Completed
- Model Architecture
Class Central Classrooms beta
YouTube videos curated by Class Central.
Classroom Contents
OpenAI DALL·E - Creating Images from Text - Blog Post Explained
Automatically move to the next video in the Classroom when playback concludes
- 1 - Introduction
- 2 - Overview
- 3 - Dataset
- 4 - Comparison to GPT-3
- 5 - Model Architecture
- 6 - VQ-VAE
- 7 - Combining VQ-VAE with GPT-3
- 8 - Pre-Training with Relaxation
- 9 - Experimental Results
- 10 - My Hypothesis about DALL·E's inner workings
- 11 - Sparse Attention Patterns
- 12 - DALL·E can't count
- 13 - DALL·E can't global order
- 14 - DALL·E renders different views
- 15 - DALL·E is very good at texture
- 16 - DALL·E can complete a bust
- 17 - DALL·E can do some reflections, but not others
- 18 - DALL·E can do cross-sections of some objects
- 19 - DALL·E is amazing at style
- 20 - DALL·E can generate logos
- 21 - DALL·E can generate bedrooms
- 22 - DALL·E can combine unusual concepts
- 23 - DALL·E can generate illustrations
- 24 - DALL·E sometimes understands complicated prompts
- 25 - DALL·E can pass part of an IQ test
- 26 - DALL·E probably does not have geographical / temporal knowledge
- 27 - Reranking dramatically improves quality
- 28 - Conclusions & Comments
