DynaPAT (ICCAD 2022) Implementation

This repository is the official implementation of DynaPAT paper. Details of the proposed methods are described in: Thai-Hoang Nguyen, Muhammad Imran, Joon-Sung Yang (2022). DynaPAT: A Dynamic Pattern-Aware Encoding Technique for Robust MLC PCM-Based Deep Neural Networks. ICCAD'22: Proceedings of the 41st IEEE/ACM International Conference on Computer-Aided Design.

Requirements

Install required packages:

$ pip install -r requirement.txt

GETTING STARTED

1. Clone the repo git clone https://github.com/thnguyen996/DynaPAT_ICCAD_22.git
2. Download pretrained weights and put it in checkpoint/ folder
3. Run the code python main.py --help to see all the options provided by the code

Baseline

Run the following code to evaluate resistance drift on a baseline model

$ python main.py --model [name of the model] --method baseline -gpu [GPU ID] --num_bits 8

Proposed method (DynaPAT):

Run state_encoding.py to generate state encoding for DynaPAT

$ python state_encoding.py --model [Name of the model] --method proposed_method 

Run main.py to evaluate DynaPAT for a given state encoding

$ python main.py --method proposed_method --model [Name of the model] --gpu [GPU ID]  --num_bits 8

Flipcy and Helmet

For Flipcy and Helmet, run the main.py file with addtional options --encode first, this will generate encodings for the 2 methods

$ python main.py --method ECP 

Research papers citing

If you use this code for your research paper, please use the following citation:

@inproceedings{nguyen2022dynapat,
 author = {Nguyen, Thai-Hoang and Imran, Muhammad and Yang, Joon-Sung},
 booktitle = {Proceedings of the 41st IEEE/ACM International Conference on Computer-Aided Design},
 pages = {1--9},
 title = {DynaPAT: A Dynamic Pattern-Aware Encoding Technique for Robust MLC PCM-Based Deep Neural Networks},
 year = {2022}
}