enas in pytorch

We have hosted the application enas in pytorch in order to run this application in our online workstations with Wine or directly.

Quick description about enas in pytorch:

ENAS in PyTorch is a PyTorch implementation of Efficient Neural Architecture Search (ENAS), a method that automates the design of neural network architectures through reinforcement learning and parameter sharing. The repository demonstrates how a controller network can explore a large search space and discover high-performing architectures while dramatically reducing the computational cost traditionally associated with neural architecture search. It is primarily intended as a research and educational codebase, helping practitioners understand how ENAS works in practice and how to reproduce results on benchmark datasets. The project includes training scripts, model definitions, and search procedures that show the full workflow from architecture sampling to evaluation. Because ENAS relies on shared weights among candidate models, the implementation emphasizes efficiency and experiment reproducibility.

Features:

• PyTorch implementation of Efficient Neural Architecture Search
• Controller-based architecture sampling workflow
• Parameter sharing for faster search
• Training scripts for benchmark datasets
• Research-oriented modular code structure
• Reproducible experiment configuration

Programming Language: Python.
Categories: