README.md

NAS(Neural Architecture Search)

How to Run

1. train base model with imagenet100 dataset

   • Train Base Model (resnet18)

2. NAS(Neural Architecture Search) and export nas onnx

   python onnx_export_nas.py
   

   1. Convert Model
   2. NAS traing
   3. Subnet architecture search
   4. Fine-tuning

3. generate tensorrt model

   python onnx2trt.py
   

• fp16 nas
• Gpu Mem: M

Reference

• TensorRT-Model-Optimizer