nvitop/examples/ml-framework-callbacks/README.md

ML-Framework Callbacks

Self-contained GpuStatsLogger callbacks for Keras and Lightning, plus a tiny TensorBoard helper. Each file has zero dependencies on the others — copy a single .py into your project and adapt the imports.

Files and Dependencies

File	Install command	Purpose
keras.py	pip install -r requirements-keras.txt	Keras Callback logging GPU stats per training batch.
lightning.py	pip install -r requirements-lightning.txt	Lightning Callback (lightning.pytorch.callbacks.Callback) for GPU stats.
tensorboard.py	pip install -r requirements-tensorboard.txt	add_scalar_dict — batched SummaryWriter.add_scalar for a flat metric dict.

Each requirements-*.txt pins only what that single file needs, so you do not pull in TensorFlow when you only want the Lightning callback.

See ../README.md for the full example index.