deepbrain-dbme

DeepBrain DBME (Dynamic Brain Memory Expansion) - Research Repository.

Quick Start (One-Click Reproduction)

1. Environment Setup

Prerequisites:

• Python 3.10+
• CUDA-capable GPU (Recommended: A100 or similar for full reproduction, but code is adaptable)
• Conda (Miniconda/Anaconda)

# Clone the repository
git clone https://github.com/yourusername/deepbrain-dbme.git
cd deepbrain-dbme

# Create environment (approx 5-10 mins)
conda env create -f environment.yml
conda activate dbme

# Verify CUDA/PyTorch
python -c "import torch; print(f'CUDA available: {torch.cuda.is_available()}')"

2. Experiment Tracking

We use Weights & Biases (WandB).

1. Log in to WandB: wandb login
2. Experiments will be logged to project dbme-experiments.

3. Running an Experiment

To run a base experiment using the provided seed control and config:

# This uses the config at configs/base_config.yaml
# Ensure you are in the root directory
python -m src.main --config configs/base_config.yaml

(Note: src.main is a placeholder for your entry point. Adjust according to your specific script implementations later).

Directory Structure

• configs/: YAML configuration files for experiments.
• src/: Source code.
• utils/: Utility scripts (seeding, logging, etc.).
• environment.yml: Conda environment spec.
• Dockerfile: Container definition.

Development

Branching strategy:

• main: Protected, stable.
• dev: Integration.
• feat/*: Feature branches.

License

Apache-2.0