atpgllm

Installation

Requirements

• Python 3.9 or higher
• pip package manager
• CUDA-capable GPU (recommended)

Quick Install

1. Clone the repository

git clone https://github.com/christos-vasileiou/atpgllm.git
cd atpgllm

2. Create and activate a virtual environment

python -m venv myenv
source myenv/bin/activate  # On Linux/Mac

3. Install the package and dependencies

pip install -r requirements.txt --use-pep517 --no-cache-dir

Troubleshooting

If you encounter installation issues: 1. Make sure you have CUDA installed for GPU support 2. Try removing any existing installations:

pip uninstall atpgllm -y
pip cache purge

3. Then reinstall with:

pip install -r requirements.txt --use-pep517 --no-cache-dir

Usage

Training

For multi/single-gpu training use:

torchrun --proc_per_node=<NODES> script_name.py
# Example: torchrun --proc_per_node=4 sequence_multilabel_classification.py

For gpu/cpu training use:

python script_name.py
# Example: python sequence_multilabel_classification.py

Project Goals

Large Language Model for Design Testing and Fault-Modeling:

1. Stuck-at: A specific net is "stuck" at a constant logic value (either 0 or 1)
2. Transition: Related to signal transitions between different logic levels (e.g., 0 to 1 or 1 to 0)
3. Coupling: ...
4. ...

Ultimate Goal: To cover as many Fault Models as possible.

What may go wrong?

• Physical Defects:
  • Shorts between two points (bridges)
  • Open in a line
  • Improper doping
  • Masking error
  • Particles on surface
  • Corrosion
• Main Focus: Safety and Reliability