks.

Tweet Image 1
Tweet Image 2

Here is my solution for creating a selectable list with objects of different types in each section: - Select by using UUIDs, which are unique across types - Retrieve the object associated with the selected UUID Is there a better way to do this? #iosdev #SwiftUI

Tweet Image 1
Tweet Image 2

While watching #DunePart2 in the theater, I was strongly reminded of #AttackOnTitan and how it made me feel, though perhaps it's more accurate to frame it the other way around. Now that I think about it, they are telling the same story at the core! And I love them for that! ♥️


How to create a popover sheet that doesn't cover the entire screen: - Use detents to define the heights at which the sheet can stop. - The "upThrough" method prevents the sheet from being dragged upward to cover the entire screen. #iosdev #SwiftUI

Tweet Image 1
Tweet Image 2

Sora makes me think that, maybe in the future, 2D image generation really should be a single frame from a video generation model in some cases. Alternatively, it would be great if we can distill the more accurate 3D geometry and anatomy understanding/learning into a 2D model.

Sora from @OpenAI is super impressive, but how consistent are the geometries? We ran this through our fast 3DGS pipeline, and here are some of the early results. This is a reconstruction 👉 1/n



Ziyue Li Reposted

Sora performance scales with compute


Great video quality and consistency! 🤯

Introducing Sora, our text-to-video model. Sora can create videos of up to 60 seconds featuring highly detailed scenes, complex camera motion, and multiple characters with vibrant emotions. openai.com/sora Prompt: “Beautiful, snowy…



Two life lessons about business: - none of your business - none of my business


#AppUpdate Re-built the keyboard toolbar, fixed scrolling stutter problem and some small bugs. Convert Super Long Text to Pic: apple.co/46avN7E

Tweet Image 1

Loading...

Something went wrong.


Something went wrong.