FBNetV3: Joint Architecture-Recipe Search using Predictor Pretraining

Abstract

Neural Architecture Search (NAS) yields state-of-the-art neural networks that outperform their best manually-designed counterparts. However, previous NAS methods search for architectures under one set of training hyper-parameters (i.e., a training recipe), overlooking superior architecture-recipe combinations. To address this, we present Neural Architecture-Recipe Search (NARS) to search both (a) architectures and (b) their corresponding training recipes, simultaneously. NARS utilizes an accuracy predictor that scores architecture and training recipes jointly, guiding both sample selection and ranking. Furthermore, to compensate for the enlarged search space, we leverage "free" architecture statistics (e.g., FLOP count) to pretrain the predictor, significantly improving its sample efficiency and prediction reliability. After training the predictor via constrained iterative optimization, we run fast evolutionary searches in just CPU minutes to generate architecture-recipe pairs for a variety of resource constraints, called FBNetV3. FBNetV3 makes up a family of state-of-the-art compact neural networks that outperform both automatically and manually-designed competitors. For example, FBNetV3 matches both EfficientNet and ResNeSt accuracy on ImageNet with up to 2.0x and 7.1x fewer FLOPs, respectively. Furthermore, FBNetV3 yields significant performance gains for downstream object detection tasks, improving mAP despite 18% fewer FLOPs and 34% fewer parameters than EfficientNet-based equivalents.

References in corpus (17)

• MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications
• Neural Architecture Search with Reinforcement Learning
• Large Scale GAN Training for High Fidelity Natural Image Synthesis
• Binarized Neural Networks
• Trained Ternary Quantization
• Tune: A Research Platform for Distributed Model Selection and Training
• Large-Scale Evolution of Image Classifiers
• SNAS: Stochastic Neural Architecture Search
• Single Path One-Shot Neural Architecture Search with Uniform Sampling
• Progressive Neural Architecture Search
• A Bayesian Data Augmentation Approach for Learning Deep Models
• Towards Automated Deep Learning: Efficient Joint Neural Architecture and Hyperparameter Search
• Dreaming to Distill: Data-free Knowledge Transfer via DeepInversion
• BigNAS: Scaling Up Neural Architecture Search with Big Single-Stage Models
• SqueezeSegV3: Spatially-Adaptive Convolution for Efficient Point-Cloud Segmentation
• Neural Predictor for Neural Architecture Search
• Using a thousand optimization tasks to learn hyperparameter search strategies

Cited by in corpus (13)

• AutoML: A Survey of the State-of-the-Art
• Visual Transformers: Token-based Image Representation and Processing for Computer Vision
• Early Convolutions Help Transformers See Better
• A Comprehensive Survey of Neural Architecture Search: Challenges and Solutions
• You Only Look at One Sequence: Rethinking Transformer in Vision through Object Detection
• Sample-Efficient Neural Architecture Search by Learning Action Space
• Weight-Sharing Neural Architecture Search: A Battle to Shrink the Optimization Gap
• Stronger NAS with Weaker Predictors
• AttentiveNAS: Improving Neural Architecture Search via Attentive Sampling
• Collaboration of Experts: Achieving 80% Top-1 Accuracy on ImageNet with 100M FLOPs
• Evolving Search Space for Neural Architecture Search
• AlphaNet: Improved Training of Supernets with Alpha-Divergence
• AutoLoss-Zero: Searching Loss Functions from Scratch for Generic Tasks