The GAN is dead; long live the GAN! A Modern GAN Baseline
Authors: Nick Huang, Aaron Gokaslan, Volodymyr Kuleshov, James Tompkin
NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Despite being simple, our approach surpasses Style GAN2 on FFHQ, Image Net, CIFAR, and Stacked MNIST datasets, and compares favorably against state-of-the-art GANs and diffusion models. |
| Researcher Affiliation | Academia | Yiwen Huang Brown University Aaron Gokaslan Cornell University Volodymyr Kuleshov Cornell University James Tompkin Brown University |
| Pseudocode | No | The paper describes algorithms and architectures in prose and mathematical equations but does not include any clearly labeled 'Pseudocode' or 'Algorithm' blocks. |
| Open Source Code | Yes | Code: https://www.github.com/brownvc/R3GAN |
| Open Datasets | Yes | Using Style GAN2 as an example, we present a roadmap of simplification and modernization that results in a new minimalist baseline R3GAN ( Re-GAN ). Despite being simple, our approach surpasses Style GAN2 on FFHQ, Image Net, CIFAR, and Stacked MNIST datasets, and compares favorably against state-of-the-art GANs and diffusion models. |
| Dataset Splits | No | The paper mentions training durations and evaluation metrics but does not explicitly state the use of a validation set, its size, or how it was split from the main dataset for model tuning or early stopping. |
| Hardware Specification | Yes | We train the Stacked MNIST and CIFAR-10 models on an 8 NVIDIA L40 node. Training took 7 hours for Stacked MNIST and 4 days for CIFAR-10. The FFHQ model was trained on an 8 NVIDIA A6000 f0r roughly 3 weeks. The Image Net model was trained on NVIDIA A100/H100 clusters and training took one day on 32 H100s (about 5000 H100 hours). |
| Software Dependencies | No | The paper mentions implementing models on top of the 'official Style GAN3 code base' but does not specify version numbers for programming languages (e.g., Python), deep learning frameworks (e.g., PyTorch, TensorFlow), or other key software libraries. |
| Experiment Setup | Yes | We inherit training hyperparameters (e.g., optimizer settings, batch size, EMA decay length) from Config A unless otherwise specified. We tune the training hyperparameters for our final model and show the converged result in Sec. 4. |