tensorflow
TensorFlow machine learning platform with Keras. Use for ML models.
Install
npx skillscat add g1joshi/agent-skills/tensorflow Install via the SkillsCat registry.
TensorFlow is a machine learning platform that provides tools such as Keras for building, training, and deploying models, and TFLite for running them on mobile and edge devices. It is appropriate for maintaining legacy enterprise pipelines, supporting TFX MLOps workflows, or deploying models to Android or iOS platforms.
TensorFlow
TensorFlow is Google's mature ML framework. In 2025, it is largely in maintenance mode compared to JAX/PyTorch, but remains specific for TFLite and legacy production.
When to Use
- Mobile (TFLite): Deploying to Android/iOS (though
LiteRTis separating). - Legacy Enterprise: Many Fortune 500s still run TFv2 pipelines.
- TFX: Complex MLOps pipelines.
Core Concepts
SavedModel
The universal serialization format.
TFLite
Optimized runtime for edge devices.
XLA
Accelerated Linear Algebra compiler (now shared with JAX).
Best Practices (2025)
Do:
- Use Keras 3: If you must use generic TensorFlow, use Keras 3 as the interface.
- Migrate to JAX: For new Google-stack research projects.
Don't:
- Don't start new research in TF: Use PyTorch or JAX.
References
Categories
Install
npx skillscat add g1joshi/agent-skills/tensorflow Install via the SkillsCat registry.