dimitreOliveira/bert-as-a-service_TFX

Implements a complete MLOps pipeline using TFX components (ExampleGen, StatisticsGen, Transform, Trainer, Evaluator, Pusher) with support for multiple orchestrators—Google Colab, Kubeflow, and Vertex AI—enabling reproducible model governance from data validation through production deployment. Leverages transfer learning with pre-trained BERT from TensorFlow Hub or HuggingFace, fine-tuning only a custom classifier head on IMDB sentiment data, while integrating Keras Tuner for hyperparameter optimization and InfraValidator for pre-production model validation. Supports both REST and gRPC serving interfaces for consuming the deployed model.

No commits in the last 6 months.