tensorflow/hub

Health v2 · maintenance only

Adoption not part of the score

3523 stars · 1641 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

TensorFlow Hub is a Python library for reusing parts of trained TensorFlow models (SavedModels) for transfer learning, wrapping them as Keras layers or loading them directly. The model repository itself has moved to Kaggle Models, and the library remains supported for downloading and loading those models.

Use cases

• fine-tune pretrained models like BERT for text classification
• extract embeddings from text or images with a few lines of code
• reuse a Faster R-CNN model for object detection
• do transfer learning on image classifiers
• load SavedModels from tfhub.dev or Kaggle Models into Keras
• run image style transfer with a pretrained model

When to choose

• you are already in the TensorFlow/Keras ecosystem and want pretrained models for transfer learning
• you need to load legacy tfhub.dev or Kaggle Models assets without rewriting code
• you want quick embeddings or fine-tuning with minimal code

When to avoid

• you are starting a new project and can use Kaggle Models or Hugging Face directly
• you work in PyTorch or JAX rather than TensorFlow
• you need unmigrated tfhub.dev assets, which were deleted in March 2024

Facets

library · maturity maintenance

machine-learning sdk machine-learning deep-learning large-language-models computer-vision python cross-platform transfer-learning tensorflow pretrained-models embeddings model-repository keras fine-tuning natural-language-processing

Member repositories

For agents

markdown · JSON · MCP: product_card(name="tensorflow/hub")