Machine learning metrics for distributed, scalable PyTorch applications.
What is Torchmetrics • Implementing a metric • Built-in metrics • Docs • Community • License
Installation
Simple installation from PyPI
pip install torchmetrics
Other installions
Install using conda
conda install torchmetrics
Pip from source
# with git
pip install git+https://github.com/PytorchLightning/[email protected]
Pip from archive
pip install https://github.com/PyTorchLightning/metrics/archive/master.zip
What is Torchmetrics
TorchMetrics is a collection of 25+ PyTorch metrics implementations and an easy-to-use API to create custom metrics. It offers:
- A standardized interface to increase reproducibility
- Reduces Boilerplate
- Automatic accumulation over batches
- Metrics optimized for distributed-training
- Automatic synchronization between multiple devices
You can use TorchMetrics with any PyTorch model or with PyTorch Lightning to enjoy additional features such as:
- Module metrics are automatically placed on the correct device.
- Native support for logging metrics in Lightning to reduce even more boilerplate.
Using TorchMetrics
Module metrics
The module-based metrics contain internal metric states (similar to the parameters of the PyTorch module) that automate accumulation and synchronization across devices!
- Automatic accumulation over multiple batches
- Automatic synchronization between multiple devices
- Metric arithmetic
This can be run on CPU, single GPU or multi-GPUs!
For the single GPU/CPU case:
import torch
# import our library
import torchmetrics
# initialize metric
metric = torchmetrics.Accuracy()
n_batches = 10
for i in range(n_batches):
# simulate a classification problem
preds = torch.randn(10, 5).softmax(dim=-1)
target = torch.randint(5, (10,))
# metric on current batch
acc = metric(preds, target)
print(f"Accuracy on batch {i}: {acc}")
# metric on all batches using custom accumulation
acc = metric.compute()
print(f"Accuracy on all data: {acc}")
Module metric usage remains the same when using multiple GPUs or multiple nodes.
Example using DDP
os.environ['MASTER_ADDR'] = 'localhost'
os.environ['MASTER_PORT'] = '12355'
# create default process group
dist.init_process_group("gloo", rank=rank, world_size=world_size)
# initialize model
metric = torchmetrics.Accuracy()
# define a model and append your metric to it
# this allows metric states to be placed on correct accelerators when
# .to(device) is called on the model
model = nn.Linear(10, 10)
model.metric = metric
model = model.to(rank)
# initialize DDP
model = DDP(model, device_ids=[rank])
n_epochs = 5
# this shows iteration over multiple training epochs
for n in range(n_epochs):
# this will be replaced by a DataLoader with a DistributedSampler
n_batches = 10
for i in range(n_batches):
# simulate a classification problem
preds = torch.randn(10, 5).softmax(dim=-1)
target = torch.randint(5, (10,))
# metric on current batch
acc = metric(preds, target)
if rank == 0: # print only for rank 0
print(f"Accuracy on batch {i}: {acc}")
# metric on all batches and all accelerators using custom accumulation
# accuracy is same across both accelerators
acc = metric.compute()
print(f"Accuracy on all data: {acc}, accelerator rank: {rank}")
# Reseting internal state such that metric ready for new data
metric.reset()
Implementing your own Module metric
Implementing your own metric is as easy as subclassing an torch.nn.Module. Simply, subclass torchmetrics.Metric and implement the following methods:
class MyAccuracy(Metric):
def __init__(self, dist_sync_on_step=False):
# call `self.add_state`for every internal state that is needed for the metrics computations
# dist_reduce_fx indicates the function that should be used to reduce
# state from multiple processes
super().__init__(dist_sync_on_step=dist_sync_on_step)
self.add_state("correct", default=torch.tensor(0), dist_reduce_fx="sum")
self.add_state("total", default=torch.tensor(0), dist_reduce_fx="sum")
def update(self, preds: torch.Tensor, target: torch.Tensor):
# update metric states
preds, target = self._input_format(preds, target)
assert preds.shape == target.shape
self.correct += torch.sum(preds == target)
self.total += target.numel()
def compute(self):
# compute final result
return self.correct.float() / self.total
Functional metrics
Similar to torch.nn, most metrics have both a module-based and a functional version. The functional versions are simple python functions that as input take torch.tensors and return the corresponding metric as a torch.tensor.
import torch
# import our library
import torchmetrics
# simulate a classification problem
preds = torch.randn(10, 5).softmax(dim=-1)
target = torch.randint(5, (10,))
acc = torchmetrics.functional.accuracy(preds, target)
Implemented metrics
- Accuracy
- AveragePrecision
- AUC
- AUROC
- F1
- Hamming Distance
- ROC
- ExplainedVariance
- MeanSquaredError
- R2Score
- bleu_score
- embedding_similarity
And many more!
Contribute!
The lightning + torchmetric team is hard at work adding even more metrics. But we're looking for incredible contributors like you to submit new metrics and improve existing ones!
Join our Slack to get help becoming a contributor!
Community
For help or questions, join our huge community on Slack!
Citations
We’re excited to continue the strong legacy of opensource software and have been inspired over the years by Caffee, Theano, Keras, PyTorch, torchbearer, ignite, sklearn and fast.ai. When/if a paper is written about this, we’ll be happy to cite these frameworks and the corresponding authors.
License
Please observe the Apache 2.0 license that is listed in this repository. In addition the Lightning framework is Patent Pending.
19 Dec 05, 2022
328 Dec 26, 2022
1 Dec 17, 2021
7 Feb 15, 2022
343 Dec 29, 2022
102 Nov 03, 2022
30 Dec 14, 2022
25 Dec 11, 2022
324 Dec 27, 2022
196 Dec 13, 2022
8 Nov 26, 2022
133 Dec 22, 2022
100 Dec 29, 2022
4.6k Dec 30, 2022
1 Jan 10, 2022
3 May 29, 2022
227 Jan 04, 2023
70 Dec 07, 2022
281 Dec 09, 2022
51 Dec 20, 2022