TorchMetrics

Spec sheet

Purpose

TorchMetrics is a collection of more than 100 PyTorch metrics implementations with an API for creating custom metrics.Read 11 Oct 2026 · lightning.ai

Free plan

YesRead 23 Sep 2026 · lightning.ai

Support

The project directs users to its Discord community for help and questions.Read 11 Oct 2026 · github.com

License

The project repository states that TorchMetrics is licensed under Apache 2.0.Read 11 Oct 2026 · github.com

Batch accumulation

Metrics can accumulate results across batches and compute an aggregate value.Read 11 Oct 2026 · lightning.ai

Bayesian analysis

NoRead 23 Sep 2026 · lightning.ai

Community support

The project repository directs users with questions to its Discord community.Read 11 Oct 2026 · github.com

Compatibility limit

The documentation says TorchMetrics supports the latest two major PyTorch Lightning versions and recommends keeping both frameworks up to date.Read 11 Oct 2026 · lightning.ai

Custom metrics

Users can create custom metrics by subclassing Metric and implementing update() and compute().Read 11 Oct 2026 · lightning.ai

Distributed training

Module metrics support automatic accumulation across batches and synchronization across devices for distributed training.Read 11 Oct 2026 · lightning.ai

Installation

The project documents installation with pip, conda, uv, or directly from its source repository.Read 11 Oct 2026 · github.com

Interfaces

TorchMetrics provides module-based metrics and functional metrics that take tensors as input and return metric values as tensors.Read 11 Oct 2026 · github.com

Lightning integration

When used with PyTorch Lightning, metrics can be logged with self.log and are automatically placed on the correct device when properly defined inside a LightningModule.Read 11 Oct 2026 · lightning.ai

Metric collections

MetricCollection wraps multiple metrics in one callable metric class.Read 11 Oct 2026 · lightning.ai

Metric coverage

The project lists more than 100 metrics across audio, classification, detection, information retrieval, image, multimodal, nominal, regression, segmentation, and text domains.Read 11 Oct 2026 · github.com

Optional dependencies

Specialized audio, image, and text metrics may require extra dependencies, which can be installed with package extras.Read 11 Oct 2026 · github.com

Plotting

Nearly all modular metrics support plotting through the plot method when the visualization dependencies are installed.Read 11 Oct 2026 · github.com

PyTorch integration

TorchMetrics can be used directly with PyTorch models independently of PyTorch Lightning.Read 11 Oct 2026 · lightning.ai

PyTorch Lightning integration

When metrics are properly defined inside a LightningModule, Lightning can place them on the module’s device and log them with self.log or self.log_dict.Read 11 Oct 2026 · lightning.ai

Regression models

NoRead 23 Sep 2026 · lightning.ai

Scripting support

YesRead 23 Sep 2026 · lightning.ai

Standardized interface

Its standardized interface is intended to improve reproducibility and reduce boilerplate.Read 11 Oct 2026 · lightning.ai

Statistical tests

NoRead 23 Sep 2026 · lightning.ai

Supported metric areas

The documentation organizes metrics by areas including classification, regression, image, audio, and text.Read 11 Oct 2026 · lightning.ai

Plans and prices

Every TorchMetrics plan and price →

Where Specifiction places it

• Statistical Analysis Software#35 of 1068.1