HandsomeD123/fast-DiT

Improved PyTorch Implementation of "Scalable Diffusion Models with Transformers"

Scalable Diffusion Models with Transformers (DiT)
_{Improved PyTorch Implementation}

Paper | Project Page | Run DiT-XL/2

This repo features an improved PyTorch implementation for the paper Scalable Diffusion Models with Transformers.

It contains:

• 🪐 An improved PyTorch implementation and the original implementation of DiT
• ⚡️ Pre-trained class-conditional DiT models trained on ImageNet (512x512 and 256x256)
• 💥 A self-contained Hugging Face Space and Colab notebook for running pre-trained DiT-XL/2 models
• 🛸 An improved DiT training script and several training options

Setup

First, download and set up the repo:

git clone https://github.com/chuanyangjin/fast-DiT.git
cd DiT

We provide an environment.yml file that can be used to create a Conda environment. If you only want to run pre-trained models locally on CPU, you can remove the cudatoolkit and pytorch-cuda requirements from the file.

conda env create -f environment.yml
conda activate DiT

Sampling

Pre-trained DiT checkpoints. You can sample from our pre-trained DiT models with sample.py. Weights for our pre-trained DiT model will be automatically downloaded depending on the model you use. The script has various arguments to switch between the 256x256 and 512x512 models, adjust sampling steps, change the classifier-free guidance scale, etc. For example, to sample from our 512x512 DiT-XL/2 model, you can use:

python sample.py --image-size 512 --seed 1

For convenience, our pre-trained DiT models can be downloaded directly here as well:

DiT Model	Image Resolution	FID-50K	Inception Score	Gflops
XL/2	256x256	2.27	278.24	119
XL/2	512x512	3.04	240.82	525

Custom DiT checkpoints. If you've trained a new DiT model with train.py (see below), you can add the --ckpt argument to use your own checkpoint instead. For example, to sample from the EMA weights of a custom 256x256 DiT-L/4 model, run:

python sample.py --model DiT-L/4 --image-size 256 --ckpt /path/to/model.pt

Training

Preparation Before Training

To extract ImageNet features with 1 GPUs on one node:

torchrun --nnodes=1 --nproc_per_node=1 extract_features.py --model DiT-XL/2 --data-path /path/to/imagenet/train --features-path /path/to/store/features

Training DiT

We provide a training script for DiT in train.py. This script can be used to train class-conditional DiT models, but it can be easily modified to support other types of conditioning.

To launch DiT-XL/2 (256x256) training with 1 GPUs on one node:

accelerate launch --mixed_precision fp16 train.py --model DiT-XL/2 --features-path /path/to/store/features

To launch DiT-XL/2 (256x256) training with N GPUs on one node:

accelerate launch --multi_gpu --num_processes N --mixed_precision fp16 train.py --model DiT-XL/2 --features-path /path/to/store/features

Alternatively, you have the option to extract and train the scripts located in the folder training options.

PyTorch Training Results

We've trained DiT-XL/2 and DiT-B/4 models from scratch with the PyTorch training script to verify that it reproduces the original JAX results up to several hundred thousand training iterations. Across our experiments, the PyTorch-trained models give similar (and sometimes slightly better) results compared to the JAX-trained models up to reasonable random variation. Some data points:

DiT Model	Train Steps	FID-50K (JAX Training)	FID-50K (PyTorch Training)	PyTorch Global Training Seed
XL/2	400K	19.5	18.1	42
B/4	400K	68.4	68.9	42
B/4	400K	68.4	68.3	100

These models were trained at 256x256 resolution; we used 8x A100s to train XL/2 and 4x A100s to train B/4. Note that FID here is computed with 250 DDPM sampling steps, with the mse VAE decoder and without guidance (cfg-scale=1).

Improved Training Performance

In comparison to the original implementation, we implement a selection of training speed acceleration and memory saving features including gradient checkpointing, mixed precision training, and pre-extracted VAE features, resulting in a 95% speed increase and 60% memory reduction on DiT-XL/2. Some data points using a global batch size of 128 with a A100:

gradient checkpointing	mixed precision training	feature pre-extraction	training speed	memory
❌	❌	❌	-	out of memory
✔	❌	❌	0.43 steps/sec	44045 MB
✔	✔	❌	0.56 steps/sec	40461 MB
✔	✔	✔	0.84 steps/sec	27485 MB

Evaluation (FID, Inception Score, etc.)

We include a sample_ddp.py script which samples a large number of images from a DiT model in parallel. This script generates a folder of samples as well as a .npz file which can be directly used with ADM's TensorFlow evaluation suite to compute FID, Inception Score and other metrics. For example, to sample 50K images from our pre-trained DiT-XL/2 model over N GPUs, run:

torchrun --nnodes=1 --nproc_per_node=N sample_ddp.py --model DiT-XL/2 --num-fid-samples 50000

There are several additional options; see sample_ddp.py for details.