In this dive we go into one of the papers that inspired Flux, the new state-of-the-art generative image model.

--

Use Oxen AI 🐂 https://oxen.ai/

Oxen AI makes versioning your datasets as easy as versioning your code! Even is millions of unstructured images, the tool quickly handles any type of data so you can build cutting-edge AI.

--

Paper, Links, + Notes 📝 https://www.oxen.ai/blog/arxiv-dives

Join arXiv Dives 🤿 https://oxen.ai/community

Discord 🗿   / discord  

--

Chapters
0:00 Intro to Flux
1:08 Rectified Flow Transformers
2:00 What Do Diffusion Models Do
4:05 Understanding the Loss Equation
13:42 How Latent Diffusion Transformers Work
16:39 Rectified Flow Transformers Architecture
19:55 Rectified Flow Transformer Diagram
21:00 The Datasets They Used
21:45 Improved Captions Synthetic Data
23:50 Data Preprocessing
25:13 Our Results Fine-Tuning