Torchmeta

Overview

Torchmeta is a community-maintained Python library that provides extensions and data-loaders for few-shot learning and meta-learning in PyTorch. It simplifies the process of creating episodic tasks and standardizes benchmarks for meta-learning research.

Best for

Best for
Researchers and developers building few-shot or meta-learning models in PyTorch

Use cases

• Implementing few-shot classification with episodic task sampling
• Reproducing meta-learning benchmarks like Mini-ImageNet or Omniglot
• Building custom meta-learning algorithms with modular data-loaders

Notes

Torchmeta is a community-maintained Python library that provides extensions and data-loaders for few-shot learning and meta-learning in PyTorch. It simplifies the process of creating episodic tasks and standardizes benchmarks for meta-learning research.

2,058 stars on GitHub. Last updated 2023-07-17. Licensed MIT.

Use cases

• Implementing few-shot classification with episodic task sampling
• Reproducing meta-learning benchmarks like Mini-ImageNet or Omniglot
• Building custom meta-learning algorithms with modular data-loaders

Pros

• Streamlines data-loading for few-shot learning with built-in task samplers
• Integrates directly with PyTorch, requiring minimal code changes
• Includes common benchmark datasets for reproducible research

Cons

• Limited to few-shot and meta-learning scenarios, not general-purpose
• Community-maintained with no official vendor support
• May lag behind PyTorch updates or lack newer dataset support

Indexed from awesome-llmops and enriched against its public facts.