Advanced AI Prompt for Python Data Science and ML Dockerfiles

Act as a Principal Machine Learning Operations (MLOps) Engineer. Construct a high-performance, production-ready Dockerfile for a heavy [ML_FRAMEWORK] application. The base image must utilize [CUDA_VERSION] for GPU acceleration. Implement [MAMBA_OR_CONDA] for strict environment management during the build stage. Optimize the final image size by aggressively cleaning up the package cache and removing source files after compilation. Configure a persistent [DATA_VOLUME] for training datasets. Crucially, ensure the entrypoint script correctly handles SIGTERM and SIGINT signal forwarding to guarantee graceful shutdowns and checkpoint saving during long-running distributed training jobs.