TorchMetrics
Compute machine learning metrics over PyTorch tensors, with state that can be accumulated across batches.
Published by Lightning AI
How to use it
Choose a metric that matches the task and prediction representation, such as multiclass classification or regression. Use the same preprocessing, threshold, and averaging options for models you compare.
Accumulate metric state over batches and compute the final result after the evaluation pass. Check distributed reduction behavior when predictions are split across devices.
- Compatibility
- Python, PyTorch
- Evaluation areas
- Metric library, Prediction scoring, Batch evaluation
- Access
- Local Python library for PyTorch workloads.