Shypd.aiThe largest open index of AI toolsmindall-e is an open-source image generation tool that provides a PyTorch implementation of a 1.3B text-to-image generation model. It was trained on 14 million image-text pairs for non-commercial purposes.
Trending
About
mindall-e is a PyTorch implementation of a 1.3B text-to-image generation model, trained on 14 million image-text pairs for non-commercial use. This open-source tool, named after minGPT, utilizes a two-stage autoregressive model. It replaces the original DALL-E's Discrete VAE with VQGAN for more effective high-quality sample generation. The model can generate candidate images from a text prompt and re-rank them using OpenAI's CLIP. It also supports transfer learning for class-conditional and unconditional generation tasks, allowing fine-tuning on datasets like ImageNet. The repository provides code snippets for sampling and interactive demos, along with quantitative results demonstrating its performance against other models.
Capabilities
Pricing & Plans
Open Source
Free
FAQs
Trending
