TensorFlow
deep-learning framework
Google's ML framework for production
Supported languages
Concepts
tensorskeraseager-executionsaved-model
Deployment Options:
tf-serving tf-lite tflite-micro
Pros and Cons
Ventajas
- + Mature production ecosystem
- + TensorFlow Lite for mobile
- + TensorFlow Serving for deployment
- + Excellent TensorBoard
- + Native TPU support
- + TensorFlow Serving for deploy
- + Excellent for production and deployment
- + TensorFlow Serving for inference
- + TensorFlow Lite for mobile/edge
- + Keras integrated as high-level API
- + Large ecosystem and documentation
Desventajas
- - API more complex than PyTorch
- - TF 1 vs TF 2 confusing
- - Less popular in research
- - More complex API than PyTorch
- - Steeper learning curve
- - Steep learning curve
- - Historically inconsistent API
- - Debugging harder than PyTorch
- - Verbose for prototyping
Casos de Uso
- ML in production
- ML on mobile (TF Lite)
- Edge deployment
- Recommendation systems
- Production models at scale
- Mobile applications with TF Lite
- Enterprise computer vision
- NLP and transformers
- Edge AI and IoT