SCPC 2025 AI

• 2025 Samsung Collegiate Programming Challenge : AI
• Submission Code
• Huigyoeng Son

ranking

• final private ranking 18/1445
• leaderboard : https://dacon.io/competitions/official/236500/leaderboard

environment

System

• OS: Linux
• GPU: 1x NVIDIA A100 SXM4 (40GB VRAM)
• CUDA: 12.8

Python Environment

• Python: 3.12.11
• All dependencies are managed via pip.

Key Libraries

Package	Version
torch	2.7.1+cu128
torchaudio	2.7.1+cu128
torchvision	0.22.1+cu128
transformers	4.54.0
accelerate	1.9.0
peft	0.16.0
bitsandbytes	0.46.1
trl	0.19.1
datasets	4.0.0
transformers-stream-generator	0.0.5
safetensors	0.5.3
einops	0.8.1
tiktoken	0.9.0
diffusers	0.20.0
numpy	2.1.2
pandas	2.3.1
pillow	11.0.0
tqdm	4.67.1

Jupyter (Optional)

Package	Version
ipykernel	6.30.0
jupyter_client	8.6.3
jupyter_core	5.8.1

Installation

• Install dependencies with requirements.txt

  pip install -r requirements.txt --extra-index-url https://download.pytorch.org/whl/cu128
  

• Install dependencies individually

  
  # Install PyTorch (CUDA 12.1)
  pip install torch==2.7.1+cu128 torchvision==0.22.1+cu128 torchaudio==2.7.1+cu128 \
    --index-url https://download.pytorch.org/whl/cu128
  
  # Core libraries
  pip install transformers==4.54.0 accelerate==1.9.0 peft==0.16.0 bitsandbytes==0.46.1
  pip install trl==0.19.1 datasets==4.0.0 tqdm==4.67.1
  pip install transformers-stream-generator==0.0.5 einops==0.8.1 tiktoken==0.9.0
  pip install diffusers==0.20.0 safetensors==0.5.3
  
  # Utilities
  pip install numpy==2.1.2 pandas==2.3.1 pillow==11.0.0
  
  # Jupyter (optional)
  pip install ipykernel==6.30.0 jupyter_client==8.6.3 jupyter_core==5.8.1
  
  

Run

Generate dataset

1. run generate_scene_prompt.py
2. run generate_scene_image.py
3. run generate_question_answer.py

Train & Inference

1. run train.ipynb
2. run inference.ipynb