Docker Setup

The Docker image provides a reproducible CPU development environment for tests, benchmark scripts, and public-data validation runs. It is intended for local reproduction and contributor onboarding, not for live trading.

Build

docker build -t ml-quant-trading .

Run The Test Suite

docker run --rm ml-quant-trading make test

Run Lint

docker run --rm ml-quant-trading make lint

Run The Synthetic Pipeline

docker run --rm ml-quant-trading make paper CONFIG=configs/small.yaml

Run Benchmarks

docker run --rm ml-quant-trading make benchmark

The default image uses CPU PyTorch from the Python package resolver. For GPU benchmarks, use a host environment with the correct NVIDIA driver, CUDA runtime, and PyTorch build, or extend this Dockerfile from an NVIDIA CUDA base image.

Run Public-Data Validation

docker run --rm ml-quant-trading \
  python scripts/public_data_validation.py \
    --source synthetic \
    --models equal_weight,momentum_20,alpha101_mean

For yfinance runs, the container needs network access:

docker run --rm ml-quant-trading \
  python scripts/public_data_validation.py \
    --source yfinance \
    --preset us-large-100 \
    --max-tickers 100

The generated reports stay inside the container unless you mount a host directory:

mkdir -p artifacts/public_data_validation
docker run --rm \
  -v "$PWD/artifacts:/workspace/ml-quant-trading/artifacts" \
  ml-quant-trading \
  python scripts/public_data_validation.py --source synthetic

Dev Container

The VS Code / Codespaces Dev Container reuses the root Dockerfile and then reinstalls the mounted workspace in editable mode. This keeps local Docker and Dev Container setup aligned.