composer.models#

Modules

`composer.models.base`	composer.models.base
`composer.models.bert`	composer.models.bert
`composer.models.classify_mnist`	composer.models.classify_mnist
`composer.models.deeplabv3`	composer.models.deeplabv3
`composer.models.efficientnetb0`	composer.models.efficientnetb0
`composer.models.efficientnets`	composer.models.efficientnets
`composer.models.gpt2`	composer.models.gpt2
`composer.models.initializers`	composer.models.initializers
`composer.models.loss`	composer.models.loss
`composer.models.model_hparams`	composer.models.model_hparams
`composer.models.nlp_metrics`	composer.models.nlp_metrics
`composer.models.resnet`	composer.models.resnet
`composer.models.resnet20_cifar10`	composer.models.resnet20_cifar10
`composer.models.resnet56_cifar10`	composer.models.resnet56_cifar10
`composer.models.resnet9_cifar10`	composer.models.resnet9_cifar10
`composer.models.resnets`	composer.models.resnets
`composer.models.timm`	composer.models.timm
`composer.models.transformer_hparams`	composer.models.transformer_hparams
`composer.models.transformer_shared`	composer.models.transformer_shared
`composer.models.unet`	composer.models.unet

composer.models

Classes

`BERTModel`	Implements a BERT wrapper around a ComposerTransformer.
`CIFAR10_ResNet20`	A ResNet-20 model extending `ComposerClassifier`.
`CIFAR10_ResNet56`	A ResNet-56 model extending `ComposerClassifier`.
`CIFAR10_ResNet9`	A ResNet-9 model extending `ComposerClassifier`.
`ComposerClassifier`	Implements the base logic that all classifiers can build on top of.
`ComposerDeepLabV3`	DeepLabV3 model extending the `ComposerClassifier`.
`ComposerModel`	The minimal interface needed to use a model with `composer.trainer.Trainer`.
`ComposerResNet`	ResNet model family extending `ComposerClassifier`.
`ComposerTransformer`	Implements the base logic that all Transformers can build on top of.
`EfficientNetB0`	An EfficientNet-b0 model extending `ComposerClassifier`.
`GPT2Model`	Implements a GPT-2 wrapper around a `ComposerTransformer`.
`Initializer`	An enumeration.
`MNIST_Classifier`	A simple convolutional neural network extending `ComposerClassifier`.
`Timm`	A wrapper around `timm.create_model()` used to create a `ComposerClassifier` from a timm model.
`UNet`	A U-Net model extending `ComposerClassifier`.

Hparams

These classes are used with yahp for YAML-based configuration.

`BERTForClassificationHparams`	composer.models.bert.bert_hparams.BERTForClassificationHparams
`BERTHparams`	composer.models.bert.bert_hparams.BERTHparams
`CIFARResNet20Hparams`	composer.models.resnet20_cifar10.resnet20_cifar10_hparams.CIFARResNet20Hparams
`CIFARResNet9Hparams`	composer.models.resnet9_cifar10.resnet9_cifar10_hparams.CIFARResNet9Hparams
`CIFARResNetHparams`	composer.models.resnet56_cifar10.resnet56_cifar10_hparams.CIFARResNetHparams
`DeepLabV3Hparams`	This class specifies arguments for a DeepLabV3 model and can instantiate a DeepLabV3 model.
`EfficientNetB0Hparams`	composer.models.efficientnetb0.efficientnetb0_hparams.EfficientNetB0Hparams
`GPT2Hparams`	Overrides TransformerHparams to create GPT-2 specific models and configs.
`MnistClassifierHparams`	composer.models.classify_mnist.mnist_hparams.MnistClassifierHparams
`ModelHparams`	Model Hparams.
`ResNetHparams`	composer.models.resnet.resnet_hparams.ResNetHparams
`TimmHparams`	composer.models.timm.timm_hparams.TimmHparams
`UnetHparams`	composer.models.unet.unet_hparams.UnetHparams