Project structure of: lucidrains/DALLE2-pytorch

• DALLE2-pytorch DALL-E 2: Image generation & testing, with diffusion prior model training code.
  • configs
    • README.md Configure DALLE2 model training options in PyTorch
  • dalle2_pytorch Trains VAE models with DALL-E 2 image library.
    • __init__.py DALLE2-pytorch library import for image generation.
    • cli.py DALL-E2 CLI PyTorch image generator.
    • dalle2_pytorch.py DALL-E 2 PyTorch: Image generation using AI libraries.
    • dataloaders Simple image loaders for PyTorch DALL-E 2.
      • __init__.py Imports datasets for DALLE2-pytorch data loading.
      • decoder_loader.py Decoder loader: embeddings, datasets, webdataset exceptions.
      • prior_loader.py Efficient DALL-E 2 data retrieval with PyTorch Dataloaders.
      • README.md Dataloader for image embeddings with 3 ranks.
      • simple_image_only_dataloader.py Simple image data loader for PyTorch. Dataset class, DataLoader function.
    • optimizer.py Create optimizer for Adam/AdamW with weight decay filtering.
    • tokenizer.py PyTorch BPE tokenizer for DALL-E2, efficient & human-readable.
    • trackers.py Trackers and loggers initialization, logging, saving, error handling.
    • train_configs.py Trains DALL-E 2 PyTorch with configs and utilities.
    • trainer.py DeepSpeed Trainer: Model Params, Distrib., Precision & Loss
    • utils.py Helper functions for time, print, and import.
    • version.py DALLE2-pytorch library version: '1.15.6'
    • vqgan_vae.py VQGAN-VAE model for image gen. with conv, self-attn, layer norm
    • vqgan_vae_trainer.py Train VAE model using ImageDataset and VQGanVAETrainer classes.
  • Makefile Install pip, project in editable mode, run tests with CUDA.
  • MANIFEST.in Includes all .txt files in dalle2_pytorch directory.
  • prior.md Diffusion-based image generation with CLIP
  • README.md Enhance DALL-E 2 with layers, generation, and prior models.
  • setup.py Setup script for dalle2-pytorch package, AI developers target.
  • train_decoder.py Trains DALL-E 2 UNet models with distributed support.
  • train_diffusion_prior.py Diffusion Prior model training code.