torchmeta

A collection of extensions and data-loaders for few-shot learning & meta-learning in PyTorch. The package contains popular meta-learning benchmarks, fully compatible with both torchvision and PyTorch's DataLoader.

Example

This minimal example below shows how to create a dataloader for the 5-shot 5-way Omniglot dataset with torchmeta. The dataloader loads a batch of randomly generated tasks. For more examples, check the examples folder.

from torchmeta.datasets import Omniglot
from torchmeta.transforms import Categorical, ClassSplitter
from torchvision.transforms import Resize, ToTensor, Compose

from torchmeta.utils.data import BatchMetaDataLoader

dataset = Omniglot('data', num_classes_per_task=5,
                   transform=Compose([Resize(28), ToTensor()]),
                   target_transform=Categorical(num_classes=5),
                   meta_train=True, download=True)
dataset = ClassSplitter(dataset, num_train_per_class=5, num_test_per_class=15)

dataloader = BatchMetaDataLoader(dataset, batch_size=16, num_workers=4)

for batch in dataloader:
    train_inputs, train_targets = batch['train']
    print('Train inputs shape: {0}'.format(train_inputs.shape))
    print('Train targets shape: {0}'.format(train_targets.shape))
    # Train inputs shape: torch.Size([16, 25, 1, 28, 28])
    # Train targets shape: torch.Size([16, 25])

    test_inputs, test_targets = batch['test']
    print('Test inputs shape: {0}'.format(test_inputs.shape))
    print('Test targets shape: {0}'.format(test_targets.shape))
    # Test inputs shape: torch.Size([16, 75, 1, 28, 28])
    # Test targets shape: torch.Size([16, 75])

Metadata

Release files for torchmeta 1.1.0rc2

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for torchmeta 1.1.0rc2
File	Size	Uploaded
torchmeta-1.1.0rc2.tar.gz	105.8 kB	Sep 16, 2019	Details

Built distribution (wheel)

Table of built distributions (wheels) for torchmeta 1.1.0rc2
File	Interpreter	ABI	Platform	Reset
torchmeta-1.1.0rc2-py3-none-any.whl 129.7 kB Sep 16, 2019	Python 3	none	any	Details

Total release size: 235.5 kB

Release files / torchmeta-1.1.0rc2.tar.gz

Download URL	torchmeta-1.1.0rc2.tar.gz
Size	105.8 kB
Tags	Source
SHA-256 checksum How to use checksums	`f53dba3b453e1d238be21539208585540b5046d0c38964f6907a9cb622df18b8`
BLAKE2b-256 checksum How to use checksums	`5173d250a3c5a5d90aed9bee806e5afbe0f4334d89e8ee91c3af6bd8e75a947c`
Upload date	Sep 16, 2019
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/1.13.0 pkginfo/1.5.0.1 requests/2.22.0 setuptools/41.1.0 requests-toolbelt/0.9.1 tqdm/4.33.0 CPython/3.6.2`

Release files / torchmeta-1.1.0rc2-py3-none-any.whl

Download URL	torchmeta-1.1.0rc2-py3-none-any.whl
Size	129.7 kB
Tags	Python 3
SHA-256 checksum How to use checksums	`ae54eb36c144c39e39253254d3a1a82662e063cc89d3ed02c0e1161fefaf9f21`
BLAKE2b-256 checksum How to use checksums	`0bd9c07c041950869b933372beaf67a701c6f7b2d68091c649e65c365518c6ba`
Upload date	Sep 16, 2019
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`twine/1.13.0 pkginfo/1.5.0.1 requests/2.22.0 setuptools/41.1.0 requests-toolbelt/0.9.1 tqdm/4.33.0 CPython/3.6.2`