guided-diffusion

This is the codebase for Diffusion Models Beat GANS on Image Synthesis.

This repository is based on openai/improved-diffusion, with modifications for classifier conditioning and architecture improvements.

Usage

Training diffusion models is described in the parent repository. Training a classifier is similar. We assume you have put training hyperparameters into a TRAIN_FLAGS variable, and classifier hyperparameters into a CLASSIFIER_FLAGS variable. Then you can run:

mpiexec -n N python scripts/classifier_train.py --data_dir path/to/imagenet $TRAIN_FLAGS $CLASSIFIER_FLAGS

Make sure to divide the batch size in TRAIN_FLAGS by the number of MPI processes you are using.

Here are flags for training the 128x128 classifier. You can modify these for training classifiers at other resolutions:

TRAIN_FLAGS="--iterations 300000 --anneal_lr True --batch_size 256 --lr 3e-4 --save_interval 10000 --weight_decay 0.05"
CLASSIFIER_FLAGS="--image_size 128 --classifier_attention_resolutions 32,16,8 --classifier_depth 2 --classifier_width 128 --classifier_pool attention --classifier_resblock_updown True --classifier_use_scale_shift_norm True"

For sampling from a 128x128 classifier-guided model, 25 step DDIM:

MODEL_FLAGS="--attention_resolutions 32,16,8 --class_cond True --image_size 128 --learn_sigma True --num_channels 256 --num_heads 4 --num_res_blocks 2 --resblock_updown True --use_fp16 True --use_scale_shift_norm True"
CLASSIFIER_FLAGS="--image_size 128 --classifier_attention_resolutions 32,16,8 --classifier_depth 2 --classifier_width 128 --classifier_pool attention --classifier_resblock_updown True --classifier_use_scale_shift_norm True --classifier_scale 1.0 --classifier_use_fp16 True"
SAMPLE_FLAGS="--batch_size 4 --num_samples 50000 --timestep_respacing ddim25 --use_ddim True"
mpiexec -n N python scripts/classifier_sample.py \
    --model_path /path/to/model.pt \
    --classifier_path path/to/classifier.pt \
    $MODEL_FLAGS $CLASSIFIER_FLAGS $SAMPLE_FLAGS

To sample for 250 timesteps without DDIM, replace --timestep_respacing ddim25 to --timestep_respacing 250, and replace --use_ddim True with --use_ddim False.

This is the codebase for Diffusion Models Beat GANS on Image Synthesis.

Related tags

Overview

guided-diffusion

Usage

Owner

OpenAI

Gender Classification Machine Learning Model using Sk-learn in Python with 97%+ accuracy and deployment

Official implementation for "Low-light Image Enhancement via Breaking Down the Darkness"

Pytorch Implementation of Zero-Shot Image-to-Text Generation for Visual-Semantic Arithmetic

Meta graph convolutional neural network-assisted resilient swarm communications

Understanding and Overcoming the Challenges of Efficient Transformer Quantization

Create Data & AI apps in 20 lines of code with Shimoku

An Active Automata Learning Library Written in Python

DP-CL(Continual Learning with Differential Privacy)

Research on Tabular Deep Learning (Python package & papers)

DeepMReye: magnetic resonance-based eye tracking using deep neural networks

[AAAI-2021] Visual Boundary Knowledge Translation for Foreground Segmentation

Deep learning library for solving differential equations and more

Python version of the amazing Reaction Mechanism Generator (RMG).

Differentiable scientific computing library

Code implementation of "Sparsity Probe: Analysis tool for Deep Learning Models"

Systemic Evolutionary Chemical Space Exploration for Drug Discovery

Meli Data Challenge 2021 - First Place Solution

Implementation of SE3-Transformers for Equivariant Self-Attention, in Pytorch.

TensorFlow 2 implementation of the Yahoo Open-NSFW model

DeepFaceLive - Live Deep Fake in python, Real-time face swap for PC streaming or video calls