MrGemy95/Tensorflow-Project-Template

Provides modular base classes (BaseModel, BaseTrain) that enforce separation of concerns—models implement only `build_model()` and `init_saver()`, while trainers override `train_step()` and `train_epoch()`—eliminating boilerplate for checkpointing, session management, and metric logging. Includes built-in TensorBoard logging and optional Comet.ml integration for tracking hyperparameters and real-time metrics across experiments.

3,628 stars. No commits in the last 6 months.