Software-part code of N3H-Core

Current code base is tested under following environment:

1. Python 3.8.5
2. Pytorch 1.6.0
3. torchvision 0.7.0
4. numpy 1.19.2
5. progress 1.5
6. tensorboardX 2.1

Create a link of the imagenet in the current folder:
ln -s /path/to/imagenet/ data/

Prepare imagenet100 dataset for fast search process:
python lib/utils/make_data.py

Set target latency in the bash file:
--target_latency 35

Run the bash file:
bash run/run_linear_quantize_search.sh

or (save the log file):
nohup bash run/run_linear_quantize_search.sh > search_tar35.log 2>&1 &

Copy the best policy (quantization + workload split-ratio) and then paste in the finetune.py:
strategy = [[8, 8, 0.25], [3, 7, 0.5], [3, 7, 0.5], [3, 7, 0.5], [3, 7, 0.5], [2, 3, 0.75], [2, 3, 0.88], [4, 8, 0.13], [3, 3, 0.75], [2, 3, 0.75], [2, 3, 0.82], [3, 3, 0.81], [4, 7, 0.06], [3, 3, 0.81], [2, 5, 0.82], [2, 2, 0.85], [3, 2, 0.9], [4, 6, 0.05], [3, 3, 0.85], [2, 4, 0.82], [8, 8, 0.3]] # resnet18

Retrain the base model and save the best checkpoint:
bash run/run_linear_quantize_finetune.sh