Flow++: Improving Flow-Based Generative Models with Variational Dequantization and Architecture Design
Authors: Jonathan Ho, Xi Chen, Aravind Srinivas, Yan Duan, Pieter Abbeel
ICML 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Here, we show that Flow++ achieves state-of-the-art density modeling performance among non-autoregressive models on CIFAR10 and 32x32 and 64x64 Image Net. We also present ablation experiments that quantify the improvements proposed in section 3, and we present example generative samples from Flow++ and compare them against samples from autoregressive models. |
| Researcher Affiliation | Collaboration | 1UC Berkeley, Department of Electrical Engineering and Computer Science 2covariant.ai. |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Our implementation is available at: https://github.com/ aravindsrinivas/flowpp. |
| Open Datasets | Yes | Here, we show that Flow++ achieves state-of-the-art density modeling performance among non-autoregressive models on CIFAR10 and 32x32 and 64x64 Image Net. We also present ablation experiments that quantify the improvements proposed in section 3, and we present example generative samples from Flow++ and compare them against samples from autoregressive models. We present the samples from our trained density models of Flow++ on CIFAR10, 32x32 Image Net, 64x64 Image Net, and 5-bit Celeb A in figs. 2 to 5. |
| Dataset Splits | No | The paper mentions 'ablation training (light) and validation (dark) curves' but does not provide specific data split information (exact percentages, sample counts, or citations to predefined splits) for the validation set. |
| Hardware Specification | Yes | Our listed Image Net 32x32 and 64x64 results are evaluated on a NVIDIA DGX-1; they are worse by 0.01 bits/dim when evaluated on a NVIDIA Titan X GPU. sampling is fast: our CIFAR10 model takes approximately 0.32 seconds to generate a batch of 8 samples in parallel on one NVIDIA 1080 Ti GPU |
| Software Dependencies | No | The paper does not provide specific ancillary software details, such as library or solver names with version numbers. |
| Experiment Setup | Yes | Our CIFAR10 model used 4 coupling layers with checkerboard splits at 32x32 resolution, 2 coupling layers with channel splits at 16x16 resolution, and 3 coupling layers with checkerboard splits at 16x16 resolution; each coupling layer used 10 convolutionattention blocks, all with 96 filters. (We always use 4 heads in our experiments, since we found it to be effective early on in our experimentation process.) |