PyTorch-Lightning-Bolts documentation¶

Introduction Guide¶

Welcome to PyTorch Lightning Bolts!

Bolts is a Deep learning research and production toolbox of:

SOTA pretrained models.
Model components.
Callbacks.
Losses.
Datasets.

The Main goal of Bolts is to enable trying new ideas as fast as possible!

All models are tested (daily), benchmarked, documented and work on CPUs, TPUs, GPUs and 16-bit precision.

some examples!

from pl_bolts.models import VAE, GPT2, ImageGPT, PixelCNN
from pl_bolts.models.self_supervised import AMDIM, CPCV2, SimCLR, MocoV2
from pl_bolts.models import LinearRegression, LogisticRegression
from pl_bolts.models.gans import GAN
from pl_bolts.callbacks import PrintTableMetricsCallback
from pl_bolts.datamodules import FashionMNISTDataModule, CIFAR10DataModule, ImagenetDataModule

Bolts are built for rapid idea iteration - subclass, override and train!

from pl_bolts.models import ImageGPT
from pl_bolts.self_supervised import SimCLR

class VideoGPT(ImageGPT):

    def training_step(self, batch, batch_idx):
        x, y = batch
        x = _shape_input(x)

        logits = self.gpt(x)
        simclr_features = self.simclr(x)

        # -----------------
        # do something new with GPT logits + simclr_features
        # -----------------

        loss = self.criterion(logits.view(-1, logits.size(-1)), x.view(-1).long())

        logs = {"loss": loss}
        return {"loss": loss, "log": logs}

Mix and match data, modules and components as you please!

model = GAN(datamodule=ImagenetDataModule(PATH))
model = GAN(datamodule=FashionMNISTDataModule(PATH))
model = ImageGPT(datamodule=FashionMNISTDataModule(PATH))

And train on any hardware accelerator

import pytorch_lightning as pl

model = ImageGPT(datamodule=FashionMNISTDataModule(PATH))

# cpus
pl.Trainer.fit(model)

# gpus
pl.Trainer(gpus=8).fit(model)

# tpus
pl.Trainer(tpu_cores=8).fit(model)

Or pass in any dataset of your choice

model = ImageGPT()
Trainer().fit(
    model,
    train_dataloader=DataLoader(...),
    val_dataloader=DataLoader(...)
)

Community Built¶

Then lightning community builds bolts and contributes them to Bolts. The lightning team guarantees that contributions are:

Rigorously tested (CPUs, GPUs, TPUs).
Rigorously documented.
Standardized via PyTorch Lightning.
Optimized for speed.
Checked for correctness.

How to contribute¶

We accept contributions directly to Bolts or via your own repository.

Note

We encourage you to have your own repository so we can link to it via our docs!

To contribute:

Submit a pull request to Bolts (we will help you finish it!).
We’ll help you add tests.
We’ll help you refactor models to work on (GPU, TPU, CPU)..
We’ll help you remove bottlenecks in your model.
We’ll help you write up documentation.
We’ll help you pretrain expensive models and host weights for you.
We’ll create proper attribution for you and link to your repo.
Once all of this is ready, we will merge into bolts.

After your model or other contribution is in bolts, our team will make sure it maintains compatibility with the other components of the library!

Contribution ideas¶

Don’t have something to contribute? Ping us on Slack or look at our Github issues!

We’ll help and guide you through the implementation / conversion

When to use Bolts¶

For pretrained models¶

Most bolts have pretrained weights trained on various datasets or algorithms. This is useful when you don’t have enough data, time or money to do your own training.

For example, you could use a pretrained VAE to generate features for an image dataset.

from pl_bolts.models.autoencoders import VAE
from pl_bolts.models.self_supervised import CPCV2

model1 = VAE(pretrained='imagenet2012')
encoder = model1.encoder
encoder.freeze()

# bolts are pretrained on different datasets
model2 = CPCV2(encoder='resnet18', pretrained='imagenet128').freeze()
model3 = CPCV2(encoder='resnet18', pretrained='stl10').freeze()

for (x, y) in own_data
    features = encoder(x)
    feat2 = model2(x)
    feat3 = model3(x)

# which is better?

To finetune on your data¶

If you have your own data, finetuning can often increase the performance. Since this is pure PyTorch you can use any finetuning protocol you prefer.

Example 1: Unfrozen finetune

# unfrozen finetune
model = CPCV2(encoder='resnet18', pretrained='imagenet128')
resnet18 = model.encoder
# don't call .freeze()

classifier = LogisticRegression()

for (x, y) in own_data:
    feats = resnet18(x)
    y_hat = classifier(feats)

Example 2: Freeze then unfreeze

# FREEZE!
model = CPCV2(encoder='resnet18', pretrained='imagenet128')
resnet18 = model.encoder
resnet18.freeze()

classifier = LogisticRegression()

for epoch in epochs:
    for (x, y) in own_data:
        feats = resnet18(x)
        y_hat = classifier(feats)
        loss = cross_entropy_with_logits(y_hat, y)

    # UNFREEZE after 10 epochs
    if epoch == 10:
        resnet18.unfreeze()

For research¶

Here is where bolts is very different than other libraries with models. It’s not just designed for production, but each module is written to be easily extended for research.

from pl_bolts.models import ImageGPT
from pl_bolts.self_supervised import SimCLR

class VideoGPT(ImageGPT):

    def training_step(self, batch, batch_idx):
        x, y = batch
        x = _shape_input(x)

        logits = self.gpt(x)
        simclr_features = self.simclr(x)

        # -----------------
        # do something new with GPT logits + simclr_features
        # -----------------

        loss = self.criterion(logits.view(-1, logits.size(-1)), x.view(-1).long())

        logs = {"loss": loss}
        return {"loss": loss, "log": logs}

Or perhaps your research is in self_supervised_learning and you want to do a new SimCLR. In this case, the only thing you want to change is the loss.

By subclassing you can focus on changing a single piece of a system without worrying that the other parts work (because if they are in Bolts, then they do and we’ve tested it).

# subclass SimCLR and change ONLY what you want to try
class ComplexCLR(SimCLR):

    def init_loss(self):
        return self.new_xent_loss

    def new_xent_loss(self):
        out = torch.cat([out_1, out_2], dim=0) n_samples = len(out)

        # Full similarity matrix
        cov = torch.mm(out, out.t().contiguous())
        sim = torch.exp(cov / temperature)

        # Negative similarity
        mask = ~torch.eye(n_samples, device=sim.device).bool()
        neg = sim.masked_select(mask).view(n_samples, -1).sum(dim=-1)

        # ------------------
        # some new thing we want to do
        # ------------------

        # Positive similarity :
        pos = torch.exp(torch.sum(out_1 * out_2, dim=-1) / temperature)
        pos = torch.cat([pos, pos], dim=0)
        loss = -torch.log(pos / neg).mean()

        return loss

Callbacks¶

Callbacks are arbitrary programs which can run at any points in time within a training loop in Lightning.

Bolts houses a collection of callbacks that are community contributed and can work in any Lightning Module!

from pl_bolts.callbacks import PrintTableMetricsCallback
import pytorch_lightning as pl

trainer = pl.Trainer(callbacks=[PrintTableMetricsCallback()])

DataModules¶

In PyTorch, working with data has these major elements.

Downloading, saving and preparing the dataset.

Splitting into train, val and test.

For each split, applying different transforms

A DataModule groups together those actions into a single reproducible DataModule that can be shared around to guarantee:

Consistent data preprocessing (download, splits, etc…)

The same exact splits

The same exact transforms

from pl_bolts.datamodules import ImagenetDataModule

dm = ImagenetDataModule(data_dir=PATH)

# standard PyTorch!
train_loader = dm.train_dataloader()
val_loader = dm.val_dataloader()
test_loader = dm.test_dataloader()

Trainer().fit(
    model,
    train_loader,
    val_loader
)

But when paired with PyTorch LightningModules (all bolts models), you can plug and play full dataset definitions with the same splits, transforms, etc…

imagenet = ImagenetDataModule(PATH)
model = VAE(datamodule=imagenet)
model = ImageGPT(datamodule=imagenet)
model = GAN(datamodule=imagenet)

We even have prebuilt modules to bridge the gap between Numpy, Sklearn and PyTorch

from sklearn.datasets import load_boston
from pl_bolts.datamodules import SklearnDataModule

X, y = load_boston(return_X_y=True)
datamodule = SklearnDataModule(X, y)

model = LitModel(datamodule)

Regression Heroes¶

In case your job or research doesn’t need a “hammer”, we offer implementations of Classic ML models which benefit from lightning’s multi-GPU and TPU support.

So, now you can run huge workloads scalably, without needing to do any engineering. For instance, here we can run logistic Regression on Imagenet (each epoch takes about 3 minutes)!

from pl_bolts.models.regression import LogisticRegression

imagenet = ImagenetDataModule(PATH)

# 224 x 224 x 3
pixels_per_image = 150528
model = LogisticRegression(input_dim=pixels_per_image, num_classes=1000)
model.prepare_data = imagenet.prepare_data

trainer = Trainer(gpus=2)
trainer.fit(
    model,
    imagenet.train_dataloader(batch_size=256),
    imagenet.val_dataloader(batch_size=256)
)

Linear Regression¶

Here’s an example for Linear regression

import pytorch_lightning as pl
from pl_bolts.datamodules import SklearnDataModule
from sklearn.datasets import load_boston

# link the numpy dataset to PyTorch
X, y = load_boston(return_X_y=True)
loaders = SklearnDataModule(X, y)

# training runs training batches while validating against a validation set
model = LinearRegression()
trainer = pl.Trainer(num_gpus=8)
trainer.fit(model, loaders.train_dataloader(), loaders.val_dataloader())

Once you’re done, you can run the test set if needed.

trainer.test(test_dataloaders=loaders.test_dataloader())

But more importantly, you can scale up to many GPUs, TPUs or even CPUs

# 8 GPUs
trainer = pl.Trainer(num_gpus=8)

# 8 TPU cores
trainer = pl.Trainer(tpu_cores=8)

# 32 GPUs
trainer = pl.Trainer(num_gpus=8, num_nodes=4)

# 128 CPUs
trainer = pl.Trainer(num_processes=128)

Logistic Regression¶

Here’s an example for logistic regression

from sklearn.datasets import load_iris
from pl_bolts.models.regression import LogisticRegression
from pl_bolts.datamodules import SklearnDataModule
import pytorch_lightning as pl

# use any numpy or sklearn dataset
X, y = load_iris(return_X_y=True)
dm = SklearnDataModule(X, y)

# build model
model = LogisticRegression(input_dim=4, num_classes=3)

# fit
trainer = pl.Trainer(tpu_cores=8, precision=16)
trainer.fit(model, dm.train_dataloader(), dm.val_dataloader())

trainer.test(test_dataloaders=dm.test_dataloader(batch_size=12))

Any input will be flattened across all dimensions except the first one (batch). This means images, sound, etc… work out of the box.

# create dataset
dm = MNISTDataModule(num_workers=0, data_dir=tmpdir)

model = LogisticRegression(input_dim=28 * 28, num_classes=10, learning_rate=0.001)
model.prepare_data = dm.prepare_data
model.train_dataloader = dm.train_dataloader
model.val_dataloader = dm.val_dataloader
model.test_dataloader = dm.test_dataloader

trainer = pl.Trainer(max_epochs=2)
trainer.fit(model)
trainer.test(model)
# {test_acc: 0.92}

But more importantly, you can scale up to many GPUs, TPUs or even CPUs

# 8 GPUs
trainer = pl.Trainer(num_gpus=8)

# 8 TPUs
trainer = pl.Trainer(tpu_cores=8)

# 32 GPUs
trainer = pl.Trainer(num_gpus=8, num_nodes=4)

# 128 CPUs
trainer = pl.Trainer(num_processes=128)

Regular PyTorch¶

Everything in bolts also works with regular PyTorch since they are all just nn.Modules!

However, if you train using Lightning you don’t have to deal with engineering code :)

Command line support¶

Any bolt module can also be trained from the command line

cd pl_bolts/models/autoencoders/basic_vae
python basic_vae_pl_module.py

Each script accepts Argparse arguments for both the lightning trainer and the model

python basic_vae_pl_module.py --latent_dim 32 --batch_size 32 --gpus 4 --max_epochs 12

Model quality control¶

For bolts to be added to the library we have a rigorous quality control checklist

Bolts vs my own repo¶

We hope you keep your own repo still! We want to link to it to let people know. However, by adding your contribution to bolts you get these additional benefits!

More visibility! (more people all over the world use your code)

We test your code on every PR (CPUs, GPUs, TPUs).

We host the docs (and test on every PR).

We help you build thorough, beautiful documentation.

We help you build robust tests.

We’ll pretrain expensive models for you and host weights.

We will improve the speed of your models!

Eligible for invited talks to discuss your implementation.

Lightning swag + involvement in the broader contributor community :)

Note

You still get to keep your attribution and be recognized for your work!

Note

Bolts is a community library built by incredible people like you!

Contribution requirements¶

Benchmarked¶

Models have known performance results on common baseline datasets.

Device agnostic¶

Models must work on CPUs, GPUs and TPUs without changing code. We help authors with this.

# bad
encoder.to(device)

Fast¶

We inspect models for computational inefficiencies and help authors meet the bar. Granted, sometimes the approaches are slow for mathematical reasons. But anything related to engineering we help overcome.

# bad
mtx = ...
for xi in rows:
    for yi in cols
        mxt[xi, yi] = ...

# good
x = x.item().numpy()
x = np.some_fx(x)
x = torch.tensor(x)

Tested¶

Models are tested on every PR (on CPUs, GPUs and soon TPUs).

Modular¶

Models are modularized to be extended and reused easily.

# GOOD!
class LitVAE(pl.LightningModule):

    def init_prior(self, ...):
        # enable users to override interesting parts of each model

    def init_posterior(self, ...):
        # enable users to override interesting parts of each model

# BAD
class LitVAE(pl.LightningModule):

    def __init__(self):
        self.prior = ...
        self.posterior = ...

Attribution¶

Any models and weights that are contributed are attributed to you as the author(s).

We request that each contribution have:

The original paper link

The list of paper authors

The link to the original paper code (if available)

The link to your repo

Your name and your team’s name as the implementation authors.

Your team’s affiliation

Any generated examples, or result plots.

Hyperparameter configurations for the results.

Thank you for all your amazing contributions!

The bar seems high¶

If your model doesn’t yet meet this bar, no worries! Please open the PR and our team of core contributors will help you get there!

Do you have contribution ideas?¶

Yes! Check the Github issues for requests from the Lightning team and the community! We’ll even work with you to finish your implementation! Then we’ll help you pretrain it and cover the compute costs when possible.

Build a Callback¶

This module houses a collection of callbacks that can be passed into the trainer

from pl_bolts.callbacks import PrintTableMetricsCallback
import pytorch_lightning as pl

trainer = pl.Trainer(callbacks=[PrintTableMetricsCallback()])

# loss│train_loss│val_loss│epoch
# ──────────────────────────────
# 2.2541470527648926│2.2541470527648926│2.2158432006835938│0

What is a Callback¶

A callback is a self-contained program that can be intertwined into a training pipeline without polluting the main research logic.

Create a Callback¶

Creating a callback is simple:

from pytorch_lightning.callbacks import Callback

class MyCallback(Callback)
    def on_epoch_end(self, trainer, pl_module):
        # do something

Please refer to Callback docs for a full list of the 20+ hooks available.

Info Callbacks¶

These callbacks give all sorts of useful information during training.

Print Table Metrics¶

This callbacks prints training metrics to a table. It’s very bare-bones for speed purposes.

class pl_bolts.callbacks.printing.PrintTableMetricsCallback[source]

Bases: pytorch_lightning.callbacks.Callback

Prints a table with the metrics in columns on every epoch end

Example:

from pl_bolts.callbacks import PrintTableMetricsCallback

callback = PrintTableMetricsCallback()

pass into trainer like so:

trainer = pl.Trainer(callbacks=[callback])
trainer.fit(...)

# ------------------------------
# at the end of every epoch it will print
# ------------------------------

# loss│train_loss│val_loss│epoch
# ──────────────────────────────
# 2.2541470527648926│2.2541470527648926│2.2158432006835938│0

Self-supervised Callbacks¶

Useful callbacks for self-supervised learning models

BYOLMAWeightUpdate¶

The exponential moving average weight-update rule from Bring Your Own Latent (BYOL).

class pl_bolts.callbacks.self_supervised.BYOLMAWeightUpdate(initial_tau=0.996)[source]

Bases: pytorch_lightning.Callback

Weight update rule from BYOL.

Your model should have a:

self.online_network.

self.target_network.

Updates the target_network params using an exponential moving average update rule weighted by tau. BYOL claims this keeps the online_network from collapsing.

Note

Automatically increases tau from initial_tau to 1.0 with every training step

Example:

from pl_bolts.callbacks.self_supervised import BYOLMAWeightUpdate

# model must have 2 attributes
model = Model()
model.online_network = ...
model.target_network = ...

trainer = Trainer(callbacks=[BYOLMAWeightUpdate()])

Parameters: initial_tau¶ – starting tau. Auto-updates with every training step

SSLOnlineEvaluator¶

Appends a MLP for fine-tuning to the given model. Callback has its own mini-inner loop.

class pl_bolts.callbacks.self_supervised.SSLOnlineEvaluator(drop_p=0.2, hidden_dim=1024, z_dim=None, num_classes=None)[source]

Bases: pytorch_lightning.Callback

Attaches a MLP for finetuning using the standard self-supervised protocol.

Example:

from pl_bolts.callbacks.self_supervised import SSLOnlineEvaluator

# your model must have 2 attributes
model = Model()
model.z_dim = ... # the representation dim
model.num_classes = ... # the num of classes in the model

Parameters

drop_p¶ (float) – (0.2) dropout probability
hidden_dim¶ (int) –
1. the hidden dimension for the finetune MLP

get_representations(pl_module, x)[source]: Override this to customize for the particular model :param _sphinx_paramlinks_pl_bolts.callbacks.self_supervised.SSLOnlineEvaluator.get_representations.pl_module: :param _sphinx_paramlinks_pl_bolts.callbacks.self_supervised.SSLOnlineEvaluator.get_representations.x:

Variational Callbacks¶

Useful callbacks for GANs, variational-autoencoders or anything with latent spaces.

Latent Dim Interpolator¶

Interpolates latent dims.

Example output:

class pl_bolts.callbacks.variational.LatentDimInterpolator(interpolate_epoch_interval=20, range_start=-5, range_end=5, num_samples=2)[source]

Bases: pytorch_lightning.callbacks.Callback

Interpolates the latent space for a model by setting all dims to zero and stepping through the first two dims increasing one unit at a time.

Default interpolates between [-5, 5] (-5, -4, -3, …, 3, 4, 5)

Example:

from pl_bolts.callbacks import LatentDimInterpolator

Trainer(callbacks=[LatentDimInterpolator()])

Parameters

interpolate_epoch_interval¶ –
range_start¶ – default -5
range_end¶ – default 5
num_samples¶ – default 2

Vision Callbacks¶

Useful callbacks for vision models

Confused Logit¶

Shows how the input would have to change to move the prediction from one logit to the other

Example outputs:

class pl_bolts.callbacks.vision.confused_logit.ConfusedLogitCallback(top_k, projection_factor=3, min_logit_value=5.0, logging_batch_interval=20, max_logit_difference=0.1)[source]

Bases: pytorch_lightning.Callback

Takes the logit predictions of a model and when the probabilities of two classes are very close, the model doesn’t have high certainty that it should pick one vs the other class.

This callback shows how the input would have to change to swing the model from one label prediction to the other.

In this case, the network predicts a 5… but gives almost equal probability to an 8. The images show what about the original 5 would have to change to make it more like a 5 or more like an 8.

For each confused logit the confused images are generated by taking the gradient from a logit wrt an input for the top two closest logits.

Example:

from pl_bolts.callbacks.vision import ConfusedLogitCallback
trainer = Trainer(callbacks=[ConfusedLogitCallback()])

Note

whenever called, this model will look for self.last_batch and self.last_logits in the LightningModule

Note

this callback supports tensorboard only right now

Parameters

top_k¶ – How many “offending” images we should plot
projection_factor¶ – How much to multiply the input image to make it look more like this logit label
min_logit_value¶ – Only consider logit values above this threshold
logging_batch_interval¶ – how frequently to inspect/potentially plot something
max_logit_difference¶ – when the top 2 logits are within this threshold we consider them confused

Authored by:

Alfredo Canziani

Tensorboard Image Generator¶

Generates images from a generative model and plots to tensorboard

class pl_bolts.callbacks.vision.image_generation.TensorboardGenerativeModelImageSampler(num_samples=3)[source]

Bases: pytorch_lightning.Callback

Generates images and logs to tensorboard. Your model must implement the forward function for generation

Requirements:

# model must have img_dim arg
model.img_dim = (1, 28, 28)

# model forward must work for sampling
z = torch.rand(batch_size, latent_dim)
img_samples = your_model(z)

Example:

from pl_bolts.callbacks import TensorboardGenerativeModelImageSampler

trainer = Trainer(callbacks=[TensorboardGenerativeModelImageSampler()])

DataModules¶

DataModules (introduced in PyTorch Lightning 0.9.0) decouple the data from a model. A DataModule is simply a collection of a training dataloder, val dataloader and test dataloader. In addition, it specifies how to:

Download/prepare data.
Train/val/test splits.
Transform

Then you can use it like this:

Example:

dm = MNISTDataModule('path/to/data')
model = LitModel()

trainer = Trainer()
trainer.fit(model, dm)

Or use it manually with plain PyTorch

Example:

dm = MNISTDataModule('path/to/data')
for batch in dm.train_dataloader():
    ...
for batch in dm.val_dataloader():
    ...
for batch in dm.test_dataloader():
    ...

Please visit the PyTorch Lightning documentation for more details on DataModules

Sklearn Datamodule¶

Utilities to map sklearn or numpy datasets to PyTorch Dataloaders with automatic data splits and GPU/TPU support.

from sklearn.datasets import load_boston
from pl_bolts.datamodules import SklearnDataModule

X, y = load_boston(return_X_y=True)
loaders = SklearnDataModule(X, y)

train_loader = loaders.train_dataloader(batch_size=32)
val_loader = loaders.val_dataloader(batch_size=32)
test_loader = loaders.test_dataloader(batch_size=32)

Or build your own torch datasets

from sklearn.datasets import load_boston
from pl_bolts.datamodules import SklearnDataset

X, y = load_boston(return_X_y=True)
dataset = SklearnDataset(X, y)
loader = DataLoader(dataset)

Sklearn Dataset Class¶

Transforms a sklearn or numpy dataset to a PyTorch Dataset.

class pl_bolts.datamodules.sklearn_datamodule.SklearnDataset(X, y, X_transform=None, y_transform=None)[source]

Bases: torch.utils.data.Dataset

Mapping between numpy (or sklearn) datasets to PyTorch datasets.

Parameters

X¶ (ndarray) – Numpy ndarray
y¶ (ndarray) – Numpy ndarray
X_transform¶ (Optional[Any]) – Any transform that works with Numpy arrays
y_transform¶ (Optional[Any]) – Any transform that works with Numpy arrays

Example

>>> from sklearn.datasets import load_boston
>>> from pl_bolts.datamodules import SklearnDataset
...
>>> X, y = load_boston(return_X_y=True)
>>> dataset = SklearnDataset(X, y)
>>> len(dataset)
506

Sklearn DataModule Class¶

Automatically generates the train, validation and test splits for a Numpy dataset. They are set up as dataloaders for convenience. Optionally, you can pass in your own validation and test splits.

class pl_bolts.datamodules.sklearn_datamodule.SklearnDataModule(X, y, x_val=None, y_val=None, x_test=None, y_test=None, val_split=0.2, test_split=0.1, num_workers=2, random_state=1234, shuffle=True, *args, **kwargs)[source]

Bases: pytorch_lightning.LightningDataModule

Automatically generates the train, validation and test splits for a Numpy dataset. They are set up as dataloaders for convenience. Optionally, you can pass in your own validation and test splits.

Example

>>> from sklearn.datasets import load_boston
>>> from pl_bolts.datamodules import SklearnDataModule
...
>>> X, y = load_boston(return_X_y=True)
>>> loaders = SklearnDataModule(X, y)
...
>>> # train set
>>> train_loader = loaders.train_dataloader(batch_size=32)
>>> len(train_loader.dataset)
355
>>> len(train_loader)
11
>>> # validation set
>>> val_loader = loaders.val_dataloader(batch_size=32)
>>> len(val_loader.dataset)
100
>>> len(val_loader)
3
>>> # test set
>>> test_loader = loaders.test_dataloader(batch_size=32)
>>> len(test_loader.dataset)
51
>>> len(test_loader)
1

Vision DataModules¶

The following are pre-built datamodules for computer-vision.

Supervised learning¶

These are standard vision datasets with the train, test, val splits pre-generated in DataLoaders with the standard transforms (and Normalization) values

BinaryMNIST¶

class pl_bolts.datamodules.binary_mnist_datamodule.BinaryMNISTDataModule(data_dir, val_split=5000, num_workers=16, normalize=False, seed=42, *args, **kwargs)[source]

Bases: pytorch_lightning.LightningDataModule

Specs:

10 classes (1 per digit)
Each image is (1 x 28 x 28)

Binary MNIST, train, val, test splits and transforms

Transforms:

mnist_transforms = transform_lib.Compose([
    transform_lib.ToTensor()
])

Example:

from pl_bolts.datamodules import BinaryMNISTDataModule

dm = BinaryMNISTDataModule('.')
model = LitModel()

Trainer().fit(model, dm)

Parameters

data_dir¶ (str) – where to save/load the data
val_split¶ (int) – how many of the training images to use for the validation split
num_workers¶ (int) – how many workers to use for loading data
normalize¶ (bool) – If true applies image normalize

prepare_data()[source]: Saves MNIST files to data_dir

test_dataloader(batch_size=32, transforms=None)[source]

MNIST test set uses the test split

Parameters

batch_size¶ – size of batch
transforms¶ – custom transforms

train_dataloader(batch_size=32, transforms=None)[source]

MNIST train set removes a subset to use for validation

Parameters

batch_size¶ – size of batch
transforms¶ – custom transforms

val_dataloader(batch_size=32, transforms=None)[source]

MNIST val set uses a subset of the training set for validation

Parameters

batch_size¶ – size of batch
transforms¶ – custom transforms

property num_classes[source]: Return: 10

CityScapes¶

class pl_bolts.datamodules.cityscapes_datamodule.CityscapesDataModule(data_dir, val_split=5000, num_workers=16, batch_size=32, seed=42, *args, **kwargs)[source]

Bases: pytorch_lightning.LightningDataModule

Standard Cityscapes, train, val, test splits and transforms

Specs:

30 classes (road, person, sidewalk, etc…)
(image, target) - image dims: (3 x 32 x 32), target dims: (3 x 32 x 32)

Transforms:

transforms = transform_lib.Compose([
    transform_lib.ToTensor(),
    transform_lib.Normalize(
        mean=[0.28689554, 0.32513303, 0.28389177],
        std=[0.18696375, 0.19017339, 0.18720214]
    )
])

Example:

from pl_bolts.datamodules import CityscapesDataModule

dm = CityscapesDataModule(PATH)
model = LitModel()

Trainer().fit(model, dm)

Or you can set your own transforms

Example:

dm.train_transforms = ...
dm.test_transforms = ...
dm.val_transforms  = ...

Parameters

data_dir¶ – where to save/load the data
val_split¶ – how many of the training images to use for the validation split
num_workers¶ – how many workers to use for loading data
batch_size¶ – number of examples per training/eval step

prepare_data()[source]: Saves Cityscapes files to data_dir

test_dataloader()[source]: Cityscapes test set uses the test split

train_dataloader()[source]: Cityscapes train set with removed subset to use for validation

val_dataloader()[source]: Cityscapes val set uses a subset of the training set for validation

property num_classes[source]: Return: 30

CIFAR-10¶

class pl_bolts.datamodules.cifar10_datamodule.CIFAR10DataModule(data_dir=None, val_split=5000, num_workers=16, batch_size=32, seed=42, *args, **kwargs)[source]

Bases: pytorch_lightning.LightningDataModule

Specs:

10 classes (1 per class)
Each image is (3 x 32 x 32)

Standard CIFAR10, train, val, test splits and transforms

Transforms:

mnist_transforms = transform_lib.Compose([
    transform_lib.ToTensor(),
    transforms.Normalize(
        mean=[x / 255.0 for x in [125.3, 123.0, 113.9]],
        std=[x / 255.0 for x in [63.0, 62.1, 66.7]]
    )
])

Example:

from pl_bolts.datamodules import CIFAR10DataModule

dm = CIFAR10DataModule(PATH)
model = LitModel()

Trainer().fit(model, dm)

Or you can set your own transforms

Example:

dm.train_transforms = ...
dm.test_transforms = ...
dm.val_transforms  = ...

Parameters

data_dir¶ (Optional[str]) – where to save/load the data
val_split¶ (int) – how many of the training images to use for the validation split
num_workers¶ (int) – how many workers to use for loading data
batch_size¶ (int) – number of examples per training/eval step

prepare_data()[source]: Saves CIFAR10 files to data_dir

test_dataloader()[source]: CIFAR10 test set uses the test split

train_dataloader()[source]: CIFAR train set removes a subset to use for validation

val_dataloader()[source]: CIFAR10 val set uses a subset of the training set for validation

property num_classes[source]: Return: 10

FashionMNIST¶

class pl_bolts.datamodules.fashion_mnist_datamodule.FashionMNISTDataModule(data_dir, val_split=5000, num_workers=16, seed=42, *args, **kwargs)[source]

Bases: pytorch_lightning.LightningDataModule

Specs:

10 classes (1 per type)
Each image is (1 x 28 x 28)

Standard FashionMNIST, train, val, test splits and transforms

Transforms:

mnist_transforms = transform_lib.Compose([
    transform_lib.ToTensor()
])

Example:

from pl_bolts.datamodules import FashionMNISTDataModule

dm = FashionMNISTDataModule('.')
model = LitModel()

Trainer().fit(model, dm)

Parameters

data_dir¶ (str) – where to save/load the data
val_split¶ (int) – how many of the training images to use for the validation split
num_workers¶ (int) – how many workers to use for loading data

prepare_data()[source]: Saves FashionMNIST files to data_dir

test_dataloader(batch_size=32, transforms=None)[source]

FashionMNIST test set uses the test split

Parameters

batch_size¶ – size of batch
transforms¶ – custom transforms

train_dataloader(batch_size=32, transforms=None)[source]

FashionMNIST train set removes a subset to use for validation

Parameters

batch_size¶ – size of batch
transforms¶ – custom transforms

val_dataloader(batch_size=32, transforms=None)[source]

FashionMNIST val set uses a subset of the training set for validation

Parameters

batch_size¶ – size of batch
transforms¶ – custom transforms

property num_classes[source]: Return: 10

Imagenet¶

class pl_bolts.datamodules.imagenet_datamodule.ImagenetDataModule(data_dir, meta_dir=None, num_imgs_per_val_class=50, image_size=224, num_workers=16, batch_size=32, *args, **kwargs)[source]

Bases: pytorch_lightning.LightningDataModule

Specs:

1000 classes
Each image is (3 x varies x varies) (here we default to 3 x 224 x 224)

Imagenet train, val and test dataloaders.

The train set is the imagenet train.

The val set is taken from the train set with num_imgs_per_val_class images per class. For example if num_imgs_per_val_class=2 then there will be 2,000 images in the validation set.

The test set is the official imagenet validation set.

Example:

from pl_bolts.datamodules import ImagenetDataModule

dm = ImagenetDataModule(IMAGENET_PATH)
model = LitModel()

Trainer().fit(model, dm)

Parameters

data_dir¶ (str) – path to the imagenet dataset file
meta_dir¶ (Optional[str]) – path to meta.bin file
num_imgs_per_val_class¶ (int) – how many images per class for the validation set
image_size¶ (int) – final image size
num_workers¶ (int) – how many data workers
batch_size¶ (int) – batch_size

prepare_data()[source]: This method already assumes you have imagenet2012 downloaded. It validates the data using the meta.bin.

Warning

Please download imagenet on your own first.

test_dataloader()[source]: Uses the validation split of imagenet2012 for testing

train_dataloader()[source]: Uses the train split of imagenet2012 and puts away a portion of it for the validation split

train_transform()[source]

The standard imagenet transforms

transform_lib.Compose([
    transform_lib.RandomResizedCrop(self.image_size),
    transform_lib.RandomHorizontalFlip(),
    transform_lib.ToTensor(),
    transform_lib.Normalize(
        mean=[0.485, 0.456, 0.406],
        std=[0.229, 0.224, 0.225]
    ),
])

val_dataloader()[source]

Uses the part of the train split of imagenet2012 that was not used for training via num_imgs_per_val_class

Parameters

batch_size¶ – the batch size
transforms¶ – the transforms

val_transform()[source]

The standard imagenet transforms for validation

transform_lib.Compose([
    transform_lib.Resize(self.image_size + 32),
    transform_lib.CenterCrop(self.image_size),
    transform_lib.ToTensor(),
    transform_lib.Normalize(
        mean=[0.485, 0.456, 0.406],
        std=[0.229, 0.224, 0.225]
    ),
])

property num_classes[source]

Return:

1000

MNIST¶

class pl_bolts.datamodules.mnist_datamodule.MNISTDataModule(data_dir='./', val_split=5000, num_workers=16, normalize=False, seed=42, batch_size=32, *args, **kwargs)[source]

Bases: pytorch_lightning.LightningDataModule

Specs:

10 classes (1 per digit)
Each image is (1 x 28 x 28)

Standard MNIST, train, val, test splits and transforms

Transforms:

mnist_transforms = transform_lib.Compose([
    transform_lib.ToTensor()
])

Example:

from pl_bolts.datamodules import MNISTDataModule

dm = MNISTDataModule('.')
model = LitModel()

Trainer().fit(model, dm)

Parameters

data_dir¶ (str) – where to save/load the data
val_split¶ (int) – how many of the training images to use for the validation split
num_workers¶ (int) – how many workers to use for loading data
normalize¶ (bool) – If true applies image normalize

prepare_data()[source]: Saves MNIST files to data_dir

test_dataloader(batch_size=32, transforms=None)[source]

MNIST test set uses the test split

Parameters

batch_size¶ – size of batch
transforms¶ – custom transforms

train_dataloader(batch_size=32, transforms=None)[source]

MNIST train set removes a subset to use for validation

Parameters

batch_size¶ – size of batch
transforms¶ – custom transforms

val_dataloader(batch_size=32, transforms=None)[source]

MNIST val set uses a subset of the training set for validation

Parameters

batch_size¶ – size of batch
transforms¶ – custom transforms

property num_classes[source]: Return: 10

Semi-supervised learning¶

The following datasets have support for unlabeled training and semi-supervised learning where only a few examples are labeled.

Imagenet (ssl)¶

class pl_bolts.datamodules.ssl_imagenet_datamodule.SSLImagenetDataModule(data_dir, meta_dir=None, num_workers=16, *args, **kwargs)[source]: Bases: pytorch_lightning.LightningDataModule

STL-10¶

class pl_bolts.datamodules.stl10_datamodule.STL10DataModule(data_dir=None, unlabeled_val_split=5000, train_val_split=500, num_workers=16, batch_size=32, seed=42, *args, **kwargs)[source]

Bases: pytorch_lightning.LightningDataModule

Specs:

10 classes (1 per type)
Each image is (3 x 96 x 96)

Standard STL-10, train, val, test splits and transforms. STL-10 has support for doing validation splits on the labeled or unlabeled splits

Transforms:

mnist_transforms = transform_lib.Compose([
    transform_lib.ToTensor(),
    transforms.Normalize(
        mean=(0.43, 0.42, 0.39),
        std=(0.27, 0.26, 0.27)
    )
])

Example:

from pl_bolts.datamodules import STL10DataModule

dm = STL10DataModule(PATH)
model = LitModel()

Trainer().fit(model, dm)

Parameters

data_dir¶ (Optional[str]) – where to save/load the data
unlabeled_val_split¶ (int) – how many images from the unlabeled training split to use for validation
train_val_split¶ (int) – how many images from the labeled training split to use for validation
num_workers¶ (int) – how many workers to use for loading data
batch_size¶ (int) – the batch size

prepare_data()[source]: Downloads the unlabeled, train and test split

test_dataloader()[source]

Loads the test split of STL10

Parameters

batch_size¶ – the batch size
transforms¶ – the transforms

train_dataloader()[source]: Loads the ‘unlabeled’ split minus a portion set aside for validation via unlabeled_val_split.

train_dataloader_mixed()[source]

Loads a portion of the ‘unlabeled’ training data and ‘train’ (labeled) data. both portions have a subset removed for validation via unlabeled_val_split and train_val_split

Parameters

batch_size¶ – the batch size
transforms¶ – a sequence of transforms

val_dataloader()[source]

Loads a portion of the ‘unlabeled’ training data set aside for validation The val dataset = (unlabeled - train_val_split)

Parameters

batch_size¶ – the batch size
transforms¶ – a sequence of transforms

val_dataloader_mixed()[source]

Loads a portion of the ‘unlabeled’ training data set aside for validation along with the portion of the ‘train’ dataset to be used for validation

unlabeled_val = (unlabeled - train_val_split)

labeled_val = (train- train_val_split)

full_val = unlabeled_val + labeled_val

Parameters

batch_size¶ – the batch size
transforms¶ – a sequence of transforms

AsynchronousLoader¶

This dataloader behaves identically to the standard pytorch dataloader, but will transfer data asynchronously to the GPU with training. You can also use it to wrap an existing dataloader.

Example:

dataloader = AsynchronousLoader(DataLoader(ds, batch_size=16), device=device)

for b in dataloader:
    ...

class pl_bolts.datamodules.async_dataloader.AsynchronousLoader(data, device=torch.device, q_size=10, num_batches=None, **kwargs)[source]

Bases: object

Class for asynchronously loading from CPU memory to device memory with DataLoader.

Note that this only works for single GPU training, multiGPU uses PyTorch’s DataParallel or DistributedDataParallel which uses its own code for transferring data across GPUs. This could just break or make things slower with DataParallel or DistributedDataParallel.

Parameters

data¶ – The PyTorch Dataset or DataLoader we’re using to load.
device¶ – The PyTorch device we are loading to
q_size¶ – Size of the queue used to store the data loaded to the device
num_batches¶ – Number of batches to load. This must be set if the dataloader doesn’t have a finite __len__. It will also override DataLoader.__len__ if set and DataLoader has a __len__. Otherwise it can be left as None
**kwargs¶ – Any additional arguments to pass to the dataloader if we’re constructing one here

Losses¶

This package lists common losses across research domains (This is a work in progress. If you have any losses you want to contribute, please submit a PR!)

Note

this module is a work in progress

Your Loss¶

We’re cleaning up many of our losses, but in the meantime, submit a PR to add your loss here!

Reinforcement Learning¶

These are common losses used in RL.

DQN Loss¶

pl_bolts.losses.rl.dqn_loss(batch, net, target_net, gamma=0.99)[source]

Calculates the mse loss using a mini batch from the replay buffer :type _sphinx_paramlinks_pl_bolts.losses.rl.dqn_loss.batch: Tuple[Tensor, Tensor] :param _sphinx_paramlinks_pl_bolts.losses.rl.dqn_loss.batch: current mini batch of replay data :type _sphinx_paramlinks_pl_bolts.losses.rl.dqn_loss.net: Module :param _sphinx_paramlinks_pl_bolts.losses.rl.dqn_loss.net: main training network :type _sphinx_paramlinks_pl_bolts.losses.rl.dqn_loss.target_net: Module :param _sphinx_paramlinks_pl_bolts.losses.rl.dqn_loss.target_net: target network of the main training network :type _sphinx_paramlinks_pl_bolts.losses.rl.dqn_loss.gamma: float :param _sphinx_paramlinks_pl_bolts.losses.rl.dqn_loss.gamma: discount factor

Return type: Tensor
Returns: loss

Double DQN Loss¶

pl_bolts.losses.rl.double_dqn_loss(batch, net, target_net, gamma=0.99)[source]

Calculates the mse loss using a mini batch from the replay buffer. This uses an improvement to the original DQN loss by using the double dqn. This is shown by using the actions of the train network to pick the value from the target network. This code is heavily commented in order to explain the process clearly :type _sphinx_paramlinks_pl_bolts.losses.rl.double_dqn_loss.batch: Tuple[Tensor, Tensor] :param _sphinx_paramlinks_pl_bolts.losses.rl.double_dqn_loss.batch: current mini batch of replay data :type _sphinx_paramlinks_pl_bolts.losses.rl.double_dqn_loss.net: Module :param _sphinx_paramlinks_pl_bolts.losses.rl.double_dqn_loss.net: main training network :type _sphinx_paramlinks_pl_bolts.losses.rl.double_dqn_loss.target_net: Module :param _sphinx_paramlinks_pl_bolts.losses.rl.double_dqn_loss.target_net: target network of the main training network :type _sphinx_paramlinks_pl_bolts.losses.rl.double_dqn_loss.gamma: float :param _sphinx_paramlinks_pl_bolts.losses.rl.double_dqn_loss.gamma: discount factor

Return type: Tensor
Returns: loss

Per DQN Loss¶

pl_bolts.losses.rl.per_dqn_loss(batch, batch_weights, net, target_net, gamma=0.99)[source]

Calculates the mse loss with the priority weights of the batch from the PER buffer :type _sphinx_paramlinks_pl_bolts.losses.rl.per_dqn_loss.batch: Tuple[Tensor, Tensor] :param _sphinx_paramlinks_pl_bolts.losses.rl.per_dqn_loss.batch: current mini batch of replay data :type _sphinx_paramlinks_pl_bolts.losses.rl.per_dqn_loss.batch_weights: List :param _sphinx_paramlinks_pl_bolts.losses.rl.per_dqn_loss.batch_weights: how each of these samples are weighted in terms of priority :type _sphinx_paramlinks_pl_bolts.losses.rl.per_dqn_loss.net: Module :param _sphinx_paramlinks_pl_bolts.losses.rl.per_dqn_loss.net: main training network :type _sphinx_paramlinks_pl_bolts.losses.rl.per_dqn_loss.target_net: Module :param _sphinx_paramlinks_pl_bolts.losses.rl.per_dqn_loss.target_net: target network of the main training network :type _sphinx_paramlinks_pl_bolts.losses.rl.per_dqn_loss.gamma: float :param _sphinx_paramlinks_pl_bolts.losses.rl.per_dqn_loss.gamma: discount factor

Return type: Tuple[Tensor, ndarray]
Returns: loss and batch_weights

How to use models¶

Models are meant to be “bolted” onto your research or production cases.

Bolts are meant to be used in the following ways

Predicting on your data¶

Most bolts have pretrained weights trained on various datasets or algorithms. This is useful when you don’t have enough data, time or money to do your own training.

For example, you could use a pretrained VAE to generate features for an image dataset.

from pl_bolts.models.autoencoders import VAE

model = VAE(pretrained='imagenet2012')
encoder = model.encoder
encoder.freeze()

for (x, y) in own_data
    features = encoder(x)

The advantage of bolts is that each system can be decomposed and used in interesting ways. For instance, this resnet18 was trained using self-supervised learning (no labels) on Imagenet, and thus might perform better than the same resnet18 trained with labels

# trained without labels
from pl_bolts.models.self_supervised import CPCV2

model = CPCV2(encoder='resnet18', pretrained='imagenet128')
resnet18_unsupervised = model.encoder.freeze()

# trained with labels
from torchvision.models import resnet18
resnet18_supervised = resnet18(pretrained=True)

# perhaps the features when trained without labels are much better for classification or other tasks
x = image_sample()
unsup_feats = resnet18_unsupervised(x)
sup_feats = resnet18_supervised(x)

# which one will be better?

Bolts are often trained on more than just one dataset.

model = CPCV2(encoder='resnet18', pretrained='stl10')

Finetuning on your data¶

If you have a little bit of data and can pay for a bit of training, it’s often better to finetune on your own data.

To finetune you have two options unfrozen finetuning or unfrozen later.

Unfrozen Finetuning¶

In this approach, we load the pretrained model and unfreeze from the beginning

model = CPCV2(encoder='resnet18', pretrained='imagenet128')
resnet18 = model.encoder
# don't call .freeze()

classifier = LogisticRegression()

for (x, y) in own_data:
    feats = resnet18(x)
    y_hat = classifier(feats)
    ...

Or as a LightningModule

class FineTuner(pl.LightningModule):

    def __init__(self, encoder):
        self.encoder = encoder
        self.classifier = LogisticRegression()

    def training_step(self, batch, batch_idx):
        (x, y) = batch
        feats = self.encoder(x)
        y_hat = self.classifier(feats)
        loss = cross_entropy_with_logits(y_hat, y)
        return loss

trainer = Trainer(gpus=2)
model = FineTuner(resnet18)
trainer.fit(model)

Sometimes this works well, but more often it’s better to keep the encoder frozen for a while

Freeze then unfreeze¶

The approach that works best most often is to freeze first then unfreeze later

# freeze!
model = CPCV2(encoder='resnet18', pretrained='imagenet128')
resnet18 = model.encoder
resnet18.freeze()

classifier = LogisticRegression()

for epoch in epochs:
    for (x, y) in own_data:
        feats = resnet18(x)
        y_hat = classifier(feats)
        loss = cross_entropy_with_logits(y_hat, y)

    # unfreeze after 10 epochs
    if epoch == 10:
        resnet18.unfreeze()

Note

In practice, unfreezing later works MUCH better.

Or in Lightning as a Callback so you don’t pollute your research code.

class UnFreezeCallback(Callback):

    def on_epoch_end(self, trainer, pl_module):
        if trainer.current_epoch == 10.
            encoder.unfreeze()

trainer = Trainer(gpus=2, callbacks=[UnFreezeCallback()])
model = FineTuner(resnet18)
trainer.fit(model)

Unless you still need to mix it into your research code.

class FineTuner(pl.LightningModule):

    def __init__(self, encoder):
        self.encoder = encoder
        self.classifier = LogisticRegression()

    def training_step(self, batch, batch_idx):

        # option 1 - (not recommended because it's messy)
        if self.trainer.current_epoch == 10:
            self.encoder.unfreeze()

        (x, y) = batch
        feats = self.encoder(x)
        y_hat = self.classifier(feats)
        loss = cross_entropy_with_logits(y_hat, y)
        return loss

    def on_epoch_end(self, trainer, pl_module):
        # a hook is cleaner (but a callback is much better)
        if self.trainer.current_epoch == 10:
            self.encoder.unfreeze()

Hyperparameter search¶

For finetuning to work well, you should try many versions of the model hyperparameters. Otherwise you’re unlikely to get the most value out of your data.

learning_rates = [0.01, 0.001, 0.0001]
hidden_dim = [128, 256, 512]

for lr in learning_rates:
    for hd in hidden_dim:
        vae = VAE(hidden_dim=hd, learning_rate=lr)
        trainer = Trainer()
        trainer.fit(vae)

Train from scratch¶

If you do have enough data and compute resources, then you could try training from scratch.

# get data
train_data = DataLoader(YourDataset)
val_data = DataLoader(YourDataset)

# use any bolts model without pretraining
model = VAE()

# fit!
trainer = Trainer(gpus=2)
trainer.fit(model, train_data, val_data)

Note

For this to work well, make sure you have enough data and time to train these models!

For research¶

What separates bolts from all the other libraries out there is that bolts is built by and used by AI researchers. This means every single bolt is modularized so that it can be easily extended or mixed with arbitrary parts of the rest of the code-base.

Extending work¶

Perhaps a research project requires modifying a part of a know approach. In this case, you’re better off only changing that part of a system that is already know to perform well. Otherwise, you risk not implementing the work correctly.

Example 1: Changing the prior or approx posterior of a VAE

from pl_bolts.models.autoencoders import VAE

class MyVAEFlavor(VAE):

    def init_prior(self, z_mu, z_std):
        P = MyPriorDistribution

        # default is standard normal
        # P = distributions.normal.Normal(loc=torch.zeros_like(z_mu), scale=torch.ones_like(z_std))
        return P

    def init_posterior(self, z_mu, z_std):
        Q = MyPosteriorDistribution
        # default is normal(z_mu, z_sigma)
        # Q = distributions.normal.Normal(loc=z_mu, scale=z_std)
        return Q

And of course train it with lightning.

model = MyVAEFlavor()
trainer = Trainer()
trainer.fit(model)

In just a few lines of code you changed something fundamental about a VAE… This means you can iterate through ideas much faster knowing that the bolt implementation and the training loop are CORRECT and TESTED.

If your model doesn’t work with the new P, Q, then you can discard that research idea much faster than trying to figure out if your VAE implementation was correct, or if your training loop was correct.

Example 2: Changing the generator step of a GAN

from pl_bolts.models.gans import GAN

class FancyGAN(GAN):

    def generator_step(self, x):
        # sample noise
        z = torch.randn(x.shape[0], self.hparams.latent_dim)
        z = z.type_as(x)

        # generate images
        self.generated_imgs = self(z)

        # ground truth result (ie: all real)
        real = torch.ones(x.size(0), 1)
        real = real.type_as(x)
        g_loss = self.generator_loss(real)

        tqdm_dict = {'g_loss': g_loss}
        output = OrderedDict({
            'loss': g_loss,
            'progress_bar': tqdm_dict,
            'log': tqdm_dict
        })
        return output

Example 3: Changing the way the loss is calculated in a contrastive self-supervised learning approach

from pl_bolts.models.self_supervised import AMDIM

class MyDIM(AMDIM):

    def validation_step(self, batch, batch_nb):
        [img_1, img_2], labels = batch

        # generate features
        r1_x1, r5_x1, r7_x1, r1_x2, r5_x2, r7_x2 = self.forward(img_1, img_2)

        # Contrastive task
        loss, lgt_reg = self.contrastive_task((r1_x1, r5_x1, r7_x1), (r1_x2, r5_x2, r7_x2))
        unsupervised_loss = loss.sum() + lgt_reg

        result = {
            'val_nce': unsupervised_loss
        }
        return result

Importing parts¶

All the bolts are modular. This means you can also arbitrarily mix and match fundamental blocks from across approaches.

Example 1: Use the VAE encoder for a GAN as a generator

from pl_bolts.models.gans import GAN
from pl_bolts.models.autoencoders.basic_vae import Encoder

class FancyGAN(GAN):

    def init_generator(self, img_dim):
        generator = Encoder(...)
        return generator

trainer = Trainer(...)
trainer.fit(FancyGAN())

Example 2: Use the contrastive task of AMDIM in CPC

from pl_bolts.models.self_supervised import AMDIM, CPCV2

default_amdim_task = AMDIM().contrastive_task
model = CPCV2(contrastive_task=default_amdim_task, encoder='cpc_default')
# you might need to modify the cpc encoder depending on what you use

Compose new ideas¶

You may also be interested in creating completely new approaches that mix and match all sorts of different pieces together

# this model is for illustration purposes, it makes no research sense but it's intended to show
# that you can be as creative and expressive as you want.
class MyNewContrastiveApproach(pl.LightningModule):

    def __init__(self):
        suoer().__init_()

        self.gan = GAN()
        self.vae = VAE()
        self.amdim = AMDIM()
        self.cpc = CPCV2

    def training_step(self, batch, batch_idx):
        (x, y) = batch

        feat_a = self.gan.generator(x)
        feat_b = self.vae.encoder(x)

        unsup_loss = self.amdim(feat_a) + self.cpc(feat_b)

        vae_loss = self.vae._step(batch)
        gan_loss = self.gan.generator_loss(x)

        return unsup_loss + vae_loss + gan_loss

Classic ML Models¶

This module implements classic machine learning models in PyTorch Lightning, including linear regression and logistic regression. Unlike other libraries that implement these models, here we use PyTorch to enable multi-GPU, multi-TPU and half-precision training.

Linear Regression¶

Linear regression fits a linear model between a real-valued target variable $y$ and one or more features $X$ . We estimate the regression coefficients that minimize the mean squared error between the predicted and true target values.

We formulate the linear regression model as a single-layer neural network. By default we include only one neuron in the output layer, although you can specify the output_dim yourself.

Add either L1 or L2 regularization, or both, by specifying the regularization strength (default 0).

from pl_bolts.models.regression import LinearRegression
import pytorch_lightning as pl
from pl_bolts.datamodules import SklearnDataModule
from sklearn.datasets import load_boston

X, y = load_boston(return_X_y=True)
loaders = SklearnDataModule(X, y)

model = LinearRegression(input_dim=13)
trainer = pl.Trainer()
trainer.fit(model, loaders.train_dataloader(), loaders.val_dataloader())
trainer.test(test_dataloaders=loaders.test_dataloader())

class pl_bolts.models.regression.linear_regression.LinearRegression(input_dim, output_dim=1, bias=True, learning_rate=0.0001, optimizer=torch.optim.Adam, l1_strength=0.0, l2_strength=0.0, **kwargs)[source]

Bases: pytorch_lightning.LightningModule

Linear regression model implementing - with optional L1/L2 regularization $$min_{W} ||(Wx + b) - y ||_2^2 $$

Parameters

input_dim¶ (int) – number of dimensions of the input (1+)
output_dim¶ (int) – number of dimensions of the output (default=1)
bias¶ (bool) – If false, will not use $+b$
learning_rate¶ (float) – learning_rate for the optimizer
optimizer¶ (Optimizer) – the optimizer to use (default=’Adam’)
l1_strength¶ (float) – L1 regularization strength (default=None)
l2_strength¶ (float) – L2 regularization strength (default=None)

Logistic Regression¶

Logistic regression is a linear model used for classification, i.e. when we have a categorical target variable. This implementation supports both binary and multi-class classification.

In the binary case, we formulate the logistic regression model as a one-layer neural network with one neuron in the output layer and a sigmoid activation function. In the multi-class case, we use a single-layer neural network but now with $k$ neurons in the output, where $k$ is the number of classes. This is also referred to as multinomial logistic regression.

Add either L1 or L2 regularization, or both, by specifying the regularization strength (default 0).

from sklearn.datasets import load_iris
from pl_bolts.models.regression import LogisticRegression
from pl_bolts.datamodules import SklearnDataModule
import pytorch_lightning as pl

# use any numpy or sklearn dataset
X, y = load_iris(return_X_y=True)
dm = SklearnDataModule(X, y)

# build model
model = LogisticRegression(input_dim=4, num_classes=3)

# fit
trainer = pl.Trainer(tpu_cores=8, precision=16)
trainer.fit(model, dm.train_dataloader(), dm.val_dataloader())

trainer.test(test_dataloaders=dm.test_dataloader(batch_size=12))

Any input will be flattened across all dimensions except the first one (batch). This means images, sound, etc… work out of the box.

# create dataset
dm = MNISTDataModule(num_workers=0, data_dir=tmpdir)

model = LogisticRegression(input_dim=28 * 28, num_classes=10, learning_rate=0.001)
model.prepare_data = dm.prepare_data
model.train_dataloader = dm.train_dataloader
model.val_dataloader = dm.val_dataloader
model.test_dataloader = dm.test_dataloader

trainer = pl.Trainer(max_epochs=2)
trainer.fit(model)
trainer.test(model)
# {test_acc: 0.92}

class pl_bolts.models.regression.logistic_regression.LogisticRegression(input_dim, num_classes, bias=True, learning_rate=0.0001, optimizer=torch.optim.Adam, l1_strength=0.0, l2_strength=0.0, **kwargs)[source]

Bases: pytorch_lightning.LightningModule

Logistic regression model

Parameters

input_dim¶ (int) – number of dimensions of the input (at least 1)
num_classes¶ (int) – number of class labels (binary: 2, multi-class: >2)
bias¶ (bool) – specifies if a constant or intercept should be fitted (equivalent to fit_intercept in sklearn)
learning_rate¶ (float) – learning_rate for the optimizer
optimizer¶ (Optimizer) – the optimizer to use (default=’Adam’)
l1_strength¶ (float) – L1 regularization strength (default=None)
l2_strength¶ (float) – L2 regularization strength (default=None)

Autoencoders¶

This section houses autoencoders and variational autoencoders.

Basic AE¶

This is the simplest autoencoder. You can use it like so

from pl_bolts.models.autoencoders import AE

model = AE()
trainer = Trainer()
trainer.fit(model)

You can override any part of this AE to build your own variation.

from pl_bolts.models.autoencoders import AE

class MyAEFlavor(AE):

    def init_encoder(self, hidden_dim, latent_dim, input_width, input_height):
        encoder = YourSuperFancyEncoder(...)
        return encoder

You can use the pretrained models present in bolts.

CIFAR-10 pretrained model:

from pl_bolts.models.autoencoders import AE

ae = AE(input_height=32)
print(AE.pretrained_weights_available())
ae = ae.from_pretrained('cifar10-resnet18')

ae.freeze()

Tensorboard for AE on CIFAR-10

Training:

Reconstructions:

Both input and generated images are normalized versions as the training was done with such images.

class pl_bolts.models.autoencoders.AE(input_height, enc_type='resnet18', first_conv=False, maxpool1=False, enc_out_dim=512, kl_coeff=0.1, latent_dim=256, lr=0.0001, **kwargs)[source]

Bases: pytorch_lightning.LightningModule

Standard AE

Model is available pretrained on different datasets:

Example:

# not pretrained
ae = AE()

# pretrained on cifar10
ae = AE.from_pretrained('cifar10-resnet18')

Parameters

input_height¶ – height of the images
enc_type¶ – option between resnet18 or resnet50
first_conv¶ – use standard kernel_size 7, stride 2 at start or replace it with kernel_size 3, stride 1 conv
maxpool1¶ – use standard maxpool to reduce spatial dim of feat by a factor of 2
enc_out_dim¶ – set according to the out_channel count of encoder used (512 for resnet18, 2048 for resnet50)
latent_dim¶ – dim of latent space
lr¶ – learning rate for Adam

Variational Autoencoders¶

Basic VAE¶

Use the VAE like so.

from pl_bolts.models.autoencoders import VAE

model = VAE()
trainer = Trainer()
trainer.fit(model)

You can override any part of this VAE to build your own variation.

from pl_bolts.models.autoencoders import VAE

class MyVAEFlavor(VAE):

    def get_posterior(self, mu, std):
        # do something other than the default
        # P = self.get_distribution(self.prior, loc=torch.zeros_like(mu), scale=torch.ones_like(std))

        return P

You can use the pretrained models present in bolts.

CIFAR-10 pretrained model:

from pl_bolts.models.autoencoders import VAE

vae = VAE(input_height=32)
print(VAE.pretrained_weights_available())
vae = vae.from_pretrained('cifar10-resnet18')

vae.freeze()

Tensorboard for VAE on CIFAR-10

Training:

Reconstructions:

Both input and generated images are normalized versions as the training was done with such images.

STL-10 pretrained model:

from pl_bolts.models.autoencoders import VAE

vae = VAE(input_height=96, first_conv=True)
print(VAE.pretrained_weights_available())
vae = vae.from_pretrained('cifar10-resnet18')

vae.freeze()

Tensorboard for VAE on STL-10

Training:

class pl_bolts.models.autoencoders.VAE(input_height, enc_type='resnet18', first_conv=False, maxpool1=False, enc_out_dim=512, kl_coeff=0.1, latent_dim=256, lr=0.0001, **kwargs)[source]

Bases: pytorch_lightning.LightningModule

Standard VAE with Gaussian Prior and approx posterior.

Model is available pretrained on different datasets:

Example:

# not pretrained
vae = VAE()

# pretrained on cifar10
vae = VAE.from_pretrained('cifar10-resnet18')

# pretrained on stl10
vae = VAE.from_pretrained('stl10-resnet18')

Parameters

input_height¶ – height of the images
enc_type¶ – option between resnet18 or resnet50
first_conv¶ – use standard kernel_size 7, stride 2 at start or replace it with kernel_size 3, stride 1 conv
maxpool1¶ – use standard maxpool to reduce spatial dim of feat by a factor of 2
enc_out_dim¶ – set according to the out_channel count of encoder used (512 for resnet18, 2048 for resnet50)
kl_coeff¶ – coefficient for kl term of the loss
latent_dim¶ – dim of latent space
lr¶ – learning rate for Adam

Convolutional Architectures¶

This package lists contributed convolutional architectures.

GPT-2¶

class pl_bolts.models.vision.image_gpt.gpt2.GPT2(embed_dim, heads, layers, num_positions, vocab_size, num_classes)[source]

Bases: pytorch_lightning.LightningModule

GPT-2 from language Models are Unsupervised Multitask Learners

Paper by: Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever

Implementation contributed by:

Teddy Koker

Example:

from pl_bolts.models import GPT2

seq_len = 17
batch_size = 32
vocab_size = 16
x = torch.randint(0, vocab_size, (seq_len, batch_size))
model = GPT2(embed_dim=32, heads=2, layers=2, num_positions=seq_len, vocab_size=vocab_size, num_classes=4)
results = model(x)

forward(x, classify=False)[source]: Expect input as shape [sequence len, batch] If classify, return classification logits

Image GPT¶

class pl_bolts.models.vision.image_gpt.igpt_module.ImageGPT(datamodule=None, embed_dim=16, heads=2, layers=2, pixels=28, vocab_size=16, num_classes=10, classify=False, batch_size=64, learning_rate=0.01, steps=25000, data_dir='.', num_workers=8, **kwargs)[source]

Bases: pytorch_lightning.LightningModule

Paper: Generative Pretraining from Pixels [original paper code].

Paper by: Mark Che, Alec Radford, Rewon Child, Jeff Wu, Heewoo Jun, Prafulla Dhariwal, David Luan, Ilya Sutskever

Implementation contributed by:

Teddy Koker

Original repo with results and more implementation details:

https://github.com/teddykoker/image-gpt

Example Results (Photo credits: Teddy Koker):

Default arguments:

Argument Defaults¶
Argument	Default	iGPT-S (Chen et al.)
–embed_dim	16	512
–heads	2	8
–layers	8	24
–pixels	28	32
–vocab_size	16	512
–num_classes	10	10
–batch_size	64	128
–learning_rate	0.01	0.01
–steps	25000	1000000

Example:

import pytorch_lightning as pl
from pl_bolts.models.vision import ImageGPT

dm = MNISTDataModule('.')
model = ImageGPT(dm)

pl.Trainer(gpu=4).fit(model)

As script:

cd pl_bolts/models/vision/image_gpt
python igpt_module.py --learning_rate 1e-2 --batch_size 32 --gpus 4

Parameters

datamodule¶ (Optional[LightningDataModule]) – LightningDataModule
embed_dim¶ (int) – the embedding dim
heads¶ (int) – number of attention heads
layers¶ (int) – number of layers
pixels¶ (int) – number of input pixels
vocab_size¶ (int) – vocab size
num_classes¶ (int) – number of classes in the input
classify¶ (bool) – true if should classify
batch_size¶ (int) – the batch size
learning_rate¶ (float) – learning rate
steps¶ (int) – number of steps for cosine annealing
data_dir¶ (str) – where to store data
num_workers¶ (int) – num_data workers

Pixel CNN¶

class pl_bolts.models.vision.pixel_cnn.PixelCNN(input_channels, hidden_channels=256, num_blocks=5)[source]

Bases: torch.nn.Module

Implementation of Pixel CNN.

Paper authors: Aaron van den Oord, Nal Kalchbrenner, Oriol Vinyals, Lasse Espeholt, Alex Graves, Koray Kavukcuoglu

Implemented by:

William Falcon

Example:

>>> from pl_bolts.models.vision import PixelCNN
>>> import torch
...
>>> model = PixelCNN(input_channels=3)
>>> x = torch.rand(5, 3, 64, 64)
>>> out = model(x)
...
>>> out.shape
torch.Size([5, 3, 64, 64])

GANs¶

Collection of Generative Adversarial Networks

Basic GAN¶

This is a vanilla GAN. This model can work on any dataset size but results are shown for MNIST. Replace the encoder, decoder or any part of the training loop to build a new method, or simply finetune on your data.

Implemented by:

William Falcon

Example outputs:

Loss curves:

from pl_bolts.models.gans import GAN
...
gan = GAN()
trainer = Trainer()
trainer.fit(gan)

class pl_bolts.models.gans.GAN(input_channels, input_height, input_width, latent_dim=32, learning_rate=0.0002, **kwargs)[source]

Bases: pytorch_lightning.LightningModule

Vanilla GAN implementation.

Example:

from pl_bolts.models.gan import GAN

m = GAN()
Trainer(gpus=2).fit(m)

Example CLI:

# mnist
python  basic_gan_module.py --gpus 1

# imagenet
python  basic_gan_module.py --gpus 1 --dataset 'imagenet2012'
--data_dir /path/to/imagenet/folder/ --meta_dir ~/path/to/meta/bin/folder
--batch_size 256 --learning_rate 0.0001

Parameters

datamodule¶ – the datamodule (train, val, test splits)
latent_dim¶ (int) – emb dim for encoder
batch_size¶ – the batch size
learning_rate¶ (float) – the learning rate
data_dir¶ – where to store data
num_workers¶ – data workers

forward(z)[source]

Generates an image given input noise z

Example:

z = torch.rand(batch_size, latent_dim)
gan = GAN.load_from_checkpoint(PATH)
img = gan(z)

Reinforcement Learning¶

This module is a collection of common RL approaches implemented in Lightning.

Module authors¶

Contributions by: Donal Byrne

DQN
Double DQN
Dueling DQN
Noisy DQN
NStep DQN
Prioritized Experience Replay DQN
Reinforce
Vanilla Policy Gradient

Note

RL models currently only support CPU and single GPU training with distributed_backend=dp. Full GPU support will be added in later updates.

DQN Models¶

The following models are based on DQN. DQN uses value based learning where it is deciding what action to take based on the model’s current learned value (V), or the state action value (Q) of the current state. These values are defined as the discounted total reward of the agents state or state action pair.

Deep-Q-Network (DQN)¶

DQN model introduced in Playing Atari with Deep Reinforcement Learning. Paper authors: Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, Martin Riedmiller.

Original implementation by: Donal Byrne

The DQN was introduced in Playing Atari with Deep Reinforcement Learning by researchers at DeepMind. This took the concept of tabular Q learning and scaled it to much larger problems by apporximating the Q function using a deep neural network.

The goal behind DQN was to take the simple control method of Q learning and scale it up in order to solve complicated tasks. As well as this, the method needed to be stable. The DQN solves these issues with the following additions.

Approximated Q Function

Storing Q values in a table works well in theory, but is completely unscalable. Instead, the authors approximate the Q function using a deep neural network. This allows the DQN to be used for much more complicated tasks

Replay Buffer

Similar to supervised learning, the DQN learns on randomly sampled batches of previous data stored in an Experience Replay Buffer. The ‘target’ is calculated using the Bellman equation

$Q(s,a)<-(r+{\gamma}\max_{a'{\in}A}Q(s',a'))^2$

and then we optimize using SGD just like a standard supervised learning problem.

$L=(Q(s,a)-(r+{\gamma}\max_{a'{\in}A}Q(s',a'))^2$

DQN Results¶

DQN: Pong

Example:

from pl_bolts.models.rl import DQN
dqn = DQN("PongNoFrameskip-v4")
trainer = Trainer()
trainer.fit(dqn)

class pl_bolts.models.rl.dqn_model.DQN(env, eps_start=1.0, eps_end=0.02, eps_last_frame=150000, sync_rate=1000, gamma=0.99, learning_rate=0.0001, batch_size=32, replay_size=100000, warm_start_size=10000, avg_reward_len=100, min_episode_reward=-21, seed=123, batches_per_epoch=1000, n_steps=1, **kwargs)[source]

Bases: pytorch_lightning.LightningModule

Basic DQN Model

PyTorch Lightning implementation of DQN Paper authors: Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, Martin Riedmiller. Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.dqn_model import DQN
...
>>> model = DQN("PongNoFrameskip-v4")

Train:

trainer = Trainer()
trainer.fit(model)

Parameters

env¶ (str) – gym environment tag
eps_start¶ (float) – starting value of epsilon for the epsilon-greedy exploration
eps_end¶ (float) – final value of epsilon for the epsilon-greedy exploration
eps_last_frame¶ (int) – the final frame in for the decrease of epsilon. At this frame espilon = eps_end
sync_rate¶ (int) – the number of iterations between syncing up the target network with the train network
gamma¶ (float) – discount factor
learning_rate¶ (float) – learning rate
batch_size¶ (int) – size of minibatch pulled from the DataLoader
replay_size¶ (int) – total capacity of the replay buffer
warm_start_size¶ (int) – how many random steps through the environment to be carried out at the start of training to fill the buffer with a starting point
avg_reward_len¶ (int) – how many episodes to take into account when calculating the avg reward
min_episode_reward¶ (int) – the minimum score that can be achieved in an episode. Used for filling the avg buffer before training begins
seed¶ (int) – seed value for all RNG used
batches_per_epoch¶ (int) – number of batches per epoch
n_steps¶ (int) – size of n step look ahead

Note

This example is based on: https://github.com/PacktPublishing/Deep-Reinforcement-Learning-Hands-On-Second-Edition/blob/master/Chapter06/02_dqn_pong.py

Note

Currently only supports CPU and single GPU training with distributed_backend=dp

_dataloader()[source]

Initialize the Replay Buffer dataset used for retrieving experiences

Return type: DataLoader

static add_model_specific_args(arg_parser)[source]

Adds arguments for DQN model Note: these params are fine tuned for Pong env :type _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.add_model_specific_args.arg_parser: ArgumentParser :param _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.add_model_specific_args.arg_parser: parent parser

Return type: ArgumentParser

build_networks()[source]

Initializes the DQN train and target networks

Return type: None

configure_optimizers()[source]

Initialize Adam optimizer

Return type: List[Optimizer]

forward(x)[source]

Passes in a state x through the network and gets the q_values of each action as an output :type _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.forward.x: Tensor :param _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.forward.x: environment state

Return type: Tensor
Returns: q values

static make_environment(env_name, seed=None)[source]

Initialise gym environment :type _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.make_environment.env_name: str :param _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.make_environment.env_name: environment name or tag :type _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.make_environment.seed: Optional[int] :param _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.make_environment.seed: value to seed the environment RNG for reproducibility

Return type: Env
Returns: gym environment

populate(warm_start)[source]

Populates the buffer with initial experience

Return type: None

run_n_episodes(env, n_epsiodes=1, epsilon=1.0)[source]

Carries out N episodes of the environment with the current agent :param _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.run_n_episodes.env: environment to use, either train environment or test environment :type _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.run_n_episodes.n_epsiodes: int :param _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.run_n_episodes.n_epsiodes: number of episodes to run :type _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.run_n_episodes.epsilon: float :param _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.run_n_episodes.epsilon: epsilon value for DQN agent

Return type: List[int]

test_dataloader()[source]

Get test loader

Return type: DataLoader

test_epoch_end(outputs)[source]

Log the avg of the test results

Return type: Dict[str, Tensor]

test_step(*args, **kwargs)[source]

Evaluate the agent for 10 episodes

Return type: Dict[str, Tensor]

train_batch()[source]: Contains the logic for generating a new batch of data to be passed to the DataLoader :rtype: Tuple[Tensor, Tensor, Tensor, Tensor, Tensor] :returns: yields a Experience tuple containing the state, action, reward, done and next_state.

train_dataloader()[source]

Get train loader

Return type: DataLoader

training_step(batch, _)[source]

Carries out a single step through the environment to update the replay buffer. Then calculates loss based on the minibatch recieved :type _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.training_step.batch: Tuple[Tensor, Tensor] :param _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.training_step.batch: current mini batch of replay data :param _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.training_step._: batch number, not used

Return type: OrderedDict
Returns: Training loss and log metrics

Double DQN¶

Double DQN model introduced in Deep Reinforcement Learning with Double Q-learning Paper authors: Hado van Hasselt, Arthur Guez, David Silver

Original implementation by: Donal Byrne

The original DQN tends to overestimate Q values during the Bellman update, leading to instability and is harmful to training. This is due to the max operation in the Bellman equation.

We are constantly taking the max of our agents estimates during our update. This may seem reasonable, if we could trust these estimates. However during the early stages of training, the estimates for these values will be off center and can lead to instability in training until our estimates become more reliable

The Double DQN fixes this overestimation by choosing actions for the next state using the main trained network but uses the values of these actions from the more stable target network. So we are still going to take the greedy action, but the value will be less “optimisitc” because it is chosen by the target network.

DQN expected return

$Q(s_t, a_t) = r_t + \gamma * \max_{Q'}(S_{t+1}, a)$

Double DQN expected return

$Q(s_t, a_t) = r_t + \gamma * \max{Q'}(S_{t+1}, \arg\max_Q(S_{t+1}, a))$

Double DQN Results¶

Double DQN: Pong

DQN vs Double DQN: Pong

orange: DQN

blue: Double DQN

Example:

from pl_bolts.models.rl import DoubleDQN
ddqn = DoubleDQN("PongNoFrameskip-v4")
trainer = Trainer()
trainer.fit(ddqn)

class pl_bolts.models.rl.double_dqn_model.DoubleDQN(env, eps_start=1.0, eps_end=0.02, eps_last_frame=150000, sync_rate=1000, gamma=0.99, learning_rate=0.0001, batch_size=32, replay_size=100000, warm_start_size=10000, avg_reward_len=100, min_episode_reward=-21, seed=123, batches_per_epoch=1000, n_steps=1, **kwargs)[source]

Bases: pl_bolts.models.rl.dqn_model.DQN

Double Deep Q-network (DDQN) PyTorch Lightning implementation of Double DQN

Paper authors: Hado van Hasselt, Arthur Guez, David Silver

Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.double_dqn_model import DoubleDQN
...
>>> model = DoubleDQN("PongNoFrameskip-v4")

Train:

trainer = Trainer()
trainer.fit(model)

Parameters

env¶ (str) – gym environment tag
gpus¶ – number of gpus being used
eps_start¶ (float) – starting value of epsilon for the epsilon-greedy exploration
eps_end¶ (float) – final value of epsilon for the epsilon-greedy exploration
eps_last_frame¶ (int) – the final frame in for the decrease of epsilon. At this frame espilon = eps_end
sync_rate¶ (int) – the number of iterations between syncing up the target network with the train network
gamma¶ (float) – discount factor
lr¶ – learning rate
batch_size¶ (int) – size of minibatch pulled from the DataLoader
replay_size¶ (int) – total capacity of the replay buffer
warm_start_size¶ (int) – how many random steps through the environment to be carried out at the start of training to fill the buffer with a starting point
sample_len¶ – the number of samples to pull from the dataset iterator and feed to the DataLoader

Note

This example is based on https://github.com/PacktPublishing/Deep-Reinforcement-Learning-Hands-On-Second-Edition/blob/master/Chapter08/03_dqn_double.py

Note

Currently only supports CPU and single GPU training with distributed_backend=dp

PyTorch Lightning implementation of DQN Paper authors: Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, Martin Riedmiller. Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.dqn_model import DQN
...
>>> model = DQN("PongNoFrameskip-v4")

Train:

trainer = Trainer()
trainer.fit(model)

Parameters

env¶ (str) – gym environment tag
eps_start¶ (float) – starting value of epsilon for the epsilon-greedy exploration
eps_end¶ (float) – final value of epsilon for the epsilon-greedy exploration
eps_last_frame¶ (int) – the final frame in for the decrease of epsilon. At this frame espilon = eps_end
sync_rate¶ (int) – the number of iterations between syncing up the target network with the train network
gamma¶ (float) – discount factor
learning_rate¶ (float) – learning rate
batch_size¶ (int) – size of minibatch pulled from the DataLoader
replay_size¶ (int) – total capacity of the replay buffer
warm_start_size¶ (int) – how many random steps through the environment to be carried out at the start of training to fill the buffer with a starting point
avg_reward_len¶ (int) – how many episodes to take into account when calculating the avg reward
min_episode_reward¶ (int) – the minimum score that can be achieved in an episode. Used for filling the avg buffer before training begins
seed¶ (int) – seed value for all RNG used
batches_per_epoch¶ (int) – number of batches per epoch
n_steps¶ (int) – size of n step look ahead

Note

This example is based on: https://github.com/PacktPublishing/Deep-Reinforcement-Learning-Hands-On-Second-Edition/blob/master/Chapter06/02_dqn_pong.py

Note

Currently only supports CPU and single GPU training with distributed_backend=dp

training_step(batch, _)[source]

Carries out a single step through the environment to update the replay buffer. Then calculates loss based on the minibatch recieved :type _sphinx_paramlinks_pl_bolts.models.rl.double_dqn_model.DoubleDQN.training_step.batch: Tuple[Tensor, Tensor] :param _sphinx_paramlinks_pl_bolts.models.rl.double_dqn_model.DoubleDQN.training_step.batch: current mini batch of replay data :param _sphinx_paramlinks_pl_bolts.models.rl.double_dqn_model.DoubleDQN.training_step._: batch number, not used

Return type: OrderedDict
Returns: Training loss and log metrics

Dueling DQN¶

Dueling DQN model introduced in Dueling Network Architectures for Deep Reinforcement Learning Paper authors: Ziyu Wang, Tom Schaul, Matteo Hessel, Hado van Hasselt, Marc Lanctot, Nando de Freitas

Original implementation by: Donal Byrne

The Q value that we are trying to approximate can be divided into two parts, the value state V(s) and the ‘advantage’ of actions in that state A(s, a). Instead of having one full network estimate the entire Q value, Dueling DQN uses two estimator heads in order to separate the estimation of the two parts.

The value is the same as in value iteration. It is the discounted expected reward achieved from state s. Think of the value as the ‘base reward’ from being in state s.

The advantage tells us how much ‘extra’ reward we get from taking action a while in state s. The advantage bridges the gap between Q(s, a) and V(s) as Q(s, a) = V(s) + A(s, a).

In the paper Dueling Network Architectures for Deep Reinforcement Learning <https://arxiv.org/abs/1511.06581> the network uses two heads, one outputs the value state and the other outputs the advantage. This leads to better training stability, faster convergence and overall better results. The V head outputs a single scalar (the state value), while the advantage head outputs a tensor equal to the size of the action space, containing an advantage value for each action in state s.

Changing the network architecture is not enough, we also need to ensure that the advantage mean is 0. This is done by subtracting the mean advantage from the Q value. This essentially pulls the mean advantage to 0.

$Q(s, a) = V(s) + A(s, a) - 1/N * \sum_k(A(s, k)$

Dueling DQN Benefits¶

Ability to efficiently learn the state value function. In the dueling network, every Q update also updates the value stream, where as in DQN only the value of the chosen action is updated. This provides a better approximation of the values
The differences between total Q values for a given state are quite small in relation to the magnitude of Q. The difference in the Q values between the best action and the second best action can be very small, while the average state value can be much larger. The differences in scale can introduce noise, which may lead to the greedy policy switching the priority of these actions. The seperate estimators for state value and advantage makes the Dueling DQN robust to this type of scenario

Dueling DQN Results¶

The results below a noticeable improvement from the original DQN network.

Dueling DQN baseline: Pong

Similar to the results of the DQN baseline, the agent has a period where the number of steps per episodes increase as it begins to hold its own against the heuristic oppoent, but then the steps per episode quickly begins to drop as it gets better and starts to beat its opponent faster and faster. There is a noticable point at step ~250k where the agent goes from losing to winning.

As you can see by the total rewards, the dueling network’s training progression is very stable and continues to trend upward until it finally plateus.

DQN vs Dueling DQN: Pong

In comparison to the base DQN, we see that the Dueling network’s training is much more stable and is able to reach a score in the high teens faster than the DQN agent. Even though the Dueling network is more stable and out performs DQN early in training, by the end of training the two networks end up at the same point.

This could very well be due to the simplicity of the Pong environment.

Orange: DQN

Red: Dueling DQN

Example:

from pl_bolts.models.rl import DuelingDQN
dueling_dqn = DuelingDQN("PongNoFrameskip-v4")
trainer = Trainer()
trainer.fit(dueling_dqn)

class pl_bolts.models.rl.dueling_dqn_model.DuelingDQN(env, eps_start=1.0, eps_end=0.02, eps_last_frame=150000, sync_rate=1000, gamma=0.99, learning_rate=0.0001, batch_size=32, replay_size=100000, warm_start_size=10000, avg_reward_len=100, min_episode_reward=-21, seed=123, batches_per_epoch=1000, n_steps=1, **kwargs)[source]

Bases: pl_bolts.models.rl.dqn_model.DQN

PyTorch Lightning implementation of Dueling DQN

Paper authors: Ziyu Wang, Tom Schaul, Matteo Hessel, Hado van Hasselt, Marc Lanctot, Nando de Freitas

Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.dueling_dqn_model import DuelingDQN
...
>>> model = DuelingDQN("PongNoFrameskip-v4")

Train:

trainer = Trainer()
trainer.fit(model)

Parameters

env¶ (str) – gym environment tag
gpus¶ – number of gpus being used
eps_start¶ (float) – starting value of epsilon for the epsilon-greedy exploration
eps_end¶ (float) – final value of epsilon for the epsilon-greedy exploration
eps_last_frame¶ (int) – the final frame in for the decrease of epsilon. At this frame espilon = eps_end
sync_rate¶ (int) – the number of iterations between syncing up the target network with the train network
gamma¶ (float) – discount factor
lr¶ – learning rate
batch_size¶ (int) – size of minibatch pulled from the DataLoader
replay_size¶ (int) – total capacity of the replay buffer
warm_start_size¶ (int) – how many random steps through the environment to be carried out at the start of training to fill the buffer with a starting point
sample_len¶ – the number of samples to pull from the dataset iterator and feed to the DataLoader

Note

Currently only supports CPU and single GPU training with distributed_backend=dp

PyTorch Lightning implementation of DQN Paper authors: Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, Martin Riedmiller. Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.dqn_model import DQN
...
>>> model = DQN("PongNoFrameskip-v4")

Train:

trainer = Trainer()
trainer.fit(model)

Parameters

env¶ (str) – gym environment tag
eps_start¶ (float) – starting value of epsilon for the epsilon-greedy exploration
eps_end¶ (float) – final value of epsilon for the epsilon-greedy exploration
eps_last_frame¶ (int) – the final frame in for the decrease of epsilon. At this frame espilon = eps_end
sync_rate¶ (int) – the number of iterations between syncing up the target network with the train network
gamma¶ (float) – discount factor
learning_rate¶ (float) – learning rate
batch_size¶ (int) – size of minibatch pulled from the DataLoader
replay_size¶ (int) – total capacity of the replay buffer
warm_start_size¶ (int) – how many random steps through the environment to be carried out at the start of training to fill the buffer with a starting point
avg_reward_len¶ (int) – how many episodes to take into account when calculating the avg reward
min_episode_reward¶ (int) – the minimum score that can be achieved in an episode. Used for filling the avg buffer before training begins
seed¶ (int) – seed value for all RNG used
batches_per_epoch¶ (int) – number of batches per epoch
n_steps¶ (int) – size of n step look ahead

Note

This example is based on: https://github.com/PacktPublishing/Deep-Reinforcement-Learning-Hands-On-Second-Edition/blob/master/Chapter06/02_dqn_pong.py

Note

Currently only supports CPU and single GPU training with distributed_backend=dp

build_networks()[source]

Initializes the Dueling DQN train and target networks

Return type: None

Noisy DQN¶

Noisy DQN model introduced in Noisy Networks for Exploration Paper authors: Meire Fortunato, Mohammad Gheshlaghi Azar, Bilal Piot, Jacob Menick, Ian Osband, Alex Graves, Vlad Mnih, Remi Munos, Demis Hassabis, Olivier Pietquin, Charles Blundell, Shane Legg

Original implementation by: Donal Byrne

Up until now the DQN agent uses a seperate exploration policy, generally epsilon-greedy where start and end values are set for its exploration. Noisy Networks For Exploration <https://arxiv.org/abs/1706.10295> introduces a new exploration strategy by adding noise parameters to the weights of the fully connect layers which get updated during backpropagation of the network. The noise parameters drive the exploration of the network instead of simply taking random actions more frequently at the start of training and less frequently towards the end. The of authors of propose two ways of doing this.

During the optimization step a new set of noisy parameters are sampled. During training the agent acts according to the fixed set of parameters. At the next optimization step, the parameters are updated with a new sample. This ensures the agent always acts based on the parameters that are drawn from the current noise distribution.

The authors propose two methods of injecting noise to the network.

Independent Gaussian Noise: This injects noise per weight. For each weight a random value is taken from the distribution. Noise parameters are stored inside the layer and are updated during backpropagation. The output of the layer is calculated as normal.
Factorized Gaussian Noise: This injects nosier per input/ouput. In order to minimize the number of random values this method stores two random vectors, one with the size of the input and the other with the size of the output. Using these two vectors, a random matrix is generated for the layer by calculating the outer products of the vector

Noisy DQN Benefits¶

Improved exploration function. Instead of just performing completely random actions, we add decreasing amount of noise and uncertainty to our policy allowing to explore while still utilising its policy.
The fact that this method is automatically tuned means that we do not have to tune hyper parameters for epsilon-greedy!

Note

For now I have just implemented the Independant Gaussian as it has been reported there isn’t much difference in results for these benchmark environments.

In order to update the basic DQN to a Noisy DQN we need to do the following

Noisy DQN Results¶

The results below improved stability and faster performance growth.

Noisy DQN baseline: Pong

Similar to the other improvements, the average score of the agent reaches positive numbers around the 250k mark and steadily increases till convergence.

DQN vs Dueling DQN: Pong

In comparison to the base DQN, the Noisy DQN is more stable and is able to converge on an optimal policy much faster than the original. It seems that the replacement of the epsilon-greedy strategy with network noise provides a better form of exploration.

Orange: DQN
Red: Noisy DQN

Example:

from pl_bolts.models.rl import NoisyDQN
noisy_dqn = NoisyDQN("PongNoFrameskip-v4")
trainer = Trainer()
trainer.fit(noisy_dqn)

class pl_bolts.models.rl.noisy_dqn_model.NoisyDQN(env, eps_start=1.0, eps_end=0.02, eps_last_frame=150000, sync_rate=1000, gamma=0.99, learning_rate=0.0001, batch_size=32, replay_size=100000, warm_start_size=10000, avg_reward_len=100, min_episode_reward=-21, seed=123, batches_per_epoch=1000, n_steps=1, **kwargs)[source]

Bases: pl_bolts.models.rl.dqn_model.DQN

PyTorch Lightning implementation of Noisy DQN

Paper authors: Meire Fortunato, Mohammad Gheshlaghi Azar, Bilal Piot, Jacob Menick, Ian Osband, Alex Graves, Vlad Mnih, Remi Munos, Demis Hassabis, Olivier Pietquin, Charles Blundell, Shane Legg

Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.noisy_dqn_model import NoisyDQN
...
>>> model = NoisyDQN("PongNoFrameskip-v4")

Train:

trainer = Trainer()
trainer.fit(model)

Parameters

env¶ (str) – gym environment tag
gpus¶ – number of gpus being used
eps_start¶ (float) – starting value of epsilon for the epsilon-greedy exploration
eps_end¶ (float) – final value of epsilon for the epsilon-greedy exploration
eps_last_frame¶ (int) – the final frame in for the decrease of epsilon. At this frame espilon = eps_end
sync_rate¶ (int) – the number of iterations between syncing up the target network with the train network
gamma¶ (float) – discount factor
lr¶ – learning rate
batch_size¶ (int) – size of minibatch pulled from the DataLoader
replay_size¶ (int) – total capacity of the replay buffer
warm_start_size¶ (int) – how many random steps through the environment to be carried out at the start of
to fill the buffer with a starting point¶ (training) –
sample_len¶ – the number of samples to pull from the dataset iterator and feed to the DataLoader

Note

Currently only supports CPU and single GPU training with distributed_backend=dp

PyTorch Lightning implementation of DQN Paper authors: Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, Martin Riedmiller. Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.dqn_model import DQN
...
>>> model = DQN("PongNoFrameskip-v4")

Train:

trainer = Trainer()
trainer.fit(model)

Parameters

env¶ (str) – gym environment tag
eps_start¶ (float) – starting value of epsilon for the epsilon-greedy exploration
eps_end¶ (float) – final value of epsilon for the epsilon-greedy exploration
eps_last_frame¶ (int) – the final frame in for the decrease of epsilon. At this frame espilon = eps_end
sync_rate¶ (int) – the number of iterations between syncing up the target network with the train network
gamma¶ (float) – discount factor
learning_rate¶ (float) – learning rate
batch_size¶ (int) – size of minibatch pulled from the DataLoader
replay_size¶ (int) – total capacity of the replay buffer
warm_start_size¶ (int) – how many random steps through the environment to be carried out at the start of training to fill the buffer with a starting point
avg_reward_len¶ (int) – how many episodes to take into account when calculating the avg reward
min_episode_reward¶ (int) – the minimum score that can be achieved in an episode. Used for filling the avg buffer before training begins
seed¶ (int) – seed value for all RNG used
batches_per_epoch¶ (int) – number of batches per epoch
n_steps¶ (int) – size of n step look ahead

Note

This example is based on: https://github.com/PacktPublishing/Deep-Reinforcement-Learning-Hands-On-Second-Edition/blob/master/Chapter06/02_dqn_pong.py

Note

Currently only supports CPU and single GPU training with distributed_backend=dp

build_networks()[source]

Initializes the Noisy DQN train and target networks

Return type: None

on_train_start()[source]

Set the agents epsilon to 0 as the exploration comes from the network

Return type: None

train_batch()[source]: Contains the logic for generating a new batch of data to be passed to the DataLoader. This is the same function as the standard DQN except that we dont update epsilon as it is always 0. The exploration comes from the noisy network. :rtype: Tuple[Tensor, Tensor, Tensor, Tensor, Tensor] :returns: yields a Experience tuple containing the state, action, reward, done and next_state.

N-Step DQN¶

N-Step DQN model introduced in Learning to Predict by the Methods of Temporal Differences Paper authors: Richard S. Sutton

Original implementation by: Donal Byrne

N Step DQN was introduced in Learning to Predict by the Methods of Temporal Differences. This method improves upon the original DQN by updating our Q values with the expected reward from multiple steps in the future as opposed to the expected reward from the immediate next state. When getting the Q values for a state action pair using a single step which looks like this

$Q(s_t,a_t)=r_t+{\gamma}\max_aQ(s_{t+1},a_{t+1})$

but because the Q function is recursive we can continue to roll this out into multiple steps, looking at the expected return for each step into the future.

$Q(s_t,a_t)=r_t+{\gamma}r_{t+1}+{\gamma}^2\max_{a'}Q(s_{t+2},a')$

The above example shows a 2-Step look ahead, but this could be rolled out to the end of the episode, which is just Monte Carlo learning. Although we could just do a monte carlo update and look forward to the end of the episode, it wouldn’t be a good idea. Every time we take another step into the future, we are basing our approximation off our current policy. For a large portion of training, our policy is going to be less than optimal. For example, at the start of training, our policy will be in a state of high exploration, and will be little better than random.

Note

For each rollout step you must scale the discount factor accordingly by the number of steps. As you can see from the equation above, the second gamma value is to the power of 2. If we rolled this out one step further, we would use gamma to the power of 3 and so.

So if we are aproximating future rewards off a bad policy, chances are those approximations are going to be pretty bad and every time we unroll our update equation, the worse it will get. The fact that we are using an off policy method like DQN with a large replay buffer will make this even worse, as there is a high chance that we will be training on experiences using an old policy that was worse than our current policy.

So we need to strike a balance between looking far enough ahead to improve the convergence of our agent, but not so far that are updates become unstable. In general, small values of 2-4 work best.

N-Step Benefits¶

Multi-Step learning is capable of learning faster than typical 1 step learning methods.
Note that this method introduces a new hyperparameter n. Although n=4 is generally a good starting point and provides good results across the board.

N-Step Results¶

As expected, the N-Step DQN converges much faster than the standard DQN, however it also adds more instability to the loss of the agent. This can be seen in the following experiments.

N-Step DQN: Pong

The N-Step DQN shows the greatest increase in performance with respect to the other DQN variations. After less than 150k steps the agent begins to consistently win games and achieves the top score after ~170K steps. This is reflected in the sharp peak of the total episode steps and of course, the total episode rewards.

DQN vs N-Step DQN: Pong

This improvement is shown in stark contrast to the base DQN, which only begins to win games after 250k steps and requires over twice as many steps (450k) as the N-Step agent to achieve the high score of 21. One important thing to notice is the large increase in the loss of the N-Step agent. This is expected as the agent is building its expected reward off approximations of the future states. The large the size of N, the greater the instability. Previous literature, listed below, shows the best results for the Pong environment with an N step between 3-5. For these experiments I opted with an N step of 4.

Example:

from pl_bolts.models.rl import DQN
n_step_dqn = DQN("PongNoFrameskip-v4", n_steps=4)
trainer = Trainer()
trainer.fit(n_step_dqn)

Prioritized Experience Replay DQN¶

Double DQN model introduced in Prioritized Experience Replay Paper authors: Tom Schaul, John Quan, Ioannis Antonoglou, David Silver

Original implementation by: Donal Byrne

The standard DQN uses a buffer to break up the correlation between experiences and uniform random samples for each batch. Instead of just randomly sampling from the buffer prioritized experience replay (PER) prioritizes these samples based on training loss. This concept was introduced in the paper Prioritized Experience Replay

Essentially we want to train more on the samples that sunrise the agent.

The priority of each sample is defined below where

$P(i) = P^\alpha_i / \sum_k P_k^\alpha$

where pi is the priority of the ith sample in the buffer and 𝛼 is the number that shows how much emphasis we give to the priority. If 𝛼 = 0 , our sampling will become uniform as in the classic DQN method. Larger values for 𝛼 put more stress on samples with higher priority

Its important that new samples are set to the highest priority so that they are sampled soon. This however introduces bias to new samples in our dataset. In order to compensate for this bias, the value of the weight is defined as

$w_i=(N . P(i))^{-\beta}$

Where beta is a hyper parameter between 0-1. When beta is 1 the bias is fully compensated. However authors noted that in practice it is better to start beta with a small value near 0 and slowly increase it to 1.

PER Benefits¶

The benefits of this technique are that the agent sees more samples that it struggled with and gets more chances to improve upon it.

Memory Buffer

First step is to replace the standard experience replay buffer with the prioritized experience replay buffer. This is pretty large (100+ lines) so I wont go through it here. There are two buffers implemented. The first is a naive list based buffer found in memory.PERBuffer and the second is more efficient buffer using a Sum Tree datastructure.

The list based version is simpler, but has a sample complexity of O(N). The Sum Tree in comparison has a complexity of O(1) for sampling and O(logN) for updating priorities.

Update loss function

The next thing we do is to use the sample weights that we get from PER. Add the following code to the end of the loss function. This applies the weights of our sample to the batch loss. Then we return the mean loss and weighted loss for each datum, with the addition of a small epsilon value.

PER Results¶

The results below show improved stability and faster performance growth.

PER DQN: Pong

Similar to the other improvements, we see that PER improves the stability of the agents training and shows to converged on an optimal policy faster.

DQN vs PER DQN: Pong

In comparison to the base DQN, the PER DQN does show improved stability and performance. As expected, the loss of the PER DQN is siginificantly lower. This is the main objective of PER by focusing on experiences with high loss.

It is important to note that loss is not the only metric we should be looking at. Although the agent may have very low loss during training, it may still perform poorly due to lack of exploration.

Orange: DQN
Pink: PER DQN

Example:

from pl_bolts.models.rl import PERDQN
per_dqn = PERDQN("PongNoFrameskip-v4")
trainer = Trainer()
trainer.fit(per_dqn)

class pl_bolts.models.rl.per_dqn_model.PERDQN(env, eps_start=1.0, eps_end=0.02, eps_last_frame=150000, sync_rate=1000, gamma=0.99, learning_rate=0.0001, batch_size=32, replay_size=100000, warm_start_size=10000, avg_reward_len=100, min_episode_reward=-21, seed=123, batches_per_epoch=1000, n_steps=1, **kwargs)[source]

Bases: pl_bolts.models.rl.dqn_model.DQN

PyTorch Lightning implementation of DQN With Prioritized Experience Replay

Paper authors: Tom Schaul, John Quan, Ioannis Antonoglou, David Silver

Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.per_dqn_model import PERDQN
...
>>> model = PERDQN("PongNoFrameskip-v4")

Train:

trainer = Trainer()
trainer.fit(model)

Args:
    env: gym environment tag
    gpus: number of gpus being used
    eps_start: starting value of epsilon for the epsilon-greedy exploration
    eps_end: final value of epsilon for the epsilon-greedy exploration
    eps_last_frame: the final frame in for the decrease of epsilon. At this frame espilon = eps_end
    sync_rate: the number of iterations between syncing up the target network with the train network
    gamma: discount factor
    learning_rate: learning rate
    batch_size: size of minibatch pulled from the DataLoader
    replay_size: total capacity of the replay buffer
    warm_start_size: how many random steps through the environment to be carried out at the start of
        training to fill the buffer with a starting point
    num_samples: the number of samples to pull from the dataset iterator and feed to the DataLoader

.. note::
    This example is based on:
     https://github.com/PacktPublishing/Deep-Reinforcement-Learning-Hands-On-Second-Edition/blob/master/Chapter08/05_dqn_prio_replay.py

.. note:: Currently only supports CPU and single GPU training with `distributed_backend=dp`

PyTorch Lightning implementation of DQN Paper authors: Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, Martin Riedmiller. Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.dqn_model import DQN
...
>>> model = DQN("PongNoFrameskip-v4")

Train:

trainer = Trainer()
trainer.fit(model)

Parameters

env¶ (str) – gym environment tag
eps_start¶ (float) – starting value of epsilon for the epsilon-greedy exploration
eps_end¶ (float) – final value of epsilon for the epsilon-greedy exploration
eps_last_frame¶ (int) – the final frame in for the decrease of epsilon. At this frame espilon = eps_end
sync_rate¶ (int) – the number of iterations between syncing up the target network with the train network
gamma¶ (float) – discount factor
learning_rate¶ (float) – learning rate
batch_size¶ (int) – size of minibatch pulled from the DataLoader
replay_size¶ (int) – total capacity of the replay buffer
warm_start_size¶ (int) – how many random steps through the environment to be carried out at the start of training to fill the buffer with a starting point
avg_reward_len¶ (int) – how many episodes to take into account when calculating the avg reward
min_episode_reward¶ (int) – the minimum score that can be achieved in an episode. Used for filling the avg buffer before training begins
seed¶ (int) – seed value for all RNG used
batches_per_epoch¶ (int) – number of batches per epoch
n_steps¶ (int) – size of n step look ahead

Note

This example is based on: https://github.com/PacktPublishing/Deep-Reinforcement-Learning-Hands-On-Second-Edition/blob/master/Chapter06/02_dqn_pong.py

Note

Currently only supports CPU and single GPU training with distributed_backend=dp

_dataloader()[source]

Initialize the Replay Buffer dataset used for retrieving experiences

Return type: DataLoader

train_batch()[source]: Contains the logic for generating a new batch of data to be passed to the DataLoader :rtype: Tuple[Tensor, Tensor, Tensor, Tensor, Tensor] :returns: yields a Experience tuple containing the state, action, reward, done and next_state.

training_step(batch, _)[source]

Carries out a single step through the environment to update the replay buffer. Then calculates loss based on the minibatch recieved :param _sphinx_paramlinks_pl_bolts.models.rl.per_dqn_model.PERDQN.training_step.batch: current mini batch of replay data :param _sphinx_paramlinks_pl_bolts.models.rl.per_dqn_model.PERDQN.training_step._: batch number, not used

Return type: OrderedDict
Returns: Training loss and log metrics

Policy Gradient Models¶

The following models are based on Policy Gradients. Unlike the Q learning models shown before, Policy based models do not try and learn the specifc values of state or state action pairs. Instead it cuts out the middle man and directly learns the policy distribution. In Policy Gradient models we update our network parameters in the direction suggested by our policy gradient in order to find a policy that produces the highest results.

Policy Gradient Key Points:

Outputs a distribution of actions instead of discrete Q values
Optimizes the policy directly, instead of indirectly through the optimization of Q values
The policy distribution of actions allows the model to handle more complex action spaces, such as continuous actions
The policy distribution introduces stochasticity, providing natural exploration to the model
The policy distribution provides a more stable update as a change in weights will only change the total distribution slightly, as opposed to changing weights based on the Q value of state S will change all Q values with similar states.
Policy gradients tend to converge faste, however they are not as sample efficient and generally require more interactions with the environment.

REINFORCE¶

REINFORCE model introduced in Policy Gradient Methods For Reinforcement Learning With Function Approximation Paper authors: Richard S. Sutton, David McAllester, Satinder Singh, Yishay Mansour

Original implementation by: Donal Byrne

REINFORCE is one of the simplest forms of the Policy Gradient method of RL. This method uses a Monte Carlo rollout, where its steps through entire episodes of the environment to build up trajectories computing the total rewards. The algorithm is as follows:

Initialize our network.
Play N full episodes saving the transitions through the environment.
For every step t in each episode k we calculate the discounted reward of the subsequent steps.

$Q_{k,t} = \sum_{i=0}\gamma^i r_i$

Calculate the loss for all transitions.

$L = - \sum_{k,t} Q_{k,t} \log(\pi(S_{k,t}, A_{k,t}))$

Perform SGD on the loss and repeat.

What this loss function is saying is simply that we want to take the log probability of action A at state S given our policy (network output). This is then scaled by the discounted reward that we calculated in the previous step. We then take the negative of our sum. This is because the loss is minimized during SGD, but we want to maximize our policy.

Note

The current implementation does not actually wait for the batch episodes the complete every time as we pass in a fixed batch size. For the time being we simply use a large batch size to accomodate this. This approach still works well for simple tasks as it still manages to get an accurate Q value by using a large batch size, but it is not as accurate or completely correct. This will be updated in a later version.

REINFORCE Benefits¶

Simple and straightforward
Computationally more efficient for simple tasks such as Cartpole than the Value Based methods.

REINFORCE Results¶

Hyperparameters:

Batch Size: 800
Learning Rate: 0.01
Episodes Per Batch: 4
Gamma: 0.99

TODO: Add results graph

Example:

from pl_bolts.models.rl import Reinforce
reinforce = Reinforce("CartPole-v0")
trainer = Trainer()
trainer.fit(reinforce)

class pl_bolts.models.rl.reinforce_model.Reinforce(env, gamma=0.99, lr=0.01, batch_size=8, n_steps=10, avg_reward_len=100, entropy_beta=0.01, epoch_len=1000, num_batch_episodes=4, **kwargs)[source]

Bases: pytorch_lightning.LightningModule

PyTorch Lightning implementation of REINFORCE Paper authors: Richard S. Sutton, David McAllester, Satinder Singh, Yishay Mansour Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.reinforce_model import Reinforce
...
>>> model = Reinforce("CartPole-v0")

Train:

trainer = Trainer()
trainer.fit(model)

Parameters

env¶ (str) – gym environment tag
gamma¶ (float) – discount factor
lr¶ (float) – learning rate
batch_size¶ (int) – size of minibatch pulled from the DataLoader
n_steps¶ (int) – number of stakes per discounted experience
entropy_beta¶ (float) – entropy coefficient
epoch_len¶ (int) – how many batches before pseudo epoch
num_batch_episodes¶ (int) – how many episodes to rollout for each batch of training
avg_reward_len¶ (int) – how many episodes to take into account when calculating the avg reward

Note

This example is based on: https://github.com/PacktPublishing/Deep-Reinforcement-Learning-Hands-On-Second-Edition/blob/master/Chapter11/02_cartpole_reinforce.py

Note

Currently only supports CPU and single GPU training with distributed_backend=dp

_dataloader()[source]

Initialize the Replay Buffer dataset used for retrieving experiences

Return type: DataLoader

static add_model_specific_args(arg_parser)[source]

Adds arguments for DQN model Note: these params are fine tuned for Pong env :param _sphinx_paramlinks_pl_bolts.models.rl.reinforce_model.Reinforce.add_model_specific_args.arg_parser: the current argument parser to add to

Return type: ArgumentParser
Returns: arg_parser with model specific cargs added

calc_qvals(rewards)[source]

Calculate the discounted rewards of all rewards in list :type _sphinx_paramlinks_pl_bolts.models.rl.reinforce_model.Reinforce.calc_qvals.rewards: List[float] :param _sphinx_paramlinks_pl_bolts.models.rl.reinforce_model.Reinforce.calc_qvals.rewards: list of rewards from latest batch

Return type: List[float]
Returns: list of discounted rewards

configure_optimizers()[source]

Initialize Adam optimizer

Return type: List[Optimizer]

discount_rewards(experiences)[source]

Calculates the discounted reward over N experiences :type _sphinx_paramlinks_pl_bolts.models.rl.reinforce_model.Reinforce.discount_rewards.experiences: Tuple[Experience] :param _sphinx_paramlinks_pl_bolts.models.rl.reinforce_model.Reinforce.discount_rewards.experiences: Tuple of Experience

Return type: float
Returns: total discounted reward

forward(x)[source]

Passes in a state x through the network and gets the q_values of each action as an output :type _sphinx_paramlinks_pl_bolts.models.rl.reinforce_model.Reinforce.forward.x: Tensor :param _sphinx_paramlinks_pl_bolts.models.rl.reinforce_model.Reinforce.forward.x: environment state

Return type: Tensor
Returns: q values

get_device(batch)[source]

Retrieve device currently being used by minibatch

Return type: str

train_batch()[source]

Contains the logic for generating a new batch of data to be passed to the DataLoader :Yields: yields a tuple of Lists containing tensors for states, actions and rewards of the batch.

Return type: Tuple[List[Tensor], List[Tensor], List[Tensor]]

train_dataloader()[source]

Get train loader

Return type: DataLoader

training_step(batch, _)[source]

Carries out a single step through the environment to update the replay buffer. Then calculates loss based on the minibatch recieved :type _sphinx_paramlinks_pl_bolts.models.rl.reinforce_model.Reinforce.training_step.batch: Tuple[Tensor, Tensor] :param _sphinx_paramlinks_pl_bolts.models.rl.reinforce_model.Reinforce.training_step.batch: current mini batch of replay data :param _sphinx_paramlinks_pl_bolts.models.rl.reinforce_model.Reinforce.training_step._: batch number, not used

Return type: OrderedDict
Returns: Training loss and log metrics

Vanilla Policy Gradient¶

Vanilla Policy Gradient model introduced in Policy Gradient Methods For Reinforcement Learning With Function Approximation Paper authors: Richard S. Sutton, David McAllester, Satinder Singh, Yishay Mansour

Original implementation by: Donal Byrne

Vanilla Policy Gradient (VPG) expands upon the REINFORCE algorithm and improves some of its major issues. The major issue with REINFORCE is that it has high variance. This can be improved by subtracting a baseline value from the Q values. For this implementation we use the average reward as our baseline.

Although Policy Gradients are able to explore naturally due to the stochastic nature of the model, the agent can still frequently be stuck in a local optima. In order to improve this, VPG adds an entropy term to improve exploration.

$H(\pi) = - \sum \pi (a | s) \log \pi (a | s)$

To further control the amount of additional entropy in our model we scale the entropy term by a small beta value. The scaled entropy is then subtracted from the policy loss.

VPG Benefits¶

Addition of the baseline reduces variance in the model
Improved exploration due to entropy bonus

VPG Results¶

Hyperparameters:

Batch Size: 8
Learning Rate: 0.001
N Steps: 10
N environments: 4
Entropy Beta: 0.01
Gamma: 0.99

Example:

from pl_bolts.models.rl import VanillaPolicyGradient
vpg = VanillaPolicyGradient("CartPole-v0")
trainer = Trainer()
trainer.fit(vpg)

class pl_bolts.models.rl.vanilla_policy_gradient_model.VanillaPolicyGradient(env, gamma=0.99, lr=0.01, batch_size=8, n_steps=10, avg_reward_len=100, entropy_beta=0.01, epoch_len=1000, **kwargs)[source]

Bases: pytorch_lightning.LightningModule

PyTorch Lightning implementation of Vanilla Policy Gradient Paper authors: Richard S. Sutton, David McAllester, Satinder Singh, Yishay Mansour Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.vanilla_policy_gradient_model import VanillaPolicyGradient
...
>>> model = VanillaPolicyGradient("CartPole-v0")

Train::: trainer = Trainer() trainer.fit(model)

Parameters

env¶ (str) – gym environment tag
gamma¶ (float) – discount factor
lr¶ (float) – learning rate
batch_size¶ (int) – size of minibatch pulled from the DataLoader
batch_episodes¶ – how many episodes to rollout for each batch of training
entropy_beta¶ (float) – dictates the level of entropy per batch
avg_reward_len¶ (int) – how many episodes to take into account when calculating the avg reward

Note

This example is based on: https://github.com/PacktPublishing/Deep-Reinforcement-Learning-Hands-On-Second-Edition/blob/master/Chapter11/04_cartpole_pg.py

Note

Currently only supports CPU and single GPU training with distributed_backend=dp

_dataloader()[source]

Initialize the Replay Buffer dataset used for retrieving experiences

Return type: DataLoader

static add_model_specific_args(arg_parser)[source]

Adds arguments for DQN model Note: these params are fine tuned for Pong env :param _sphinx_paramlinks_pl_bolts.models.rl.vanilla_policy_gradient_model.VanillaPolicyGradient.add_model_specific_args.arg_parser: the current argument parser to add to

Return type: ArgumentParser
Returns: arg_parser with model specific cargs added

compute_returns(rewards)[source]

Calculate the discounted rewards of the batched rewards

Parameters: rewards¶ – list of batched rewards
Returns: list of discounted rewards

configure_optimizers()[source]

Initialize Adam optimizer

Return type: List[Optimizer]

forward(x)[source]

Passes in a state x through the network and gets the q_values of each action as an output :type _sphinx_paramlinks_pl_bolts.models.rl.vanilla_policy_gradient_model.VanillaPolicyGradient.forward.x: Tensor :param _sphinx_paramlinks_pl_bolts.models.rl.vanilla_policy_gradient_model.VanillaPolicyGradient.forward.x: environment state

Return type: Tensor
Returns: q values

get_device(batch)[source]

Retrieve device currently being used by minibatch

Return type: str

loss(states, actions, scaled_rewards)[source]

Calculates the loss for VPG

Parameters

states¶ – batched states
actions¶ – batch actions
scaled_rewards¶ – batche Q values

Return type

Tensor

Returns

loss for the current batch

train_batch()[source]: Contains the logic for generating a new batch of data to be passed to the DataLoader :rtype: Tuple[List[Tensor], List[Tensor], List[Tensor]] :returns: yields a tuple of Lists containing tensors for states, actions and rewards of the batch.

train_dataloader()[source]

Get train loader

Return type: DataLoader

training_step(batch, _)[source]

Carries out a single step through the environment to update the replay buffer. Then calculates loss based on the minibatch recieved :type _sphinx_paramlinks_pl_bolts.models.rl.vanilla_policy_gradient_model.VanillaPolicyGradient.training_step.batch: Tuple[Tensor, Tensor] :param _sphinx_paramlinks_pl_bolts.models.rl.vanilla_policy_gradient_model.VanillaPolicyGradient.training_step.batch: current mini batch of replay data :param _sphinx_paramlinks_pl_bolts.models.rl.vanilla_policy_gradient_model.VanillaPolicyGradient.training_step._: batch number, not used

Return type: OrderedDict
Returns: Training loss and log metrics

Self-supervised Learning¶

This bolts module houses a collection of all self-supervised learning models.

Self-supervised learning extracts representations of an input by solving a pretext task. In this package, we implement many of the current state-of-the-art self-supervised algorithms.

Self-supervised models are trained with unlabeled datasets

Use cases¶

Here are some use cases for the self-supervised package.

Extracting image features¶

The models in this module are trained unsupervised and thus can capture better image representations (features).

In this example, we’ll load a resnet 18 which was pretrained on imagenet using CPC as the pretext task.

Example:

from pl_bolts.models.self_supervised import CPCV2

# load resnet18 pretrained using CPC on imagenet
model = CPCV2(pretrained='resnet18')
cpc_resnet18 = model.encoder
cpc_resnet18.freeze()

# it supports any torchvision resnet
model = CPCV2(pretrained='resnet50')

This means you can now extract image representations that were pretrained via unsupervised learning.

Example:

my_dataset = SomeDataset()
for batch in my_dataset:
    x, y = batch
    out = cpc_resnet18(x)

Train with unlabeled data¶

These models are perfect for training from scratch when you have a huge set of unlabeled images

from pl_bolts.models.self_supervised import SimCLR
from pl_bolts.models.self_supervised.simclr import SimCLREvalDataTransform, SimCLRTrainDataTransform


train_dataset = MyDataset(transforms=SimCLRTrainDataTransform())
val_dataset = MyDataset(transforms=SimCLREvalDataTransform())

# simclr needs a lot of compute!
model = SimCLR()
trainer = Trainer(tpu_cores=128)
trainer.fit(
    model,
    DataLoader(train_dataset),
    DataLoader(val_dataset),
)

Research¶

Mix and match any part, or subclass to create your own new method

from pl_bolts.models.self_supervised import CPCV2
from pl_bolts.losses.self_supervised_learning import FeatureMapContrastiveTask

amdim_task = FeatureMapContrastiveTask(comparisons='01, 11, 02', bidirectional=True)
model = CPCV2(contrastive_task=amdim_task)

Contrastive Learning Models¶

Contrastive self-supervised learning (CSL) is a self-supervised learning approach where we generate representations of instances such that similar instances are near each other and far from dissimilar ones. This is often done by comparing triplets of positive, anchor and negative representations.

In this section, we list Lightning implementations of popular contrastive learning approaches.

AMDIM¶

class pl_bolts.models.self_supervised.AMDIM(datamodule='cifar10', encoder='amdim_encoder', contrastive_task=torch.nn.Module, image_channels=3, image_height=32, encoder_feature_dim=320, embedding_fx_dim=1280, conv_block_depth=10, use_bn=False, tclip=20.0, learning_rate=0.0002, data_dir='', num_classes=10, batch_size=200, **kwargs)[source]

Bases: pytorch_lightning.LightningModule

PyTorch Lightning implementation of Augmented Multiscale Deep InfoMax (AMDIM)

Paper authors: Philip Bachman, R Devon Hjelm, William Buchwalter.

Model implemented by: William Falcon

This code is adapted to Lightning using the original author repo (the original repo).

Example

>>> from pl_bolts.models.self_supervised import AMDIM
...
>>> model = AMDIM(encoder='resnet18')

Train:

trainer = Trainer()
trainer.fit(model)

Parameters

datamodule¶ (Union[str, LightningDataModule]) – A LightningDatamodule
encoder¶ (Union[str, Module, LightningModule]) – an encoder string or model
image_channels¶ (int) – 3
image_height¶ (int) – pixels
encoder_feature_dim¶ (int) – Called ndf in the paper, this is the representation size for the encoder.
embedding_fx_dim¶ (int) – Output dim of the embedding function (nrkhs in the paper) (Reproducing Kernel Hilbert Spaces).
conv_block_depth¶ (int) – Depth of each encoder block,
use_bn¶ (bool) – If true will use batchnorm.
tclip¶ (int) – soft clipping non-linearity to the scores after computing the regularization term and before computing the log-softmax. This is the ‘second trick’ used in the paper
learning_rate¶ (int) – The learning rate
data_dir¶ (str) – Where to store data
num_classes¶ (int) – How many classes in the dataset
batch_size¶ (int) – The batch size

BYOL¶

class pl_bolts.models.self_supervised.BYOL(num_classes, learning_rate=0.2, weight_decay=1.5e-06, input_height=32, batch_size=32, num_workers=0, warmup_epochs=10, max_epochs=1000, **kwargs)[source]

Bases: pytorch_lightning.LightningModule

PyTorch Lightning implementation of Bootstrap Your Own Latent (BYOL)

Paper authors: Jean-Bastien Grill, Florian Strub, Florent Altché, Corentin Tallec, Pierre H. Richemond, Elena Buchatskaya, Carl Doersch, Bernardo Avila Pires, Zhaohan Daniel Guo, Mohammad Gheshlaghi Azar, Bilal Piot, Koray Kavukcuoglu, Rémi Munos, Michal Valko.

Model implemented by:

Annika Brundyn

Warning

Work in progress. This implementation is still being verified.

TODOs:

verify on CIFAR-10
verify on STL-10
pre-train on imagenet

Example:

import pytorch_lightning as pl
from pl_bolts.models.self_supervised import BYOL
from pl_bolts.datamodules import CIFAR10DataModule
from pl_bolts.models.self_supervised.simclr.transforms import (
    SimCLREvalDataTransform, SimCLRTrainDataTransform)

# model
model = BYOL(num_classes=10)

# data
dm = CIFAR10DataModule(num_workers=0)
dm.train_transforms = SimCLRTrainDataTransform(32)
dm.val_transforms = SimCLREvalDataTransform(32)

trainer = pl.Trainer()
trainer.fit(model, dm)

Train:

trainer = Trainer()
trainer.fit(model)

CLI command:

# cifar10
python byol_module.py --gpus 1

# imagenet
python byol_module.py
    --gpus 8
    --dataset imagenet2012
    --data_dir /path/to/imagenet/
    --meta_dir /path/to/folder/with/meta.bin/
    --batch_size 32

Parameters

datamodule¶ – The datamodule
learning_rate¶ (float) – the learning rate
weight_decay¶ (float) – optimizer weight decay
input_height¶ (int) – image input height
batch_size¶ (int) – the batch size
num_workers¶ (int) – number of workers
warmup_epochs¶ (int) – num of epochs for scheduler warm up
max_epochs¶ (int) – max epochs for scheduler

CPC (V2)¶

PyTorch Lightning implementation of Data-Efficient Image Recognition with Contrastive Predictive Coding

Paper authors: (Olivier J. Hénaff, Aravind Srinivas, Jeffrey De Fauw, Ali Razavi, Carl Doersch, S. M. Ali Eslami, Aaron van den Oord).

Model implemented by:

William Falcon

Tullie Murrell

To Train:

import pytorch_lightning as pl
from pl_bolts.models.self_supervised import CPCV2
from pl_bolts.datamodules import CIFAR10DataModule
from pl_bolts.models.self_supervised.cpc import (
    CPCTrainTransformsCIFAR10, CPCEvalTransformsCIFAR10)

# data
dm = CIFAR10DataModule(num_workers=0)
dm.train_transforms = CPCTrainTransformsCIFAR10()
dm.val_transforms = CPCEvalTransformsCIFAR10()

# model
model = CPCV2()

# fit
trainer = pl.Trainer()
trainer.fit(model, dm)

To finetune:

python cpc_finetuner.py
    --ckpt_path path/to/checkpoint.ckpt
    --dataset cifar10
    --gpus 1

CIFAR-10 and STL-10 baselines¶

CPCv2 does not report baselines on CIFAR-10 and STL-10 datasets. Results in table are reported from the YADIM paper.

CPCv2 implementation results¶
Dataset	test acc	Encoder	Optimizer	Batch	Epochs	Hardware	LR
CIFAR-10	84.52	CPCresnet101	Adam	64	1000 (upto 24 hours)	1 V100 (32GB)	4e-5
STL-10	78.36	CPCresnet101	Adam	144	1000 (upto 72 hours)	4 V100 (32GB)	1e-4
ImageNet	54.82	CPCresnet101	Adam	3072	1000 (upto 21 days)	64 V100 (32GB)	4e-5

CIFAR-10 pretrained model:

from pl_bolts.models.self_supervised import CPCV2

weight_path = 'https://pl-bolts-weights.s3.us-east-2.amazonaws.com/cpc/cpc-cifar10-v4-exp3/epoch%3D474.ckpt'
cpc_v2 = CPCV2.load_from_checkpoint(weight_path, strict=False)

cpc_v2.freeze()

Tensorboard for CIFAR10

Pre-training:

Fine-tuning:

STL-10 pretrained model:

from pl_bolts.models.self_supervised import CPCV2

weight_path = 'https://pl-bolts-weights.s3.us-east-2.amazonaws.com/cpc/cpc-stl10-v0-exp3/epoch%3D624.ckpt'
cpc_v2 = CPCV2.load_from_checkpoint(weight_path, strict=False)

cpc_v2.freeze()

Tensorboard for STL10

Pre-training:

Fine-tuning:

ImageNet pretrained model:

from pl_bolts.models.self_supervised import CPCV2

weight_path = 'https://pl-bolts-weights.s3.us-east-2.amazonaws.com/cpc/cpcv2_weights/checkpoints/epoch%3D526.ckpt'
cpc_v2 = CPCV2.load_from_checkpoint(weight_path, strict=False)

cpc_v2.freeze()

Tensorboard for ImageNet

Pre-training:

Fine-tuning:

CPCV2 API¶

class pl_bolts.models.self_supervised.CPCV2(datamodule=None, encoder_name='cpc_encoder', patch_size=8, patch_overlap=4, online_ft=True, task='cpc', num_workers=4, learning_rate=0.0001, data_dir='', batch_size=32, pretrained=None, **kwargs)[source]

Bases: pytorch_lightning.LightningModule

Parameters

datamodule¶ (Optional[LightningDataModule]) – A Datamodule (optional). Otherwise set the dataloaders directly
encoder_name¶ (str) – A string for any of the resnets in torchvision, or the original CPC encoder, or a custon nn.Module encoder
patch_size¶ (int) – How big to make the image patches
patch_overlap¶ (int) – How much overlap should each patch have.
online_ft¶ (int) – Enable a 1024-unit MLP to fine-tune online
task¶ (str) – Which self-supervised task to use (‘cpc’, ‘amdim’, etc…)
num_workers¶ (int) – num dataloader worksers
learning_rate¶ (int) – what learning rate to use
data_dir¶ (str) – where to store data
batch_size¶ (int) – batch size
pretrained¶ (Optional[str]) – If true, will use the weights pretrained (using CPC) on Imagenet

Moco (V2)¶

class pl_bolts.models.self_supervised.MocoV2(base_encoder='resnet18', emb_dim=128, num_negatives=65536, encoder_momentum=0.999, softmax_temperature=0.07, learning_rate=0.03, momentum=0.9, weight_decay=0.0001, datamodule=None, data_dir='./', batch_size=256, use_mlp=False, num_workers=8, *args, **kwargs)[source]

Bases: pytorch_lightning.LightningModule

PyTorch Lightning implementation of Moco

Paper authors: Xinlei Chen, Haoqi Fan, Ross Girshick, Kaiming He.

Code adapted from facebookresearch/moco to Lightning by:

William Falcon

Example::: from pl_bolts.models.self_supervised import MocoV2 model = MocoV2() trainer = Trainer() trainer.fit(model)

CLI command:

# cifar10
python moco2_module.py --gpus 1

# imagenet
python moco2_module.py
    --gpus 8
    --dataset imagenet2012
    --data_dir /path/to/imagenet/
    --meta_dir /path/to/folder/with/meta.bin/
    --batch_size 32

Parameters

base_encoder¶ (Union[str, Module]) – torchvision model name or torch.nn.Module
emb_dim¶ (int) – feature dimension (default: 128)
num_negatives¶ (int) – queue size; number of negative keys (default: 65536)
encoder_momentum¶ (float) – moco momentum of updating key encoder (default: 0.999)
softmax_temperature¶ (float) – softmax temperature (default: 0.07)
learning_rate¶ (float) – the learning rate
momentum¶ (float) – optimizer momentum
weight_decay¶ (float) – optimizer weight decay
datamodule¶ (Optional[LightningDataModule]) – the DataModule (train, val, test dataloaders)
data_dir¶ (str) – the directory to store data
batch_size¶ (int) – batch size
use_mlp¶ (bool) – add an mlp to the encoders
num_workers¶ (int) – workers for the loaders

_batch_shuffle_ddp(x)[source]: Batch shuffle, for making use of BatchNorm. * Only support DistributedDataParallel (DDP) model. *

_batch_unshuffle_ddp(x, idx_unshuffle)[source]: Undo batch shuffle. * Only support DistributedDataParallel (DDP) model. *

_momentum_update_key_encoder()[source]: Momentum update of the key encoder

forward(img_q, img_k)[source]

Input:: im_q: a batch of query images im_k: a batch of key images
Output:: logits, targets

init_encoders(base_encoder)[source]: Override to add your own encoders

SimCLR¶

PyTorch Lightning implementation of SimCLR

Paper authors: Ting Chen, Simon Kornblith, Mohammad Norouzi, Geoffrey Hinton.

Model implemented by:

William Falcon

Tullie Murrell

To Train:

import pytorch_lightning as pl
from pl_bolts.models.self_supervised import SimCLR
from pl_bolts.datamodules import CIFAR10DataModule
from pl_bolts.models.self_supervised.simclr.transforms import (
    SimCLREvalDataTransform, SimCLRTrainDataTransform)

# data
dm = CIFAR10DataModule(num_workers=0)
dm.train_transforms = SimCLRTrainDataTransform(32)
dm.val_transforms = SimCLREvalDataTransform(32)

# model
model = SimCLR(num_samples=dm.num_samples, batch_size=dm.batch_size)

# fit
trainer = pl.Trainer()
trainer.fit(model, dm)

CIFAR-10 baseline¶

Cifar-10 implementation results¶
Implementation	test acc	Encoder	Optimizer	Batch	Epochs	Hardware	LR
Original	92.00?	resnet50	LARS	512	1000	1 V100 (32GB)	1.0
Ours	85.68	resnet50	LARS	512	960 (12 hr)	1 V100 (32GB)	1e-6

CIFAR-10 pretrained model:

from pl_bolts.models.self_supervised import SimCLR

weight_path = 'https://pl-bolts-weights.s3.us-east-2.amazonaws.com/simclr/simclr-cifar10-v1-exp12_87_52/epoch%3D960.ckpt'
simclr = SimCLR.load_from_checkpoint(weight_path, strict=False)

simclr.freeze()

Pre-training:

Fine-tuning (Single layer MLP, 1024 hidden units):

To reproduce:

# pretrain
python simclr_module.py
    --gpus 1
    --dataset cifar10
    --batch_size 512
    --learning_rate 1e-06
    --num_workers 8

# finetune
python simclr_finetuner.py
    --ckpt_path path/to/epoch=xyz.ckpt
    --gpus 1

SimCLR API¶

class pl_bolts.models.self_supervised.SimCLR(batch_size, num_samples, warmup_epochs=10, lr=0.0001, opt_weight_decay=1e-06, loss_temperature=0.5, **kwargs)[source]

Bases: pytorch_lightning.LightningModule

Parameters

batch_size¶ – the batch size
num_samples¶ – num samples in the dataset
warmup_epochs¶ – epochs to warmup the lr for
lr¶ – the optimizer learning rate
opt_weight_decay¶ – the optimizer weight decay
loss_temperature¶ – the loss temperature

Self-supervised learning Transforms¶

These transforms are used in various self-supervised learning approaches.

CPC transforms¶

Transforms used for CPC

CIFAR-10 Train (c)¶

class pl_bolts.models.self_supervised.cpc.transforms.CPCTrainTransformsCIFAR10(patch_size=8, overlap=4)[source]

Bases: object

Transforms used for CPC:

Parameters

patch_size¶ – size of patches when cutting up the image into overlapping patches
overlap¶ – how much to overlap patches

Transforms:

random_flip
img_jitter
col_jitter
rnd_gray
transforms.ToTensor()
normalize
Patchify(patch_size=patch_size, overlap_size=patch_size // 2)

Example:

# in a regular dataset
CIFAR10(..., transforms=CPCTrainTransformsCIFAR10())

# in a DataModule
module = CIFAR10DataModule(PATH)
train_loader = module.train_dataloader(batch_size=32, transforms=CPCTrainTransformsCIFAR10())

__call__(inp)[source]: Call self as a function.

CIFAR-10 Eval (c)¶

class pl_bolts.models.self_supervised.cpc.transforms.CPCEvalTransformsCIFAR10(patch_size=8, overlap=4)[source]

Bases: object

Transforms used for CPC:

Parameters

patch_size¶ – size of patches when cutting up the image into overlapping patches
overlap¶ – how much to overlap patches

Transforms:

random_flip
transforms.ToTensor()
normalize
Patchify(patch_size=patch_size, overlap_size=overlap)

Example:

# in a regular dataset
CIFAR10(..., transforms=CPCEvalTransformsCIFAR10())

# in a DataModule
module = CIFAR10DataModule(PATH)
train_loader = module.train_dataloader(batch_size=32, transforms=CPCEvalTransformsCIFAR10())

__call__(inp)[source]: Call self as a function.

Imagenet Train (c)¶

class pl_bolts.models.self_supervised.cpc.transforms.CPCTrainTransformsImageNet128(patch_size=32, overlap=16)[source]

Bases: object

Transforms used for CPC:

Parameters

patch_size¶ – size of patches when cutting up the image into overlapping patches
overlap¶ – how much to overlap patches

Transforms:

random_flip
transforms.ToTensor()
normalize
Patchify(patch_size=patch_size, overlap_size=patch_size // 2)

Example:

# in a regular dataset
Imagenet(..., transforms=CPCTrainTransformsImageNet128())

# in a DataModule
module = ImagenetDataModule(PATH)
train_loader = module.train_dataloader(batch_size=32, transforms=CPCTrainTransformsImageNet128())

__call__(inp)[source]: Call self as a function.

Imagenet Eval (c)¶

class pl_bolts.models.self_supervised.cpc.transforms.CPCEvalTransformsImageNet128(patch_size=32, overlap=16)[source]

Bases: object

Transforms used for CPC:

Parameters

patch_size¶ – size of patches when cutting up the image into overlapping patches
overlap¶ – how much to overlap patches

Transforms:

random_flip
transforms.ToTensor()
normalize
Patchify(patch_size=patch_size, overlap_size=patch_size // 2)

Example:

# in a regular dataset
Imagenet(..., transforms=CPCEvalTransformsImageNet128())

# in a DataModule
module = ImagenetDataModule(PATH)
train_loader = module.train_dataloader(batch_size=32, transforms=CPCEvalTransformsImageNet128())

__call__(inp)[source]: Call self as a function.

STL-10 Train (c)¶

class pl_bolts.models.self_supervised.cpc.transforms.CPCTrainTransformsSTL10(patch_size=16, overlap=8)[source]

Bases: object

Transforms used for CPC:

Parameters

patch_size¶ – size of patches when cutting up the image into overlapping patches
overlap¶ – how much to overlap patches

Transforms:

random_flip
img_jitter
col_jitter
rnd_gray
transforms.ToTensor()
normalize
Patchify(patch_size=patch_size, overlap_size=patch_size // 2)

Example:

# in a regular dataset
STL10(..., transforms=CPCTrainTransformsSTL10())

# in a DataModule
module = STL10DataModule(PATH)
train_loader = module.train_dataloader(batch_size=32, transforms=CPCTrainTransformsSTL10())

__call__(inp)[source]: Call self as a function.

STL-10 Eval (c)¶

class pl_bolts.models.self_supervised.cpc.transforms.CPCEvalTransformsSTL10(patch_size=16, overlap=8)[source]

Bases: object

Transforms used for CPC:

Parameters

patch_size¶ – size of patches when cutting up the image into overlapping patches
overlap¶ – how much to overlap patches

Transforms:

random_flip
transforms.ToTensor()
normalize
Patchify(patch_size=patch_size, overlap_size=patch_size // 2)

Example:

# in a regular dataset
STL10(..., transforms=CPCEvalTransformsSTL10())

# in a DataModule
module = STL10DataModule(PATH)
train_loader = module.train_dataloader(batch_size=32, transforms=CPCEvalTransformsSTL10())

__call__(inp)[source]: Call self as a function.

AMDIM transforms¶

Transforms used for AMDIM

CIFAR-10 Train (a)¶

class pl_bolts.models.self_supervised.amdim.transforms.AMDIMTrainTransformsCIFAR10[source]

Bases: object

Transforms applied to AMDIM

Transforms:

img_jitter,
col_jitter,
rnd_gray,
transforms.ToTensor(),
normalize

Example:

x = torch.rand(5, 3, 32, 32)

transform = AMDIMTrainTransformsCIFAR10()
(view1, view2) = transform(x)

__call__(inp)[source]: Call self as a function.

CIFAR-10 Eval (a)¶

class pl_bolts.models.self_supervised.amdim.transforms.AMDIMEvalTransformsCIFAR10[source]

Bases: object

Transforms applied to AMDIM

Transforms:

transforms.ToTensor(),
normalize

Example:

x = torch.rand(5, 3, 32, 32)

transform = AMDIMEvalTransformsCIFAR10()
(view1, view2) = transform(x)

__call__(inp)[source]: Call self as a function.

Imagenet Train (a)¶

class pl_bolts.models.self_supervised.amdim.transforms.AMDIMTrainTransformsImageNet128(height=128)[source]

Bases: object

Transforms applied to AMDIM

Transforms:

img_jitter,
col_jitter,
rnd_gray,
transforms.ToTensor(),
normalize

Example:

x = torch.rand(5, 3, 128, 128)

transform = AMDIMTrainTransformsSTL10()
(view1, view2) = transform(x)

__call__(inp)[source]: Call self as a function.

Imagenet Eval (a)¶

class pl_bolts.models.self_supervised.amdim.transforms.AMDIMEvalTransformsImageNet128(height=128)[source]

Bases: object

Transforms applied to AMDIM

Transforms:

transforms.Resize(height + 6, interpolation=3),
transforms.CenterCrop(height),
transforms.ToTensor(),
normalize

Example:

x = torch.rand(5, 3, 128, 128)

transform = AMDIMEvalTransformsImageNet128()
view1 = transform(x)

__call__(inp)[source]: Call self as a function.

STL-10 Train (a)¶

class pl_bolts.models.self_supervised.amdim.transforms.AMDIMTrainTransformsSTL10(height=64)[source]

Bases: object

Transforms applied to AMDIM

Transforms:

img_jitter,
col_jitter,
rnd_gray,
transforms.ToTensor(),
normalize

Example:

x = torch.rand(5, 3, 64, 64)

transform = AMDIMTrainTransformsSTL10()
(view1, view2) = transform(x)

__call__(inp)[source]: Call self as a function.

STL-10 Eval (a)¶

class pl_bolts.models.self_supervised.amdim.transforms.AMDIMEvalTransformsSTL10(height=64)[source]

Bases: object

Transforms applied to AMDIM

Transforms:

transforms.Resize(height + 6, interpolation=3),
transforms.CenterCrop(height),
transforms.ToTensor(),
normalize

Example:

x = torch.rand(5, 3, 64, 64)

transform = AMDIMTrainTransformsSTL10()
view1 = transform(x)

__call__(inp)[source]: Call self as a function.

MOCO V2 transforms¶

Transforms used for MOCO V2

CIFAR-10 Train (m2)¶

class pl_bolts.models.self_supervised.moco.transforms.Moco2TrainCIFAR10Transforms(height=32)[source]

Bases: object

Moco 2 augmentation: https://arxiv.org/pdf/2003.04297.pdf

__call__(inp)[source]: Call self as a function.

CIFAR-10 Eval (m2)¶

class pl_bolts.models.self_supervised.moco.transforms.Moco2EvalCIFAR10Transforms(height=32)[source]

Bases: object

Moco 2 augmentation: https://arxiv.org/pdf/2003.04297.pdf

__call__(inp)[source]: Call self as a function.

Imagenet Train (m2)¶

class pl_bolts.models.self_supervised.moco.transforms.Moco2TrainSTL10Transforms(height=64)[source]

Bases: object

Moco 2 augmentation: https://arxiv.org/pdf/2003.04297.pdf

__call__(inp)[source]: Call self as a function.

Imagenet Eval (m2)¶

class pl_bolts.models.self_supervised.moco.transforms.Moco2EvalSTL10Transforms(height=64)[source]

Bases: object

Moco 2 augmentation: https://arxiv.org/pdf/2003.04297.pdf

__call__(inp)[source]: Call self as a function.

STL-10 Train (m2)¶

class pl_bolts.models.self_supervised.moco.transforms.Moco2TrainImagenetTransforms(height=128)[source]

Bases: object

Moco 2 augmentation: https://arxiv.org/pdf/2003.04297.pdf

__call__(inp)[source]: Call self as a function.

STL-10 Eval (m2)¶

class pl_bolts.models.self_supervised.moco.transforms.Moco2EvalImagenetTransforms(height=128)[source]

Bases: object

Moco 2 augmentation: https://arxiv.org/pdf/2003.04297.pdf

__call__(inp)[source]: Call self as a function.

SimCLR transforms¶

Transforms used for SimCLR

Train (sc)¶

class pl_bolts.models.self_supervised.simclr.transforms.SimCLRTrainDataTransform(input_height, s=1)[source]

Bases: object

Transforms for SimCLR

Transform:

RandomResizedCrop(size=self.input_height)
RandomHorizontalFlip()
RandomApply([color_jitter], p=0.8)
RandomGrayscale(p=0.2)
GaussianBlur(kernel_size=int(0.1 * self.input_height))
transforms.ToTensor()

Example:

from pl_bolts.models.self_supervised.simclr.transforms import SimCLRTrainDataTransform

transform = SimCLRTrainDataTransform(input_height=32)
x = sample()
(xi, xj) = transform(x)

__call__(sample)[source]: Call self as a function.

Eval (sc)¶

class pl_bolts.models.self_supervised.simclr.transforms.SimCLREvalDataTransform(input_height, s=1)[source]

Bases: object

Transforms for SimCLR

Transform:

Resize(input_height + 10, interpolation=3)
transforms.CenterCrop(input_height),
transforms.ToTensor()

Example:

from pl_bolts.models.self_supervised.simclr.transforms import SimCLREvalDataTransform

transform = SimCLREvalDataTransform(input_height=32)
x = sample()
(xi, xj) = transform(x)

__call__(sample)[source]: Call self as a function.

Self-supervised learning¶

Collection of useful functions for self-supervised learning

Identity class¶

Example:

from pl_bolts.utils import Identity

class pl_bolts.utils.self_supervised.Identity[source]

Bases: torch.nn.Module

An identity class to replace arbitrary layers in pretrained models

Example:

from pl_bolts.utils import Identity

model = resnet18()
model.fc = Identity()

SSL-ready resnets¶

Torchvision resnets with the fc layers removed and with the ability to return all feature maps instead of just the last one.

Example:

from pl_bolts.utils.self_supervised import torchvision_ssl_encoder

resnet = torchvision_ssl_encoder('resnet18', pretrained=False, return_all_feature_maps=True)
x = torch.rand(3, 3, 32, 32)

feat_maps = resnet(x)

pl_bolts.utils.self_supervised.torchvision_ssl_encoder(name, pretrained=False, return_all_feature_maps=False)[source]

SSL backbone finetuner¶

class pl_bolts.models.self_supervised.ssl_finetuner.SSLFineTuner(backbone, in_features, num_classes, hidden_dim=1024)[source]

Bases: pytorch_lightning.LightningModule

Finetunes a self-supervised learning backbone using the standard evaluation protocol of a singler layer MLP with 1024 units

Example:

from pl_bolts.utils.self_supervised import SSLFineTuner
from pl_bolts.models.self_supervised import CPCV2
from pl_bolts.datamodules import CIFAR10DataModule
from pl_bolts.models.self_supervised.cpc.transforms import CPCEvalTransformsCIFAR10,
                                                            CPCTrainTransformsCIFAR10

# pretrained model
backbone = CPCV2.load_from_checkpoint(PATH, strict=False)

# dataset + transforms
dm = CIFAR10DataModule(data_dir='.')
dm.train_transforms = CPCTrainTransformsCIFAR10()
dm.val_transforms = CPCEvalTransformsCIFAR10()

# finetuner
finetuner = SSLFineTuner(backbone, in_features=backbone.z_dim, num_classes=backbone.num_classes)

# train
trainer = pl.Trainer()
trainer.fit(finetuner, dm)

# test
trainer.test(datamodule=dm)

Parameters

backbone¶ – a pretrained model
in_features¶ – feature dim of backbone outputs
num_classes¶ – classes of the dataset
hidden_dim¶ – dim of the MLP (1024 default used in self-supervised literature)

Semi-supervised learning¶

Collection of utilities for semi-supervised learning where some part of the data is labeled and the other part is not.

Balanced classes¶

Example:

from pl_bolts.utils.semi_supervised import balance_classes

pl_bolts.utils.semi_supervised.balance_classes(X, Y, batch_size)[source]

Makes sure each batch has an equal amount of data from each class. Perfect balance

Parameters

X¶ (ndarray) – input features
Y¶ (list) – mixed labels (ints)
batch_size¶ (int) – the ultimate batch size

half labeled batches¶

Example:

from pl_bolts.utils.semi_supervised import balance_classes

pl_bolts.utils.semi_supervised.generate_half_labeled_batches(smaller_set_X, smaller_set_Y, larger_set_X, larger_set_Y, batch_size)[source]: Given a labeled dataset and an unlabeled dataset, this function generates a joint pair where half the batches are labeled and the other half is not

Self-supervised Learning Contrastive tasks¶

This section implements popular contrastive learning tasks used in self-supervised learning.

FeatureMapContrastiveTask¶

This task compares sets of feature maps.

In general the feature map comparison pretext task uses triplets of features. Here are the abstract steps of comparison.

Generate multiple views of the same image

x1_view_1 = data_augmentation(x1)
x1_view_2 = data_augmentation(x1)

Use a different example to generate additional views (usually within the same batch or a pool of candidates)

x2_view_1 = data_augmentation(x2)
x2_view_2 = data_augmentation(x2)

Pick 3 views to compare, these are the anchor, positive and negative features

anchor = x1_view_1
positive = x1_view_2
negative = x2_view_1

Generate feature maps for each view

(a0, a1, a2) = encoder(anchor)
(p0, p1, p2) = encoder(positive)

Make a comparison for a set of feature maps

phi = some_score_function()

# the '01' comparison
score = phi(a0, p1)

# and can be bidirectional
score = phi(p0, a1)

In practice the contrastive task creates a BxB matrix where B is the batch size. The diagonals for set 1 of feature maps are the anchors, the diagonals of set 2 of the feature maps are the positives, the non-diagonals of set 1 are the negatives.

class pl_bolts.losses.self_supervised_learning.FeatureMapContrastiveTask(comparisons='00, 11', tclip=10.0, bidirectional=True)[source]

Bases: torch.nn.Module

Performs an anchor, positive negative pair comparison for each each tuple of feature maps passed.

# extract feature maps
pos_0, pos_1, pos_2 = encoder(x_pos)
anc_0, anc_1, anc_2 = encoder(x_anchor)

# compare only the 0th feature maps
task = FeatureMapContrastiveTask('00')
loss, regularizer = task((pos_0), (anc_0))

# compare (pos_0 to anc_1) and (pos_0, anc_2)
task = FeatureMapContrastiveTask('01, 02')
losses, regularizer = task((pos_0, pos_1, pos_2), (anc_0, anc_1, anc_2))
loss = losses.sum()

# compare (pos_1 vs a anc_random)
task = FeatureMapContrastiveTask('0r')
loss, regularizer = task((pos_0, pos_1, pos_2), (anc_0, anc_1, anc_2))

Parameters

comparisons¶ (str) – groupings of feature map indices to compare (zero indexed, ‘r’ means random) ex: ‘00, 1r’
tclip¶ (float) – stability clipping value
bidirectional¶ (bool) – if true, does the comparison both ways

# with bidirectional the comparisons are done both ways
task = FeatureMapContrastiveTask('01, 02')

# will compare the following:
# 01: (pos_0, anc_1), (anc_0, pos_1)
# 02: (pos_0, anc_2), (anc_0, pos_2)

forward(anchor_maps, positive_maps)[source]

Takes in a set of tuples, each tuple has two feature maps with all matching dimensions

Example

>>> import torch
>>> from pytorch_lightning import seed_everything
>>> seed_everything(0)
0
>>> a1 = torch.rand(3, 5, 2, 2)
>>> a2 = torch.rand(3, 5, 2, 2)
>>> b1 = torch.rand(3, 5, 2, 2)
>>> b2 = torch.rand(3, 5, 2, 2)
...
>>> task = FeatureMapContrastiveTask('01, 11')
...
>>> losses, regularizer = task((a1, a2), (b1, b2))
>>> losses
tensor([2.2351, 2.1902])
>>> regularizer
tensor(0.0324)

static parse_map_indexes(comparisons)[source]

Example:

>>> FeatureMapContrastiveTask.parse_map_indexes('11')
[(1, 1)]
>>> FeatureMapContrastiveTask.parse_map_indexes('11,59')
[(1, 1), (5, 9)]
>>> FeatureMapContrastiveTask.parse_map_indexes('11,59, 2r')
[(1, 1), (5, 9), (2, -1)]

Context prediction tasks¶

The following tasks aim to predict a target using a context representation.

CPCContrastiveTask¶

This is the predictive task from CPC (v2).

task = CPCTask(num_input_channels=32)

# (batch, channels, rows, cols)
# this should be thought of as 49 feature vectors, each with 32 dims
Z = torch.random.rand(3, 32, 7, 7)

loss = task(Z)

class pl_bolts.losses.self_supervised_learning.CPCTask(num_input_channels, target_dim=64, embed_scale=0.1)[source]

Bases: torch.nn.Module

Loss used in CPC

Indices and tables¶

Logo

PyTorch Lightning Bolts¶

Pretrained SOTA Deep Learning models, callbacks and more for research and production with PyTorch Lightning and PyTorch

Continuous Integration¶

Install¶

Simple installation from PyPI

pip install pytorch-lightning-bolts

Install bleeding-edge (no guarantees)

pip install git+https://github.com/PytorchLightning/pytorch-lightning-bolts.git@master --upgrade

In case you wan to have full experience you can install all optional packages at once

pip install pytorch-lightning-bolts["extra"]

Docs¶

What is Bolts¶

Bolts is a Deep learning research and production toolbox of:

SOTA pretrained models.
Model components.
Callbacks.
Losses.
Datasets.

Main Goals of Bolts¶

The main goal of Bolts is to enable rapid model idea iteration.

Example 1: Finetuning on data¶

from pl_bolts.models.self_supervised import SimCLR
from pl_bolts.models.self_supervised.simclr.transforms import SimCLRTrainDataTransform, SimCLREvalDataTransform
import pytorch_lightning as pl

# data
train_data = DataLoader(MyDataset(transforms=SimCLRTrainDataTransform(input_height=32)))
val_data = DataLoader(MyDataset(transforms=SimCLREvalDataTransform(input_height=32)))

# model
weight_path = 'https://pl-bolts-weights.s3.us-east-2.amazonaws.com/simclr/simclr-cifar10-v1-exp12_87_52/epoch%3D960.ckpt'
simclr = SimCLR.load_from_checkpoint(weight_path, strict=False)

simclr.freeze()

# finetune

Example 2: Subclass and ideate¶

from pl_bolts.models import ImageGPT
from pl_bolts.models.self_supervised import SimCLR

class VideoGPT(ImageGPT):

    def training_step(self, batch, batch_idx):
        x, y = batch
        x = _shape_input(x)

        logits = self.gpt(x)
        simclr_features = self.simclr(x)

        # -----------------
        # do something new with GPT logits + simclr_features
        # -----------------

        loss = self.criterion(logits.view(-1, logits.size(-1)), x.view(-1).long())

        logs = {"loss": loss}
        return {"loss": loss, "log": logs}

Who is Bolts for?¶

Corporate production teams
Professional researchers
Ph.D. students
Linear + Logistic regression heroes

I don’t need deep learning¶

Great! We have LinearRegression and LogisticRegression implementations with numpy and sklearn bridges for datasets! But our implementations work on multiple GPUs, TPUs and scale dramatically…

Check out our Linear Regression on TPU demo

from pl_bolts.models.regression import LinearRegression
from pl_bolts.datamodules import SklearnDataModule
from sklearn.datasets import load_boston
import pytorch_lightning as pl

# sklearn dataset
X, y = load_boston(return_X_y=True)
loaders = SklearnDataModule(X, y)

model = LinearRegression(input_dim=13)

# try with gpus=4!
# trainer = pl.Trainer(gpus=4)
trainer = pl.Trainer()
trainer.fit(model, loaders.train_dataloader(), loaders.val_dataloader())
trainer.test(test_dataloaders=loaders.test_dataloader())

Is this another model zoo?¶

No!

Bolts is unique because models are implemented using PyTorch Lightning and structured so that they can be easily subclassed and iterated on.

For example, you can override the elbo loss of a VAE, or the generator_step of a GAN to quickly try out a new idea. The best part is that all the models are benchmarked so you won’t waste time trying to “reproduce” or find the bugs with your implementation.

Team¶

Bolts is supported by the PyTorch Lightning team and the PyTorch Lightning community!

pl_bolts.callbacks package¶

Collection of PyTorchLightning callbacks

Subpackages¶

pl_bolts.callbacks.vision package¶

Submodules¶

pl_bolts.callbacks.vision.confused_logit module¶

class pl_bolts.callbacks.vision.confused_logit.ConfusedLogitCallback(top_k, projection_factor=3, min_logit_value=5.0, logging_batch_interval=20, max_logit_difference=0.1)[source]¶

Bases: pytorch_lightning.Callback

Takes the logit predictions of a model and when the probabilities of two classes are very close, the model doesn’t have high certainty that it should pick one vs the other class.

This callback shows how the input would have to change to swing the model from one label prediction to the other.

In this case, the network predicts a 5… but gives almost equal probability to an 8. The images show what about the original 5 would have to change to make it more like a 5 or more like an 8.

For each confused logit the confused images are generated by taking the gradient from a logit wrt an input for the top two closest logits.

Example:

from pl_bolts.callbacks.vision import ConfusedLogitCallback
trainer = Trainer(callbacks=[ConfusedLogitCallback()])

Note

whenever called, this model will look for self.last_batch and self.last_logits in the LightningModule

Note

this callback supports tensorboard only right now

Parameters

top_k – How many “offending” images we should plot
projection_factor – How much to multiply the input image to make it look more like this logit label
min_logit_value – Only consider logit values above this threshold
logging_batch_interval – how frequently to inspect/potentially plot something
max_logit_difference – when the top 2 logits are within this threshold we consider them confused

Authored by:

Alfredo Canziani

static _ConfusedLogitCallback__draw_sample(fig, axarr, row_idx, col_idx, img, title)[source]¶

_plot(confusing_x, confusing_y, trainer, model, mask_idxs)[source]¶

on_train_batch_end(trainer, pl_module, batch, batch_idx, dataloader_idx)[source]¶

pl_bolts.callbacks.vision.image_generation module¶

class pl_bolts.callbacks.vision.image_generation.TensorboardGenerativeModelImageSampler(num_samples=3)[source]¶

Bases: pytorch_lightning.Callback

Generates images and logs to tensorboard. Your model must implement the forward function for generation

Requirements:

# model must have img_dim arg
model.img_dim = (1, 28, 28)

# model forward must work for sampling
z = torch.rand(batch_size, latent_dim)
img_samples = your_model(z)

Example:

from pl_bolts.callbacks import TensorboardGenerativeModelImageSampler

trainer = Trainer(callbacks=[TensorboardGenerativeModelImageSampler()])

on_epoch_end(trainer, pl_module)[source]¶

Submodules¶

pl_bolts.callbacks.printing module¶

class pl_bolts.callbacks.printing.PrintTableMetricsCallback[source]¶

Bases: pytorch_lightning.callbacks.Callback

Prints a table with the metrics in columns on every epoch end

Example:

from pl_bolts.callbacks import PrintTableMetricsCallback

callback = PrintTableMetricsCallback()

pass into trainer like so:

trainer = pl.Trainer(callbacks=[callback])
trainer.fit(...)

# ------------------------------
# at the end of every epoch it will print
# ------------------------------

# loss│train_loss│val_loss│epoch
# ──────────────────────────────
# 2.2541470527648926│2.2541470527648926│2.2158432006835938│0

on_epoch_end(trainer, pl_module)[source]¶

pl_bolts.callbacks.printing.dicts_to_table(dicts, keys=None, pads=None, fcodes=None, convert_headers=None, header_names=None, skip_none_lines=False, replace_values=None)[source]¶

Generate ascii table from dictionary Taken from (https://stackoverflow.com/questions/40056747/print-a-list-of-dictionaries-in-table-form)

Parameters

dicts (List[Dict]) – input dictionary list; empty lists make keys OR header_names mandatory
keys (Optional[List[str]]) – order list of keys to generate columns for; no key/dict-key should suffix with ‘____’ else adjust code-suffix
pads (Optional[List[str]]) – indicate padding direction and size, eg <10 to right pad alias left-align
fcodes (Optional[List[str]]) – formating codes for respective column type, eg .3f
convert_headers (Optional[Dict[str, Callable]]) – apply converters(dict) on column keys k, eg timestamps
header_names (Optional[List[str]]) – supply for custom column headers instead of keys
skip_none_lines (bool) – skip line if contains None
replace_values (Optional[Dict[str, Any]]) – specify per column keys k a map from seen value to new value; new value must comply with the columns fcode; CAUTION: modifies input (due speed)

Example

>>> a = {'a': 1, 'b': 2}
>>> b = {'a': 3, 'b': 4}
>>> print(dicts_to_table([a, b]))
a│b
───
1│2
3│4

pl_bolts.callbacks.self_supervised module¶

class pl_bolts.callbacks.self_supervised.BYOLMAWeightUpdate(initial_tau=0.996)[source]¶

Bases: pytorch_lightning.Callback

Weight update rule from BYOL.

Your model should have a:

self.online_network.

self.target_network.

Updates the target_network params using an exponential moving average update rule weighted by tau. BYOL claims this keeps the online_network from collapsing.

Note

Automatically increases tau from initial_tau to 1.0 with every training step

Example:

from pl_bolts.callbacks.self_supervised import BYOLMAWeightUpdate

# model must have 2 attributes
model = Model()
model.online_network = ...
model.target_network = ...

trainer = Trainer(callbacks=[BYOLMAWeightUpdate()])

Parameters: initial_tau – starting tau. Auto-updates with every training step

on_train_batch_end(trainer, pl_module, batch, batch_idx, dataloader_idx)[source]¶

update_tau(pl_module, trainer)[source]¶

update_weights(online_net, target_net)[source]¶

class pl_bolts.callbacks.self_supervised.SSLOnlineEvaluator(drop_p=0.2, hidden_dim=1024, z_dim=None, num_classes=None)[source]¶

Bases: pytorch_lightning.Callback

Attaches a MLP for finetuning using the standard self-supervised protocol.

Example:

from pl_bolts.callbacks.self_supervised import SSLOnlineEvaluator

# your model must have 2 attributes
model = Model()
model.z_dim = ... # the representation dim
model.num_classes = ... # the num of classes in the model

Parameters

drop_p (float) – (0.2) dropout probability
hidden_dim (int) –
1. the hidden dimension for the finetune MLP

get_representations(pl_module, x)[source]¶: Override this to customize for the particular model :param _sphinx_paramlinks_pl_bolts.callbacks.self_supervised.SSLOnlineEvaluator.get_representations.pl_module: :param _sphinx_paramlinks_pl_bolts.callbacks.self_supervised.SSLOnlineEvaluator.get_representations.x:

on_pretrain_routine_start(trainer, pl_module)[source]¶

on_train_batch_end(trainer, pl_module, batch, batch_idx, dataloader_idx)[source]¶

to_device(batch, device)[source]¶

pl_bolts.callbacks.variational module¶

class pl_bolts.callbacks.variational.LatentDimInterpolator(interpolate_epoch_interval=20, range_start=-5, range_end=5, num_samples=2)[source]¶

Bases: pytorch_lightning.callbacks.Callback

Interpolates the latent space for a model by setting all dims to zero and stepping through the first two dims increasing one unit at a time.

Default interpolates between [-5, 5] (-5, -4, -3, …, 3, 4, 5)

Example:

from pl_bolts.callbacks import LatentDimInterpolator

Trainer(callbacks=[LatentDimInterpolator()])

Parameters

interpolate_epoch_interval –
range_start – default -5
range_end – default 5
num_samples – default 2

interpolate_latent_space(pl_module, latent_dim)[source]¶

on_epoch_end(trainer, pl_module)[source]¶

pl_bolts.datamodules package¶

Submodules¶

pl_bolts.datamodules.async_dataloader module¶

class pl_bolts.datamodules.async_dataloader.AsynchronousLoader(data, device=torch.device, q_size=10, num_batches=None, **kwargs)[source]¶

Bases: object

Class for asynchronously loading from CPU memory to device memory with DataLoader.

Note that this only works for single GPU training, multiGPU uses PyTorch’s DataParallel or DistributedDataParallel which uses its own code for transferring data across GPUs. This could just break or make things slower with DataParallel or DistributedDataParallel.

Parameters

data – The PyTorch Dataset or DataLoader we’re using to load.
device – The PyTorch device we are loading to
q_size – Size of the queue used to store the data loaded to the device
num_batches – Number of batches to load. This must be set if the dataloader doesn’t have a finite __len__. It will also override DataLoader.__len__ if set and DataLoader has a __len__. Otherwise it can be left as None
**kwargs – Any additional arguments to pass to the dataloader if we’re constructing one here

load_instance(sample)[source]¶

load_loop()[source]¶

pl_bolts.datamodules.base_dataset module¶

class pl_bolts.datamodules.base_dataset.LightDataset(*args, **kwargs)[source]¶

Bases: abc.ABC, torch.utils.data.Dataset

_download_from_url(base_url, data_folder, file_name)[source]¶

static _prepare_subset(full_data, full_targets, num_samples, labels)[source]¶

Prepare a subset of a common dataset.

Return type: Tuple[Tensor, Tensor]

DATASET_NAME = 'light'[source]¶

cache_folder_name: str = None[source]¶

property cached_folder_path[source]¶

Return type: str

data: torch.Tensor = None[source]¶

dir_path: str = None[source]¶

normalize: tuple = None[source]¶

targets: torch.Tensor = None[source]¶

pl_bolts.datamodules.binary_mnist_datamodule module¶

class pl_bolts.datamodules.binary_mnist_datamodule.BinaryMNISTDataModule(data_dir, val_split=5000, num_workers=16, normalize=False, seed=42, *args, **kwargs)[source]¶

Bases: pytorch_lightning.LightningDataModule

Specs:

10 classes (1 per digit)
Each image is (1 x 28 x 28)

Binary MNIST, train, val, test splits and transforms

Transforms:

mnist_transforms = transform_lib.Compose([
    transform_lib.ToTensor()
])

Example:

from pl_bolts.datamodules import BinaryMNISTDataModule

dm = BinaryMNISTDataModule('.')
model = LitModel()

Trainer().fit(model, dm)

Parameters

data_dir (str) – where to save/load the data
val_split (int) – how many of the training images to use for the validation split
num_workers (int) – how many workers to use for loading data
normalize (bool) – If true applies image normalize

_default_transforms()[source]¶

prepare_data()[source]¶: Saves MNIST files to data_dir

test_dataloader(batch_size=32, transforms=None)[source]¶

MNIST test set uses the test split

Parameters

batch_size – size of batch
transforms – custom transforms

train_dataloader(batch_size=32, transforms=None)[source]¶

MNIST train set removes a subset to use for validation

Parameters

batch_size – size of batch
transforms – custom transforms

val_dataloader(batch_size=32, transforms=None)[source]¶

MNIST val set uses a subset of the training set for validation

Parameters

batch_size – size of batch
transforms – custom transforms

name = 'mnist'[source]¶

property num_classes[source]¶: Return: 10

pl_bolts.datamodules.cifar10_datamodule module¶

class pl_bolts.datamodules.cifar10_datamodule.CIFAR10DataModule(data_dir=None, val_split=5000, num_workers=16, batch_size=32, seed=42, *args, **kwargs)[source]¶

Bases: pytorch_lightning.LightningDataModule

Specs:

10 classes (1 per class)
Each image is (3 x 32 x 32)

Standard CIFAR10, train, val, test splits and transforms

Transforms:

mnist_transforms = transform_lib.Compose([
    transform_lib.ToTensor(),
    transforms.Normalize(
        mean=[x / 255.0 for x in [125.3, 123.0, 113.9]],
        std=[x / 255.0 for x in [63.0, 62.1, 66.7]]
    )
])

Example:

from pl_bolts.datamodules import CIFAR10DataModule

dm = CIFAR10DataModule(PATH)
model = LitModel()

Trainer().fit(model, dm)

Or you can set your own transforms

Example:

dm.train_transforms = ...
dm.test_transforms = ...
dm.val_transforms  = ...

Parameters

data_dir (Optional[str]) – where to save/load the data
val_split (int) – how many of the training images to use for the validation split
num_workers (int) – how many workers to use for loading data
batch_size (int) – number of examples per training/eval step

default_transforms()[source]¶

prepare_data()[source]¶: Saves CIFAR10 files to data_dir

test_dataloader()[source]¶: CIFAR10 test set uses the test split

train_dataloader()[source]¶: CIFAR train set removes a subset to use for validation

val_dataloader()[source]¶: CIFAR10 val set uses a subset of the training set for validation

extra_args = {}[source]¶

name = 'cifar10'[source]¶

property num_classes[source]¶: Return: 10

class pl_bolts.datamodules.cifar10_datamodule.TinyCIFAR10DataModule(data_dir, val_split=50, num_workers=16, num_samples=100, labels=(1, 5, 8), *args, **kwargs)[source]¶

Bases: pl_bolts.datamodules.cifar10_datamodule.CIFAR10DataModule

Standard CIFAR10, train, val, test splits and transforms

Transforms:

mnist_transforms = transform_lib.Compose([
    transform_lib.ToTensor(),
    transforms.Normalize(mean=[x / 255.0 for x in [125.3, 123.0, 113.9]],
                         std=[x / 255.0 for x in [63.0, 62.1, 66.7]])
])

Example:

from pl_bolts.datamodules import CIFAR10DataModule

dm = CIFAR10DataModule(PATH)
model = LitModel(datamodule=dm)

Parameters

data_dir (str) – where to save/load the data
val_split (int) – how many of the training images to use for the validation split
num_workers (int) – how many workers to use for loading data
num_samples (int) – number of examples per selected class/label
labels (Optional[Sequence]) – list selected CIFAR10 classes/labels

property num_classes[source]¶

Return number of classes.

Return type: int

pl_bolts.datamodules.cifar10_dataset module¶

class pl_bolts.datamodules.cifar10_dataset.CIFAR10(data_dir='.', train=True, transform=None, download=True)[source]¶

Bases: pl_bolts.datamodules.base_dataset.LightDataset

Customized CIFAR10 dataset for testing Pytorch Lightning without the torchvision dependency.

Part of the code was copied from https://github.com/pytorch/vision/blob/build/v0.5.0/torchvision/datasets/

Parameters

data_dir (str) – Root directory of dataset where CIFAR10/processed/training.pt and CIFAR10/processed/test.pt exist.
train (bool) – If True, creates dataset from training.pt, otherwise from test.pt.
download (bool) – If true, downloads the dataset from the internet and puts it in root directory. If dataset is already downloaded, it is not downloaded again.

Examples

>>> from torchvision import transforms
>>> from pl_bolts.transforms.dataset_normalizations import cifar10_normalization
>>> cf10_transforms = transforms.Compose([transforms.ToTensor(), cifar10_normalization()])
>>> dataset = CIFAR10(download=True, transform=cf10_transforms)
>>> len(dataset)
50000
>>> torch.bincount(dataset.targets)
tensor([5000, 5000, 5000, 5000, 5000, 5000, 5000, 5000, 5000, 5000])
>>> data, label = dataset[0]
>>> data.shape
torch.Size([3, 32, 32])
>>> label
6

Labels:

airplane: 0
automobile: 1
bird: 2
cat: 3
deer: 4
dog: 5
frog: 6
horse: 7
ship: 8
truck: 9

classmethod _check_exists(data_folder, file_names)[source]¶

Return type: bool

_extract_archive_save_torch(download_path)[source]¶

_unpickle(path_folder, file_name)[source]¶

Return type: Tuple[Tensor, Tensor]

download(data_folder)[source]¶

Download the data if it doesn’t exist in cached_folder_path already.

Return type: None

prepare_data(download)[source]¶

BASE_URL = 'https://www.cs.toronto.edu/~kriz/'[source]¶

DATASET_NAME = 'CIFAR10'[source]¶

FILE_NAME = 'cifar-10-python.tar.gz'[source]¶

TEST_FILE_NAME = 'test.pt'[source]¶

TRAIN_FILE_NAME = 'training.pt'[source]¶

cache_folder_name: str = 'complete'[source]¶

data = None[source]¶

dir_path = None[source]¶

labels = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9}[source]¶

normalize = None[source]¶

relabel = False[source]¶

targets = None[source]¶

class pl_bolts.datamodules.cifar10_dataset.TrialCIFAR10(data_dir='.', train=True, transform=None, download=False, num_samples=100, labels=(1, 5, 8), relabel=True)[source]¶

Bases: pl_bolts.datamodules.cifar10_dataset.CIFAR10

Customized CIFAR10 dataset for testing Pytorch Lightning without the torchvision dependency.

Parameters

data_dir (str) – Root directory of dataset where CIFAR10/processed/training.pt and CIFAR10/processed/test.pt exist.
train (bool) – If True, creates dataset from training.pt, otherwise from test.pt.
download (bool) – If true, downloads the dataset from the internet and puts it in root directory. If dataset is already downloaded, it is not downloaded again.
num_samples (int) – number of examples per selected class/digit
labels (Optional[Sequence]) – list selected CIFAR10 digits/classes

Examples

>>> dataset = TrialCIFAR10(download=True, num_samples=150, labels=(1, 5, 8))
>>> len(dataset)
450
>>> sorted(set([d.item() for d in dataset.targets]))
[1, 5, 8]
>>> torch.bincount(dataset.targets)
tensor([  0, 150,   0,   0,   0, 150,   0,   0, 150])
>>> data, label = dataset[0]
>>> data.shape
torch.Size([3, 32, 32])

prepare_data(download)[source]¶

Return type: None

data = None[source]¶

dir_path = None[source]¶

normalize = None[source]¶

targets = None[source]¶

pl_bolts.datamodules.cityscapes_datamodule module¶

class pl_bolts.datamodules.cityscapes_datamodule.CityscapesDataModule(data_dir, val_split=5000, num_workers=16, batch_size=32, seed=42, *args, **kwargs)[source]¶

Bases: pytorch_lightning.LightningDataModule

Standard Cityscapes, train, val, test splits and transforms

Specs:

30 classes (road, person, sidewalk, etc…)
(image, target) - image dims: (3 x 32 x 32), target dims: (3 x 32 x 32)

Transforms:

transforms = transform_lib.Compose([
    transform_lib.ToTensor(),
    transform_lib.Normalize(
        mean=[0.28689554, 0.32513303, 0.28389177],
        std=[0.18696375, 0.19017339, 0.18720214]
    )
])

Example:

from pl_bolts.datamodules import CityscapesDataModule

dm = CityscapesDataModule(PATH)
model = LitModel()

Trainer().fit(model, dm)

Or you can set your own transforms

Example:

dm.train_transforms = ...
dm.test_transforms = ...
dm.val_transforms  = ...

Parameters

data_dir – where to save/load the data
val_split – how many of the training images to use for the validation split
num_workers – how many workers to use for loading data
batch_size – number of examples per training/eval step

default_transforms()[source]¶

prepare_data()[source]¶: Saves Cityscapes files to data_dir

test_dataloader()[source]¶: Cityscapes test set uses the test split

train_dataloader()[source]¶: Cityscapes train set with removed subset to use for validation

val_dataloader()[source]¶: Cityscapes val set uses a subset of the training set for validation

extra_args = {}[source]¶

name = 'Cityscapes'[source]¶

property num_classes[source]¶: Return: 30

pl_bolts.datamodules.concat_dataset module¶

class pl_bolts.datamodules.concat_dataset.ConcatDataset(*datasets)[source]¶: Bases: torch.utils.data.Dataset

pl_bolts.datamodules.experience_source module¶

Datamodules for RL models that rely on experiences generated during training Based on implementations found here: https://github.com/Shmuma/ptan/blob/master/ptan/experience.py

class pl_bolts.datamodules.experience_source.BaseExperienceSource(env, agent)[source]¶

Bases: abc.ABC

Simplest form of the experience source :param _sphinx_paramlinks_pl_bolts.datamodules.experience_source.BaseExperienceSource.env: Environment that is being used :param _sphinx_paramlinks_pl_bolts.datamodules.experience_source.BaseExperienceSource.agent: Agent being used to make decisions

runner()[source]¶

Iterable method that yields steps from the experience source

Return type: Experience

class pl_bolts.datamodules.experience_source.DiscountedExperienceSource(env, agent, n_steps=1, gamma=0.99)[source]¶

Bases: pl_bolts.datamodules.experience_source.ExperienceSource

Outputs experiences with a discounted reward over N steps

discount_rewards(experiences)[source]¶

Calculates the discounted reward over N experiences :type _sphinx_paramlinks_pl_bolts.datamodules.experience_source.DiscountedExperienceSource.discount_rewards.experiences: Tuple[Experience] :param _sphinx_paramlinks_pl_bolts.datamodules.experience_source.DiscountedExperienceSource.discount_rewards.experiences: Tuple of Experience

Return type: float
Returns: total discounted reward

runner(device)[source]¶

Iterates through experience tuple and calculate discounted experience :type _sphinx_paramlinks_pl_bolts.datamodules.experience_source.DiscountedExperienceSource.runner.device: device :param _sphinx_paramlinks_pl_bolts.datamodules.experience_source.DiscountedExperienceSource.runner.device: current device to be used for executing experience steps

Yields: Discounted Experience
Return type: Experience

split_head_tail_exp(experiences)[source]¶

Takes in a tuple of experiences and returns the last state and tail experiences based on if the last state is the end of an episode :type _sphinx_paramlinks_pl_bolts.datamodules.experience_source.DiscountedExperienceSource.split_head_tail_exp.experiences: Tuple[Experience] :param _sphinx_paramlinks_pl_bolts.datamodules.experience_source.DiscountedExperienceSource.split_head_tail_exp.experiences: Tuple of N Experience

Return type: Tuple[List, Tuple[Experience]]
Returns: last state (Array or None) and remaining Experience

class pl_bolts.datamodules.experience_source.Experience(state, action, reward, done, new_state)[source]¶

Bases: tuple

Create new instance of Experience(state, action, reward, done, new_state)

_asdict()[source]¶: Return a new OrderedDict which maps field names to their values.

classmethod _make(iterable)[source]¶: Make a new Experience object from a sequence or iterable

_replace(**kwds)[source]¶: Return a new Experience object replacing specified fields with new values

_field_defaults = {}[source]¶

_fields = ('state', 'action', 'reward', 'done', 'new_state')[source]¶

_fields_defaults = {}[source]¶

property action[source]¶: Alias for field number 1

property done[source]¶: Alias for field number 3

property new_state[source]¶: Alias for field number 4

property reward[source]¶: Alias for field number 2

property state[source]¶: Alias for field number 0

class pl_bolts.datamodules.experience_source.ExperienceSource(env, agent, n_steps=1)[source]¶

Bases: pl_bolts.datamodules.experience_source.BaseExperienceSource

Experience source class handling single and multiple environment steps :param _sphinx_paramlinks_pl_bolts.datamodules.experience_source.ExperienceSource.env: Environment that is being used :param _sphinx_paramlinks_pl_bolts.datamodules.experience_source.ExperienceSource.agent: Agent being used to make decisions :type _sphinx_paramlinks_pl_bolts.datamodules.experience_source.ExperienceSource.n_steps: int :param _sphinx_paramlinks_pl_bolts.datamodules.experience_source.ExperienceSource.n_steps: Number of steps to return from each environment at once

env_actions(device)[source]¶: For each environment in the pool, get the correct action :rtype: List[List[int]] :returns: List of actions for each env, with size (num_envs, action_size)

env_step(env_idx, env, action)[source]¶

Carries out a step through the given environment using the given action :type _sphinx_paramlinks_pl_bolts.datamodules.experience_source.ExperienceSource.env_step.env_idx: int :param _sphinx_paramlinks_pl_bolts.datamodules.experience_source.ExperienceSource.env_step.env_idx: index of the current environment :type _sphinx_paramlinks_pl_bolts.datamodules.experience_source.ExperienceSource.env_step.env: Env :param _sphinx_paramlinks_pl_bolts.datamodules.experience_source.ExperienceSource.env_step.env: env at index env_idx :type _sphinx_paramlinks_pl_bolts.datamodules.experience_source.ExperienceSource.env_step.action: List[int] :param _sphinx_paramlinks_pl_bolts.datamodules.experience_source.ExperienceSource.env_step.action: action for this environment step

Return type: Experience
Returns: Experience tuple

init_environments()[source]¶

For each environment in the pool setups lists for tracking history of size n, state, current reward and current step

Return type: None

pop_rewards_steps()[source]¶: Returns the list of the current total rewards and steps collected :returns: list of total rewards and steps for all completed episodes for each environment since last pop

pop_total_rewards()[source]¶: Returns the list of the current total rewards collected :rtype: List[float] :returns: list of total rewards for all completed episodes for each environment since last pop

runner(device)[source]¶

Experience Source iterator yielding Tuple of experiences for n_steps. These come from the pool of environments provided by the user. :type _sphinx_paramlinks_pl_bolts.datamodules.experience_source.ExperienceSource.runner.device: device :param _sphinx_paramlinks_pl_bolts.datamodules.experience_source.ExperienceSource.runner.device: current device to be used for executing experience steps

Return type: Tuple[Experience]
Returns: Tuple of Experiences

update_env_stats(env_idx)[source]¶

To be called at the end of the history tail generation during the termination state. Updates the stats tracked for all environments :type _sphinx_paramlinks_pl_bolts.datamodules.experience_source.ExperienceSource.update_env_stats.env_idx: int :param _sphinx_paramlinks_pl_bolts.datamodules.experience_source.ExperienceSource.update_env_stats.env_idx: index of the environment used to update stats

Return type: None

update_history_queue(env_idx, exp, history)[source]¶

Updates the experience history queue with the lastest experiences. In the event of an experience step is in the done state, the history will be incrementally appended to the queue, removing the tail of the history each time. :param _sphinx_paramlinks_pl_bolts.datamodules.experience_source.ExperienceSource.update_history_queue.env_idx: index of the environment :param _sphinx_paramlinks_pl_bolts.datamodules.experience_source.ExperienceSource.update_history_queue.exp: the current experience :param _sphinx_paramlinks_pl_bolts.datamodules.experience_source.ExperienceSource.update_history_queue.history: history of experience steps for this environment

Return type: None

class pl_bolts.datamodules.experience_source.ExperienceSourceDataset(generate_batch)[source]¶

Bases: torch.utils.data.IterableDataset

Basic experience source dataset. Takes a generate_batch function that returns an iterator. The logic for the experience source and how the batch is generated is defined the Lightning model itself

pl_bolts.datamodules.fashion_mnist_datamodule module¶

class pl_bolts.datamodules.fashion_mnist_datamodule.FashionMNISTDataModule(data_dir, val_split=5000, num_workers=16, seed=42, *args, **kwargs)[source]¶

Bases: pytorch_lightning.LightningDataModule

Specs:

10 classes (1 per type)
Each image is (1 x 28 x 28)

Standard FashionMNIST, train, val, test splits and transforms

Transforms:

mnist_transforms = transform_lib.Compose([
    transform_lib.ToTensor()
])

Example:

from pl_bolts.datamodules import FashionMNISTDataModule

dm = FashionMNISTDataModule('.')
model = LitModel()

Trainer().fit(model, dm)

Parameters

data_dir (str) – where to save/load the data
val_split (int) – how many of the training images to use for the validation split
num_workers (int) – how many workers to use for loading data

_default_transforms()[source]¶

prepare_data()[source]¶: Saves FashionMNIST files to data_dir

test_dataloader(batch_size=32, transforms=None)[source]¶

FashionMNIST test set uses the test split

Parameters

batch_size – size of batch
transforms – custom transforms

train_dataloader(batch_size=32, transforms=None)[source]¶

FashionMNIST train set removes a subset to use for validation

Parameters

batch_size – size of batch
transforms – custom transforms

val_dataloader(batch_size=32, transforms=None)[source]¶

FashionMNIST val set uses a subset of the training set for validation

Parameters

batch_size – size of batch
transforms – custom transforms

name = 'fashion_mnist'[source]¶

property num_classes[source]¶: Return: 10

pl_bolts.datamodules.imagenet_datamodule module¶

class pl_bolts.datamodules.imagenet_datamodule.ImagenetDataModule(data_dir, meta_dir=None, num_imgs_per_val_class=50, image_size=224, num_workers=16, batch_size=32, *args, **kwargs)[source]¶

Bases: pytorch_lightning.LightningDataModule

Specs:

1000 classes
Each image is (3 x varies x varies) (here we default to 3 x 224 x 224)

Imagenet train, val and test dataloaders.

The train set is the imagenet train.

The val set is taken from the train set with num_imgs_per_val_class images per class. For example if num_imgs_per_val_class=2 then there will be 2,000 images in the validation set.

The test set is the official imagenet validation set.

Example:

from pl_bolts.datamodules import ImagenetDataModule

dm = ImagenetDataModule(IMAGENET_PATH)
model = LitModel()

Trainer().fit(model, dm)

Parameters

data_dir (str) – path to the imagenet dataset file
meta_dir (Optional[str]) – path to meta.bin file
num_imgs_per_val_class (int) – how many images per class for the validation set
image_size (int) – final image size
num_workers (int) – how many data workers
batch_size (int) – batch_size

_verify_splits(data_dir, split)[source]¶

prepare_data()[source]¶: This method already assumes you have imagenet2012 downloaded. It validates the data using the meta.bin.

Warning

Please download imagenet on your own first.

test_dataloader()[source]¶: Uses the validation split of imagenet2012 for testing

train_dataloader()[source]¶: Uses the train split of imagenet2012 and puts away a portion of it for the validation split

train_transform()[source]¶

The standard imagenet transforms

transform_lib.Compose([
    transform_lib.RandomResizedCrop(self.image_size),
    transform_lib.RandomHorizontalFlip(),
    transform_lib.ToTensor(),
    transform_lib.Normalize(
        mean=[0.485, 0.456, 0.406],
        std=[0.229, 0.224, 0.225]
    ),
])

val_dataloader()[source]¶

Uses the part of the train split of imagenet2012 that was not used for training via num_imgs_per_val_class

Parameters

batch_size – the batch size
transforms – the transforms

val_transform()[source]¶

The standard imagenet transforms for validation

transform_lib.Compose([
    transform_lib.Resize(self.image_size + 32),
    transform_lib.CenterCrop(self.image_size),
    transform_lib.ToTensor(),
    transform_lib.Normalize(
        mean=[0.485, 0.456, 0.406],
        std=[0.229, 0.224, 0.225]
    ),
])

name = 'imagenet'[source]¶

property num_classes[source]¶

Return:

1000

pl_bolts.datamodules.imagenet_dataset module¶

class pl_bolts.datamodules.imagenet_dataset.UnlabeledImagenet(root, split='train', num_classes=-1, num_imgs_per_class=-1, num_imgs_per_class_val_split=50, meta_dir=None, **kwargs)[source]¶

Bases: torchvision.datasets.ImageNet

Official train set gets split into train, val. (using nb_imgs_per_val_class for each class). Official validation becomes test set

Within each class, we further allow limiting the number of samples per class (for semi-sup lng)

Parameters

root – path of dataset
split (str) –
num_classes (int) – Sets the limit of classes
num_imgs_per_class (int) – Limits the number of images per class
num_imgs_per_class_val_split (int) – How many images per class to generate the val split
download –
kwargs –

classmethod generate_meta_bins(devkit_dir)[source]¶

partition_train_set(imgs, nb_imgs_in_val)[source]¶

pl_bolts.datamodules.imagenet_dataset._calculate_md5(fpath, chunk_size=1048576)[source]¶

pl_bolts.datamodules.imagenet_dataset._check_integrity(fpath, md5=None)[source]¶

pl_bolts.datamodules.imagenet_dataset._check_md5(fpath, md5, **kwargs)[source]¶

pl_bolts.datamodules.imagenet_dataset._is_gzip(filename)[source]¶

pl_bolts.datamodules.imagenet_dataset._is_tar(filename)[source]¶

pl_bolts.datamodules.imagenet_dataset._is_targz(filename)[source]¶

pl_bolts.datamodules.imagenet_dataset._is_tarxz(filename)[source]¶

pl_bolts.datamodules.imagenet_dataset._is_zip(filename)[source]¶

pl_bolts.datamodules.imagenet_dataset._verify_archive(root, file, md5)[source]¶

pl_bolts.datamodules.imagenet_dataset.extract_archive(from_path, to_path=None, remove_finished=False)[source]¶

pl_bolts.datamodules.imagenet_dataset.parse_devkit_archive(root, file=None)[source]¶

Parse the devkit archive of the ImageNet2012 classification dataset and save the meta information in a binary file.

Parameters

root (str) – Root directory containing the devkit archive
file (str, optional) – Name of devkit archive. Defaults to ‘ILSVRC2012_devkit_t12.tar.gz’

pl_bolts.datamodules.kitti_datamodule module¶

class pl_bolts.datamodules.kitti_datamodule.KittiDataModule(data_dir, val_split=0.2, test_split=0.1, num_workers=16, batch_size=32, seed=42, *args, **kwargs)[source]¶

Bases: pytorch_lightning.LightningDataModule

Kitti train, validation and test dataloaders.

Note: You need to have downloaded the Kitti dataset first and provide the path to where it is saved. You can download the dataset here: http://www.cvlibs.net/datasets/kitti/eval_semseg.php?benchmark=semantics2015

Specs:

200 samples
Each image is (3 x 1242 x 376)

In total there are 34 classes but some of these are not useful so by default we use only 19 of the classes specified by the valid_labels parameter.

Example:

from pl_bolts.datamodules import KittiDataModule

dm = KittiDataModule(PATH)
model = LitModel()

Trainer().fit(model, dm)

Args::: data_dir: where to load the data from path, i.e. ‘/path/to/folder/with/data_semantics/’ val_split: size of validation test (default 0.2) test_split: size of test set (default 0.1) num_workers: how many workers to use for loading data batch_size: the batch size seed: random seed to be used for train/val/test splits

test_dataloader()[source]¶

train_dataloader()[source]¶

val_dataloader()[source]¶

name = 'kitti'[source]¶

pl_bolts.datamodules.kitti_dataset module¶

class pl_bolts.datamodules.kitti_dataset.KittiDataset(data_dir, img_size=(1242, 376), void_labels=(0, 1, 2, 3, 4, 5, 6, 9, 10, 14, 15, 16, 18, 29, 30, -1), valid_labels=(7, 8, 11, 12, 13, 17, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 31, 32, 33), transform=None)[source]¶

Bases: torch.utils.data.Dataset

Note: You need to have downloaded the Kitti dataset first and provide the path to where it is saved. You can download the dataset here: http://www.cvlibs.net/datasets/kitti/eval_semseg.php?benchmark=semantics2015

There are 34 classes, however not all of them are useful for training (e.g. railings on highways). These useless classes (the pixel values of these classes) are stored in void_labels. Useful classes are stored in valid_labels.

The encode_segmap function sets all pixels with any of the void_labels to ignore_index (250 by default). It also sets all of the valid pixels to the appropriate value between 0 and len(valid_labels) (since that is the number of valid classes), so it can be used properly by the loss function when comparing with the output.

Parameters

data_dir (str) – where to load the data from path, i.e. ‘/path/to/folder/with/data_semantics/’
img_size (tuple) – image dimensions (width, height)
void_labels (list) – useless classes to be excluded from training
valid_labels (list) – useful classes to include

encode_segmap(mask)[source]¶: Sets void classes to zero so they won’t be considered for training

get_filenames(path)[source]¶: Returns a list of absolute paths to images inside given path

IMAGE_PATH = 'training/image_2'[source]¶

MASK_PATH = 'training/semantic'[source]¶

pl_bolts.datamodules.mnist_datamodule module¶

class pl_bolts.datamodules.mnist_datamodule.MNISTDataModule(data_dir='./', val_split=5000, num_workers=16, normalize=False, seed=42, batch_size=32, *args, **kwargs)[source]¶

Bases: pytorch_lightning.LightningDataModule

Specs:

10 classes (1 per digit)
Each image is (1 x 28 x 28)

Standard MNIST, train, val, test splits and transforms

Transforms:

mnist_transforms = transform_lib.Compose([
    transform_lib.ToTensor()
])

Example:

from pl_bolts.datamodules import MNISTDataModule

dm = MNISTDataModule('.')
model = LitModel()

Trainer().fit(model, dm)

Parameters

data_dir (str) – where to save/load the data
val_split (int) – how many of the training images to use for the validation split
num_workers (int) – how many workers to use for loading data
normalize (bool) – If true applies image normalize

_default_transforms()[source]¶

prepare_data()[source]¶: Saves MNIST files to data_dir

test_dataloader(batch_size=32, transforms=None)[source]¶

MNIST test set uses the test split

Parameters

batch_size – size of batch
transforms – custom transforms

train_dataloader(batch_size=32, transforms=None)[source]¶

MNIST train set removes a subset to use for validation

Parameters

batch_size – size of batch
transforms – custom transforms

val_dataloader(batch_size=32, transforms=None)[source]¶

MNIST val set uses a subset of the training set for validation

Parameters

batch_size – size of batch
transforms – custom transforms

name = 'mnist'[source]¶

property num_classes[source]¶: Return: 10

pl_bolts.datamodules.mnist_dataset module¶

class pl_bolts.datamodules.mnist_dataset.BinaryMNIST(*args, **kwargs)[source]¶: Bases: torchvision.datasets.MNIST

pl_bolts.datamodules.sklearn_datamodule module¶

class pl_bolts.datamodules.sklearn_datamodule.SklearnDataModule(X, y, x_val=None, y_val=None, x_test=None, y_test=None, val_split=0.2, test_split=0.1, num_workers=2, random_state=1234, shuffle=True, *args, **kwargs)[source]¶

Bases: pytorch_lightning.LightningDataModule

Automatically generates the train, validation and test splits for a Numpy dataset. They are set up as dataloaders for convenience. Optionally, you can pass in your own validation and test splits.

Example

>>> from sklearn.datasets import load_boston
>>> from pl_bolts.datamodules import SklearnDataModule
...
>>> X, y = load_boston(return_X_y=True)
>>> loaders = SklearnDataModule(X, y)
...
>>> # train set
>>> train_loader = loaders.train_dataloader(batch_size=32)
>>> len(train_loader.dataset)
355
>>> len(train_loader)
11
>>> # validation set
>>> val_loader = loaders.val_dataloader(batch_size=32)
>>> len(val_loader.dataset)
100
>>> len(val_loader)
3
>>> # test set
>>> test_loader = loaders.test_dataloader(batch_size=32)
>>> len(test_loader.dataset)
51
>>> len(test_loader)
1

_init_datasets(X, y, x_val, y_val, x_test, y_test)[source]¶

test_dataloader(batch_size=16)[source]¶

train_dataloader(batch_size=16)[source]¶

val_dataloader(batch_size=16)[source]¶

name = 'sklearn'[source]¶

class pl_bolts.datamodules.sklearn_datamodule.SklearnDataset(X, y, X_transform=None, y_transform=None)[source]¶

Bases: torch.utils.data.Dataset

Mapping between numpy (or sklearn) datasets to PyTorch datasets.

Parameters

X (ndarray) – Numpy ndarray
y (ndarray) – Numpy ndarray
X_transform (Optional[Any]) – Any transform that works with Numpy arrays
y_transform (Optional[Any]) – Any transform that works with Numpy arrays

Example

>>> from sklearn.datasets import load_boston
>>> from pl_bolts.datamodules import SklearnDataset
...
>>> X, y = load_boston(return_X_y=True)
>>> dataset = SklearnDataset(X, y)
>>> len(dataset)
506

class pl_bolts.datamodules.sklearn_datamodule.TensorDataset(X, y, X_transform=None, y_transform=None)[source]¶

Bases: torch.utils.data.Dataset

Prepare PyTorch tensor dataset for data loaders.

Parameters

X (Tensor) – PyTorch tensor
y (Tensor) – PyTorch tensor
X_transform (Optional[Any]) – Any transform that works with PyTorch tensors
y_transform (Optional[Any]) – Any transform that works with PyTorch tensors

Example

>>> from pl_bolts.datamodules import TensorDataset
...
>>> X = torch.rand(10, 3)
>>> y = torch.rand(10)
>>> dataset = TensorDataset(X, y)
>>> len(dataset)
10

pl_bolts.datamodules.ssl_amdim_datasets module¶

class pl_bolts.datamodules.ssl_amdim_datasets.CIFAR10Mixed(root, split='val', transform=None, target_transform=None, download=False, nb_labeled_per_class=None, val_pct=0.1)[source]¶: Bases: pl_bolts.datamodules.ssl_amdim_datasets.SSLDatasetMixin, torchvision.datasets.CIFAR10

class pl_bolts.datamodules.ssl_amdim_datasets.SSLDatasetMixin[source]¶

Bases: abc.ABC

classmethod deterministic_shuffle(x, y)[source]¶

classmethod generate_train_val_split(examples, labels, pct_val)[source]¶: Splits dataset uniformly across classes :param _sphinx_paramlinks_pl_bolts.datamodules.ssl_amdim_datasets.SSLDatasetMixin.generate_train_val_split.examples: :param _sphinx_paramlinks_pl_bolts.datamodules.ssl_amdim_datasets.SSLDatasetMixin.generate_train_val_split.labels: :param _sphinx_paramlinks_pl_bolts.datamodules.ssl_amdim_datasets.SSLDatasetMixin.generate_train_val_split.pct_val: :return:

classmethod select_nb_imgs_per_class(examples, labels, nb_imgs_in_val)[source]¶: Splits a dataset into two parts. The labeled split has nb_imgs_in_val per class :param _sphinx_paramlinks_pl_bolts.datamodules.ssl_amdim_datasets.SSLDatasetMixin.select_nb_imgs_per_class.examples: :param _sphinx_paramlinks_pl_bolts.datamodules.ssl_amdim_datasets.SSLDatasetMixin.select_nb_imgs_per_class.labels: :param _sphinx_paramlinks_pl_bolts.datamodules.ssl_amdim_datasets.SSLDatasetMixin.select_nb_imgs_per_class.nb_imgs_in_val: :return:

pl_bolts.datamodules.ssl_imagenet_datamodule module¶

class pl_bolts.datamodules.ssl_imagenet_datamodule.SSLImagenetDataModule(data_dir, meta_dir=None, num_workers=16, *args, **kwargs)[source]¶

Bases: pytorch_lightning.LightningDataModule

_default_transforms()[source]¶

_verify_splits(data_dir, split)[source]¶

prepare_data()[source]¶

test_dataloader(batch_size, num_images_per_class, add_normalize=False)[source]¶

train_dataloader(batch_size, num_images_per_class=-1, add_normalize=False)[source]¶

val_dataloader(batch_size, num_images_per_class=50, add_normalize=False)[source]¶

name = 'imagenet'[source]¶

property num_classes[source]¶

pl_bolts.datamodules.stl10_datamodule module¶

class pl_bolts.datamodules.stl10_datamodule.STL10DataModule(data_dir=None, unlabeled_val_split=5000, train_val_split=500, num_workers=16, batch_size=32, seed=42, *args, **kwargs)[source]¶

Bases: pytorch_lightning.LightningDataModule

Specs:

10 classes (1 per type)
Each image is (3 x 96 x 96)

Standard STL-10, train, val, test splits and transforms. STL-10 has support for doing validation splits on the labeled or unlabeled splits

Transforms:

mnist_transforms = transform_lib.Compose([
    transform_lib.ToTensor(),
    transforms.Normalize(
        mean=(0.43, 0.42, 0.39),
        std=(0.27, 0.26, 0.27)
    )
])

Example:

from pl_bolts.datamodules import STL10DataModule

dm = STL10DataModule(PATH)
model = LitModel()

Trainer().fit(model, dm)

Parameters

data_dir (Optional[str]) – where to save/load the data
unlabeled_val_split (int) – how many images from the unlabeled training split to use for validation
train_val_split (int) – how many images from the labeled training split to use for validation
num_workers (int) – how many workers to use for loading data
batch_size (int) – the batch size

default_transforms()[source]¶

prepare_data()[source]¶: Downloads the unlabeled, train and test split

test_dataloader()[source]¶

Loads the test split of STL10

Parameters

batch_size – the batch size
transforms – the transforms

train_dataloader()[source]¶: Loads the ‘unlabeled’ split minus a portion set aside for validation via unlabeled_val_split.

train_dataloader_labeled()[source]¶

train_dataloader_mixed()[source]¶

Loads a portion of the ‘unlabeled’ training data and ‘train’ (labeled) data. both portions have a subset removed for validation via unlabeled_val_split and train_val_split

Parameters

batch_size – the batch size
transforms – a sequence of transforms

val_dataloader()[source]¶

Loads a portion of the ‘unlabeled’ training data set aside for validation The val dataset = (unlabeled - train_val_split)

Parameters

batch_size – the batch size
transforms – a sequence of transforms

val_dataloader_labeled()[source]¶

val_dataloader_mixed()[source]¶

Loads a portion of the ‘unlabeled’ training data set aside for validation along with the portion of the ‘train’ dataset to be used for validation

unlabeled_val = (unlabeled - train_val_split)

labeled_val = (train- train_val_split)

full_val = unlabeled_val + labeled_val

Parameters

batch_size – the batch size
transforms – a sequence of transforms

name = 'stl10'[source]¶

property num_classes[source]¶

pl_bolts.datamodules.vocdetection_datamodule module¶

class pl_bolts.datamodules.vocdetection_datamodule.Compose(transforms)[source]¶

Bases: object

Like torchvision.transforms.compose but works for (image, target)

__call__(image, target)[source]¶: Call self as a function.

class pl_bolts.datamodules.vocdetection_datamodule.VOCDetectionDataModule(data_dir, year='2012', num_workers=16, normalize=False, *args, **kwargs)[source]¶

Bases: pytorch_lightning.LightningDataModule

TODO(teddykoker) docstring

_default_transforms()[source]¶

prepare_data()[source]¶: Saves VOCDetection files to data_dir

train_dataloader(batch_size=1, transforms=None)[source]¶

VOCDetection train set uses the train subset

Parameters

batch_size – size of batch
transforms – custom transforms

val_dataloader(batch_size=1, transforms=None)[source]¶

VOCDetection val set uses the val subset

Parameters

batch_size – size of batch
transforms – custom transforms

name = 'vocdetection'[source]¶

property num_classes[source]¶: Return: 21

pl_bolts.datamodules.vocdetection_datamodule._collate_fn(batch)[source]¶

pl_bolts.datamodules.vocdetection_datamodule._prepare_voc_instance(image, target)[source]¶

Prepares VOC dataset into appropriate target for fasterrcnn

https://github.com/pytorch/vision/issues/1097#issuecomment-508917489

pl_bolts.metrics package¶

Submodules¶

pl_bolts.metrics.aggregation module¶

pl_bolts.metrics.aggregation.accuracy(preds, labels)[source]¶

pl_bolts.metrics.aggregation.mean(res, key)[source]¶

pl_bolts.metrics.aggregation.precision_at_k(output, target, top_k=(1, ))[source]¶: Computes the accuracy over the k top predictions for the specified values of k

pl_bolts.models package¶

Collection of PyTorchLightning models

Subpackages¶

pl_bolts.models.autoencoders package¶

Here are a VAE and GAN

Subpackages¶

pl_bolts.models.autoencoders.basic_ae package¶

AE Template¶

This is a basic template for implementing an Autoencoder in PyTorch Lightning.

A default encoder and decoder have been provided but can easily be replaced by custom models.

This template uses the CIFAR10 dataset but image data of any dimension can be fed in as long as the image: width and image height are even values. For other types of data, such as sound, it will be necessary to change the Encoder and Decoder.
The default encoder is a resnet18 backbone followed by linear layers which map representations to latent space.: The default decoder mirrors the encoder architecture and is similar to an inverted resnet18.

from pl_bolts.models.autoencoders import AE

model = AE()
trainer = pl.Trainer()
trainer.fit(model)

Submodules¶

pl_bolts.models.autoencoders.basic_ae.basic_ae_module module¶

class pl_bolts.models.autoencoders.basic_ae.basic_ae_module.AE(input_height, enc_type='resnet18', first_conv=False, maxpool1=False, enc_out_dim=512, kl_coeff=0.1, latent_dim=256, lr=0.0001, **kwargs)[source]¶

Bases: pytorch_lightning.LightningModule

Standard AE

Model is available pretrained on different datasets:

Example:

# not pretrained
ae = AE()

# pretrained on cifar10
ae = AE.from_pretrained('cifar10-resnet18')

Parameters

input_height – height of the images
enc_type – option between resnet18 or resnet50
first_conv – use standard kernel_size 7, stride 2 at start or replace it with kernel_size 3, stride 1 conv
maxpool1 – use standard maxpool to reduce spatial dim of feat by a factor of 2
enc_out_dim – set according to the out_channel count of encoder used (512 for resnet18, 2048 for resnet50)
latent_dim – dim of latent space
lr – learning rate for Adam

static add_model_specific_args(parent_parser)[source]¶

configure_optimizers()[source]¶

forward(x)[source]¶

from_pretrained(checkpoint_name)[source]¶

static pretrained_weights_available()[source]¶

step(batch, batch_idx)[source]¶

training_step(batch, batch_idx)[source]¶

validation_step(batch, batch_idx)[source]¶

pretrained_urls = {'cifar10-resnet18': 'https://pl-bolts-weights.s3.us-east-2.amazonaws.com/ae/ae-cifar10/checkpoints/epoch%3D96.ckpt'}[source]¶

pl_bolts.models.autoencoders.basic_ae.basic_ae_module.cli_main(args=None)[source]¶

pl_bolts.models.autoencoders.basic_vae package¶

VAE Template¶

This is a basic template for implementing a Variational Autoencoder in PyTorch Lightning.

A default encoder and decoder have been provided but can easily be replaced by custom models.

This template uses the CIFAR10 dataset but image data of any dimension can be fed in as long as the image: width and image height are even values. For other types of data, such as sound, it will be necessary to change the Encoder and Decoder.
The default encoder is a resnet18 backbone followed by linear layers which map representations: to mu and var. The default decoder mirrors the encoder architecture and is similar to an inverted resnet18. The model also assumes a Gaussian prior and a Gaussian approximate posterior distribution.

Submodules¶

pl_bolts.models.autoencoders.basic_vae.basic_vae_module module¶

class pl_bolts.models.autoencoders.basic_vae.basic_vae_module.VAE(input_height, enc_type='resnet18', first_conv=False, maxpool1=False, enc_out_dim=512, kl_coeff=0.1, latent_dim=256, lr=0.0001, **kwargs)[source]¶

Bases: pytorch_lightning.LightningModule

Standard VAE with Gaussian Prior and approx posterior.

Model is available pretrained on different datasets:

Example:

# not pretrained
vae = VAE()

# pretrained on cifar10
vae = VAE.from_pretrained('cifar10-resnet18')

# pretrained on stl10
vae = VAE.from_pretrained('stl10-resnet18')

Parameters

input_height – height of the images
enc_type – option between resnet18 or resnet50
first_conv – use standard kernel_size 7, stride 2 at start or replace it with kernel_size 3, stride 1 conv
maxpool1 – use standard maxpool to reduce spatial dim of feat by a factor of 2
enc_out_dim – set according to the out_channel count of encoder used (512 for resnet18, 2048 for resnet50)
kl_coeff – coefficient for kl term of the loss
latent_dim – dim of latent space
lr – learning rate for Adam

_run_step(x)[source]¶

static add_model_specific_args(parent_parser)[source]¶

configure_optimizers()[source]¶

forward(x)[source]¶

from_pretrained(checkpoint_name)[source]¶

static pretrained_weights_available()[source]¶

sample(mu, log_var)[source]¶

step(batch, batch_idx)[source]¶

training_step(batch, batch_idx)[source]¶

validation_step(batch, batch_idx)[source]¶

pretrained_urls = {'cifar10-resnet18': 'https://pl-bolts-weights.s3.us-east-2.amazonaws.com/vae/vae-cifar10/checkpoints/epoch%3D89.ckpt', 'stl10-resnet18': 'https://pl-bolts-weights.s3.us-east-2.amazonaws.com/vae/vae-stl10/checkpoints/epoch%3D89.ckpt'}[source]¶

pl_bolts.models.autoencoders.basic_vae.basic_vae_module.cli_main(args=None)[source]¶

Submodules¶

pl_bolts.models.autoencoders.components module¶

class pl_bolts.models.autoencoders.components.DecoderBlock(inplanes, planes, scale=1, upsample=None)[source]¶

Bases: torch.nn.Module

ResNet block, but convs replaced with resize convs, and channel increase is in second conv, not first

forward(x)[source]¶

expansion = 1[source]¶

class pl_bolts.models.autoencoders.components.DecoderBottleneck(inplanes, planes, scale=1, upsample=None)[source]¶

Bases: torch.nn.Module

ResNet bottleneck, but convs replaced with resize convs

forward(x)[source]¶

expansion = 4[source]¶

class pl_bolts.models.autoencoders.components.EncoderBlock(inplanes, planes, stride=1, downsample=None)[source]¶

Bases: torch.nn.Module

ResNet block, copied from https://github.com/pytorch/vision/blob/master/torchvision/models/resnet.py#L35

forward(x)[source]¶

expansion = 1[source]¶

class pl_bolts.models.autoencoders.components.EncoderBottleneck(inplanes, planes, stride=1, downsample=None)[source]¶

Bases: torch.nn.Module

ResNet bottleneck, copied from https://github.com/pytorch/vision/blob/master/torchvision/models/resnet.py#L75

forward(x)[source]¶

expansion = 4[source]¶

class pl_bolts.models.autoencoders.components.Interpolate(size=None, scale_factor=None)[source]¶

Bases: torch.nn.Module

nn.Module wrapper for F.interpolate

forward(x)[source]¶

class pl_bolts.models.autoencoders.components.ResNetDecoder(block, layers, latent_dim, input_height, first_conv=False, maxpool1=False)[source]¶

Bases: torch.nn.Module

Resnet in reverse order

_make_layer(block, planes, blocks, scale=1)[source]¶

forward(x)[source]¶

class pl_bolts.models.autoencoders.components.ResNetEncoder(block, layers, first_conv=False, maxpool1=False)[source]¶

Bases: torch.nn.Module

_make_layer(block, planes, blocks, stride=1)[source]¶

forward(x)[source]¶

pl_bolts.models.autoencoders.components.conv1x1(in_planes, out_planes, stride=1)[source]¶: 1x1 convolution

pl_bolts.models.autoencoders.components.conv3x3(in_planes, out_planes, stride=1)[source]¶: 3x3 convolution with padding

pl_bolts.models.autoencoders.components.resize_conv1x1(in_planes, out_planes, scale=1)[source]¶: upsample + 1x1 convolution with padding to avoid checkerboard artifact

pl_bolts.models.autoencoders.components.resize_conv3x3(in_planes, out_planes, scale=1)[source]¶: upsample + 3x3 convolution with padding to avoid checkerboard artifact

pl_bolts.models.autoencoders.components.resnet18_decoder(latent_dim, input_height, first_conv, maxpool1)[source]¶

pl_bolts.models.autoencoders.components.resnet18_encoder(first_conv, maxpool1)[source]¶

pl_bolts.models.autoencoders.components.resnet50_decoder(latent_dim, input_height, first_conv, maxpool1)[source]¶

pl_bolts.models.autoencoders.components.resnet50_encoder(first_conv, maxpool1)[source]¶

pl_bolts.models.detection package¶

class pl_bolts.models.detection.FasterRCNN(learning_rate=0.0001, num_classes=91, pretrained=False, pretrained_backbone=True, trainable_backbone_layers=3, replace_head=True, **kwargs)[source]¶

Bases: pytorch_lightning.LightningModule

PyTorch Lightning implementation of Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks.

Paper authors: Shaoqing Ren, Kaiming He, Ross Girshick, Jian Sun

Model implemented by:

Teddy Koker <https://github.com/teddykoker>

During training, the model expects both the input tensors, as well as targets (list of dictionary), containing:

boxes (FloatTensor[N, 4]): the ground truth boxes in [x1, y1, x2, y2] format.
labels (Int64Tensor[N]): the class label for each ground truh box

CLI command:

# PascalVOC
python faster_rcnn.py --gpus 1 --pretrained True

Parameters

learning_rate (float) – the learning rate
num_classes (int) – number of detection classes (including background)
pretrained (bool) – if true, returns a model pre-trained on COCO train2017
pretrained_backbone (bool) – if true, returns a model with backbone pre-trained on Imagenet
trainable_backbone_layers (int) – number of trainable resnet layers starting from final block

static add_model_specific_args(parent_parser)[source]¶

configure_optimizers()[source]¶

forward(x)[source]¶

training_step(batch, batch_idx)[source]¶

validation_epoch_end(outs)[source]¶

validation_step(batch, batch_idx)[source]¶

Submodules¶

pl_bolts.models.detection.faster_rcnn module¶

class pl_bolts.models.detection.faster_rcnn.FasterRCNN(learning_rate=0.0001, num_classes=91, pretrained=False, pretrained_backbone=True, trainable_backbone_layers=3, replace_head=True, **kwargs)[source]¶

Bases: pytorch_lightning.LightningModule

PyTorch Lightning implementation of Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks.

Paper authors: Shaoqing Ren, Kaiming He, Ross Girshick, Jian Sun

Model implemented by:

Teddy Koker <https://github.com/teddykoker>

During training, the model expects both the input tensors, as well as targets (list of dictionary), containing:

boxes (FloatTensor[N, 4]): the ground truth boxes in [x1, y1, x2, y2] format.
labels (Int64Tensor[N]): the class label for each ground truh box

CLI command:

# PascalVOC
python faster_rcnn.py --gpus 1 --pretrained True

Parameters

learning_rate (float) – the learning rate
num_classes (int) – number of detection classes (including background)
pretrained (bool) – if true, returns a model pre-trained on COCO train2017
pretrained_backbone (bool) – if true, returns a model with backbone pre-trained on Imagenet
trainable_backbone_layers (int) – number of trainable resnet layers starting from final block

static add_model_specific_args(parent_parser)[source]¶

configure_optimizers()[source]¶

forward(x)[source]¶

training_step(batch, batch_idx)[source]¶

validation_epoch_end(outs)[source]¶

validation_step(batch, batch_idx)[source]¶

pl_bolts.models.detection.faster_rcnn._evaluate_iou(target, pred)[source]¶: Evaluate intersection over union (IOU) for target from dataset and output prediction from model

pl_bolts.models.detection.faster_rcnn.run_cli()[source]¶

pl_bolts.models.gans package¶

Subpackages¶

pl_bolts.models.gans.basic package¶

Submodules¶

pl_bolts.models.gans.basic.basic_gan_module module¶

class pl_bolts.models.gans.basic.basic_gan_module.GAN(input_channels, input_height, input_width, latent_dim=32, learning_rate=0.0002, **kwargs)[source]¶

Bases: pytorch_lightning.LightningModule

Vanilla GAN implementation.

Example:

from pl_bolts.models.gan import GAN

m = GAN()
Trainer(gpus=2).fit(m)

Example CLI:

# mnist
python  basic_gan_module.py --gpus 1

# imagenet
python  basic_gan_module.py --gpus 1 --dataset 'imagenet2012'
--data_dir /path/to/imagenet/folder/ --meta_dir ~/path/to/meta/bin/folder
--batch_size 256 --learning_rate 0.0001

Parameters

datamodule – the datamodule (train, val, test splits)
latent_dim (int) – emb dim for encoder
batch_size – the batch size
learning_rate (float) – the learning rate
data_dir – where to store data
num_workers – data workers

static add_model_specific_args(parent_parser)[source]¶

configure_optimizers()[source]¶

discriminator_loss(x)[source]¶

discriminator_step(x)[source]¶

forward(z)[source]¶

Generates an image given input noise z

Example:

z = torch.rand(batch_size, latent_dim)
gan = GAN.load_from_checkpoint(PATH)
img = gan(z)

generator_loss(x)[source]¶

generator_step(x)[source]¶

init_discriminator(img_dim)[source]¶

init_generator(img_dim)[source]¶

training_step(batch, batch_idx, optimizer_idx)[source]¶

pl_bolts.models.gans.basic.basic_gan_module.cli_main(args=None)[source]¶

pl_bolts.models.gans.basic.components module¶

class pl_bolts.models.gans.basic.components.Discriminator(img_shape, hidden_dim=1024)[source]¶

Bases: torch.nn.Module

forward(img)[source]¶

class pl_bolts.models.gans.basic.components.Generator(latent_dim, img_shape, hidden_dim=256)[source]¶

Bases: torch.nn.Module

forward(z)[source]¶

pl_bolts.models.regression package¶

Submodules¶

pl_bolts.models.regression.linear_regression module¶

class pl_bolts.models.regression.linear_regression.LinearRegression(input_dim, output_dim=1, bias=True, learning_rate=0.0001, optimizer=torch.optim.Adam, l1_strength=0.0, l2_strength=0.0, **kwargs)[source]¶

Bases: pytorch_lightning.LightningModule

Linear regression model implementing - with optional L1/L2 regularization $$min_{W} ||(Wx + b) - y ||_2^2 $$

Parameters

input_dim (int) – number of dimensions of the input (1+)
output_dim (int) – number of dimensions of the output (default=1)
bias (bool) – If false, will not use $+b$
learning_rate (float) – learning_rate for the optimizer
optimizer (Optimizer) – the optimizer to use (default=’Adam’)
l1_strength (float) – L1 regularization strength (default=None)
l2_strength (float) – L2 regularization strength (default=None)

static add_model_specific_args(parent_parser)[source]¶

configure_optimizers()[source]¶

forward(x)[source]¶

test_epoch_end(outputs)[source]¶

test_step(batch, batch_idx)[source]¶

training_step(batch, batch_idx)[source]¶

validation_epoch_end(outputs)[source]¶

validation_step(batch, batch_idx)[source]¶

pl_bolts.models.regression.linear_regression.cli_main()[source]¶

pl_bolts.models.regression.logistic_regression module¶

class pl_bolts.models.regression.logistic_regression.LogisticRegression(input_dim, num_classes, bias=True, learning_rate=0.0001, optimizer=torch.optim.Adam, l1_strength=0.0, l2_strength=0.0, **kwargs)[source]¶

Bases: pytorch_lightning.LightningModule

Logistic regression model

Parameters

input_dim (int) – number of dimensions of the input (at least 1)
num_classes (int) – number of class labels (binary: 2, multi-class: >2)
bias (bool) – specifies if a constant or intercept should be fitted (equivalent to fit_intercept in sklearn)
learning_rate (float) – learning_rate for the optimizer
optimizer (Optimizer) – the optimizer to use (default=’Adam’)
l1_strength (float) – L1 regularization strength (default=None)
l2_strength (float) – L2 regularization strength (default=None)

static add_model_specific_args(parent_parser)[source]¶

configure_optimizers()[source]¶

forward(x)[source]¶

test_epoch_end(outputs)[source]¶

test_step(batch, batch_idx)[source]¶

training_step(batch, batch_idx)[source]¶

validation_epoch_end(outputs)[source]¶

validation_step(batch, batch_idx)[source]¶

pl_bolts.models.regression.logistic_regression.cli_main()[source]¶

pl_bolts.models.rl package¶

Subpackages¶

pl_bolts.models.rl.common package¶

Submodules¶

pl_bolts.models.rl.common.agents module¶

Agent module containing classes for Agent logic Based on the implementations found here: https://github.com/Shmuma/ptan/blob/master/ptan/agent.py

class pl_bolts.models.rl.common.agents.Agent(net)[source]¶

Bases: abc.ABC

Basic agent that always returns 0

__call__(state, device, *args, **kwargs)[source]¶

Using the given network, decide what action to carry :type _sphinx_paramlinks_pl_bolts.models.rl.common.agents.Agent.__call__.state: Tensor :param _sphinx_paramlinks_pl_bolts.models.rl.common.agents.Agent.__call__.state: current state of the environment :type _sphinx_paramlinks_pl_bolts.models.rl.common.agents.Agent.__call__.device: str :param _sphinx_paramlinks_pl_bolts.models.rl.common.agents.Agent.__call__.device: device used for current batch

Return type: List[int]
Returns: action

class pl_bolts.models.rl.common.agents.PolicyAgent(net)[source]¶

Bases: pl_bolts.models.rl.common.agents.Agent

Policy based agent that returns an action based on the networks policy

__call__(states, device)[source]¶

Takes in the current state and returns the action based on the agents policy :type _sphinx_paramlinks_pl_bolts.models.rl.common.agents.PolicyAgent.__call__.states: Tensor :param _sphinx_paramlinks_pl_bolts.models.rl.common.agents.PolicyAgent.__call__.states: current state of the environment :type _sphinx_paramlinks_pl_bolts.models.rl.common.agents.PolicyAgent.__call__.device: str :param _sphinx_paramlinks_pl_bolts.models.rl.common.agents.PolicyAgent.__call__.device: the device used for the current batch

Return type: List[int]
Returns: action defined by policy

class pl_bolts.models.rl.common.agents.ValueAgent(net, action_space, eps_start=1.0, eps_end=0.2, eps_frames=1000)[source]¶

Bases: pl_bolts.models.rl.common.agents.Agent

Value based agent that returns an action based on the Q values from the network

__call__(state, device)[source]¶

Takes in the current state and returns the action based on the agents policy :type _sphinx_paramlinks_pl_bolts.models.rl.common.agents.ValueAgent.__call__.state: Tensor :param _sphinx_paramlinks_pl_bolts.models.rl.common.agents.ValueAgent.__call__.state: current state of the environment :type _sphinx_paramlinks_pl_bolts.models.rl.common.agents.ValueAgent.__call__.device: str :param _sphinx_paramlinks_pl_bolts.models.rl.common.agents.ValueAgent.__call__.device: the device used for the current batch

Return type: List[int]
Returns: action defined by policy

get_action(state, device)[source]¶

Returns the best action based on the Q values of the network :type _sphinx_paramlinks_pl_bolts.models.rl.common.agents.ValueAgent.get_action.state: Tensor :param _sphinx_paramlinks_pl_bolts.models.rl.common.agents.ValueAgent.get_action.state: current state of the environment :type _sphinx_paramlinks_pl_bolts.models.rl.common.agents.ValueAgent.get_action.device: device :param _sphinx_paramlinks_pl_bolts.models.rl.common.agents.ValueAgent.get_action.device: the device used for the current batch

Returns: action defined by Q values

get_random_action(state)[source]¶

returns a random action

Return type: int

update_epsilon(step)[source]¶

Updates the epsilon value based on the current step :type _sphinx_paramlinks_pl_bolts.models.rl.common.agents.ValueAgent.update_epsilon.step: int :param _sphinx_paramlinks_pl_bolts.models.rl.common.agents.ValueAgent.update_epsilon.step: current global step

Return type: None

pl_bolts.models.rl.common.cli module¶

Contains generic arguments used for all models

pl_bolts.models.rl.common.cli.add_base_args(parent)[source]¶

Adds arguments for DQN model

Note: these params are fine tuned for Pong env

Parameters: parent –
Return type: ArgumentParser

pl_bolts.models.rl.common.gym_wrappers module¶

Set of wrapper functions for gym environments taken from https://github.com/Shmuma/ptan/blob/master/ptan/common/wrappers.py

class pl_bolts.models.rl.common.gym_wrappers.BufferWrapper(env, n_steps, dtype=numpy.float32)[source]¶

Bases: gym.ObservationWrapper

“Wrapper for image stacking

observation(observation)[source]¶: convert observation

reset()[source]¶: reset env

class pl_bolts.models.rl.common.gym_wrappers.DataAugmentation(env=None)[source]¶

Bases: gym.ObservationWrapper

Carries out basic data augmentation on the env observations - ToTensor - GrayScale - RandomCrop

observation(obs)[source]¶: preprocess the obs

class pl_bolts.models.rl.common.gym_wrappers.FireResetEnv(env=None)[source]¶

Bases: gym.Wrapper

For environments where the user need to press FIRE for the game to start.

reset()[source]¶: reset the env

step(action)[source]¶: Take 1 step

class pl_bolts.models.rl.common.gym_wrappers.ImageToPyTorch(env)[source]¶

Bases: gym.ObservationWrapper

converts image to pytorch format

static observation(observation)[source]¶: convert observation

class pl_bolts.models.rl.common.gym_wrappers.MaxAndSkipEnv(env=None, skip=4)[source]¶

Bases: gym.Wrapper

Return only every skip-th frame

reset()[source]¶: Clear past frame buffer and init. to first obs. from inner env.

step(action)[source]¶: take 1 step

class pl_bolts.models.rl.common.gym_wrappers.ProcessFrame84(env=None)[source]¶

Bases: gym.ObservationWrapper

preprocessing images from env

observation(obs)[source]¶: preprocess the obs

static process(frame)[source]¶: image preprocessing, formats to 84x84

class pl_bolts.models.rl.common.gym_wrappers.ScaledFloatFrame(*args, **kwargs)[source]¶

Bases: gym.ObservationWrapper

scales the pixels

static observation(obs)[source]¶

class pl_bolts.models.rl.common.gym_wrappers.ToTensor(env=None)[source]¶

Bases: gym.Wrapper

For environments where the user need to press FIRE for the game to start.

reset()[source]¶: reset the env and cast to tensor

step(action)[source]¶: Take 1 step and cast to tensor

pl_bolts.models.rl.common.gym_wrappers.make_environment(env_name)[source]¶: Convert environment with wrappers

pl_bolts.models.rl.common.memory module¶

Series of memory buffers sued

class pl_bolts.models.rl.common.memory.Buffer(capacity)[source]¶

Bases: object

Basic Buffer for storing a single experience at a time :type _sphinx_paramlinks_pl_bolts.models.rl.common.memory.Buffer.capacity: int :param _sphinx_paramlinks_pl_bolts.models.rl.common.memory.Buffer.capacity: size of the buffer

append(experience)[source]¶

Add experience to the buffer :type _sphinx_paramlinks_pl_bolts.models.rl.common.memory.Buffer.append.experience: Experience :param _sphinx_paramlinks_pl_bolts.models.rl.common.memory.Buffer.append.experience: tuple (state, action, reward, done, new_state)

Return type: None

sample(*args)[source]¶: returns everything in the buffer so far it is then reset :rtype: Union[Tuple, List[Tuple]] :returns: a batch of tuple np arrays of state, action, reward, done, next_state

class pl_bolts.models.rl.common.memory.Experience(state, action, reward, done, new_state)[source]¶

Bases: tuple

Create new instance of Experience(state, action, reward, done, new_state)

_asdict()[source]¶: Return a new OrderedDict which maps field names to their values.

classmethod _make(iterable)[source]¶: Make a new Experience object from a sequence or iterable

_replace(**kwds)[source]¶: Return a new Experience object replacing specified fields with new values

_field_defaults = {}[source]¶

_fields = ('state', 'action', 'reward', 'done', 'new_state')[source]¶

_fields_defaults = {}[source]¶

property action[source]¶: Alias for field number 1

property done[source]¶: Alias for field number 3

property new_state[source]¶: Alias for field number 4

property reward[source]¶: Alias for field number 2

property state[source]¶: Alias for field number 0

class pl_bolts.models.rl.common.memory.MeanBuffer(capacity)[source]¶

Bases: object

Stores a deque of items and calculates the mean

add(val)[source]¶

Add to the buffer

Return type: None

mean()[source]¶

Retrieve the mean

Return type: float

class pl_bolts.models.rl.common.memory.MultiStepBuffer(capacity, n_steps=1, gamma=0.99)[source]¶

Bases: pl_bolts.models.rl.common.memory.ReplayBuffer

N Step Replay Buffer

Parameters

capacity (int) – max number of experiences that will be stored in the buffer
n_steps (int) – number of steps used for calculating discounted reward/experience
gamma (float) – discount factor when calculating n_step discounted reward of the experience being stored in buffer

append(exp)[source]¶

Add experience to the buffer :type _sphinx_paramlinks_pl_bolts.models.rl.common.memory.MultiStepBuffer.append.exp: Experience :param _sphinx_paramlinks_pl_bolts.models.rl.common.memory.MultiStepBuffer.append.exp: tuple (state, action, reward, done, new_state)

Return type: None

discount_rewards(experiences)[source]¶

Calculates the discounted reward over N experiences :type _sphinx_paramlinks_pl_bolts.models.rl.common.memory.MultiStepBuffer.discount_rewards.experiences: Tuple[Experience] :param _sphinx_paramlinks_pl_bolts.models.rl.common.memory.MultiStepBuffer.discount_rewards.experiences: Tuple of Experience

Return type: float
Returns: total discounted reward

split_head_tail_exp(experiences)[source]¶

Takes in a tuple of experiences and returns the last state and tail experiences based on if the last state is the end of an episode :type _sphinx_paramlinks_pl_bolts.models.rl.common.memory.MultiStepBuffer.split_head_tail_exp.experiences: Tuple[Experience] :param _sphinx_paramlinks_pl_bolts.models.rl.common.memory.MultiStepBuffer.split_head_tail_exp.experiences: Tuple of N Experience

Return type: Tuple[List, Tuple[Experience]]
Returns: last state (Array or None) and remaining Experience

update_history_queue(exp)[source]¶

Updates the experience history queue with the lastest experiences. In the event of an experience step is in the done state, the history will be incrementally appended to the queue, removing the tail of the history each time. :param _sphinx_paramlinks_pl_bolts.models.rl.common.memory.MultiStepBuffer.update_history_queue.env_idx: index of the environment :param _sphinx_paramlinks_pl_bolts.models.rl.common.memory.MultiStepBuffer.update_history_queue.exp: the current experience :param _sphinx_paramlinks_pl_bolts.models.rl.common.memory.MultiStepBuffer.update_history_queue.history: history of experience steps for this environment

Return type: None

class pl_bolts.models.rl.common.memory.PERBuffer(buffer_size, prob_alpha=0.6, beta_start=0.4, beta_frames=100000)[source]¶

Bases: pl_bolts.models.rl.common.memory.ReplayBuffer

simple list based Prioritized Experience Replay Buffer Based on implementation found here: https://github.com/Shmuma/ptan/blob/master/ptan/experience.py#L371

append(exp)[source]¶

Adds experiences from exp_source to the PER buffer :param _sphinx_paramlinks_pl_bolts.models.rl.common.memory.PERBuffer.append.exp: experience tuple being added to the buffer

Return type: None

sample(batch_size=32)[source]¶

Takes a prioritized sample from the buffer :param _sphinx_paramlinks_pl_bolts.models.rl.common.memory.PERBuffer.sample.batch_size: size of sample

Return type: Tuple
Returns: sample of experiences chosen with ranked probability

update_beta(step)[source]¶

Update the beta value which accounts for the bias in the PER :param _sphinx_paramlinks_pl_bolts.models.rl.common.memory.PERBuffer.update_beta.step: current global step

Return type: float
Returns: beta value for this indexed experience

update_priorities(batch_indices, batch_priorities)[source]¶

Update the priorities from the last batch, this should be called after the loss for this batch has been calculated. :type _sphinx_paramlinks_pl_bolts.models.rl.common.memory.PERBuffer.update_priorities.batch_indices: List :param _sphinx_paramlinks_pl_bolts.models.rl.common.memory.PERBuffer.update_priorities.batch_indices: index of each datum in the batch :type _sphinx_paramlinks_pl_bolts.models.rl.common.memory.PERBuffer.update_priorities.batch_priorities: List :param _sphinx_paramlinks_pl_bolts.models.rl.common.memory.PERBuffer.update_priorities.batch_priorities: priority of each datum in the batch

Return type: None

class pl_bolts.models.rl.common.memory.ReplayBuffer(capacity)[source]¶

Bases: pl_bolts.models.rl.common.memory.Buffer

Replay Buffer for storing past experiences allowing the agent to learn from them

sample(batch_size)[source]¶

Takes a sample of the buffer :type _sphinx_paramlinks_pl_bolts.models.rl.common.memory.ReplayBuffer.sample.batch_size: int :param _sphinx_paramlinks_pl_bolts.models.rl.common.memory.ReplayBuffer.sample.batch_size: current batch_size

Return type: Tuple
Returns: a batch of tuple np arrays of state, action, reward, done, next_state

pl_bolts.models.rl.common.networks module¶

Series of networks used Based on implementations found here:

class pl_bolts.models.rl.common.networks.CNN(input_shape, n_actions)[source]¶

Bases: torch.nn.Module

Simple MLP network :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.CNN.input_shape: observation shape of the environment :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.CNN.n_actions: number of discrete actions available in the environment

_get_conv_out(shape)[source]¶

Calculates the output size of the last conv layer :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.CNN._get_conv_out.shape: input dimensions

Return type: int
Returns: size of the conv output

forward(input_x)[source]¶

Forward pass through network :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.CNN.forward.x: input to network

Return type: Tensor
Returns: output of network

class pl_bolts.models.rl.common.networks.DuelingCNN(input_shape, n_actions, _=128)[source]¶

Bases: torch.nn.Module

CNN network with duel heads for val and advantage :type _sphinx_paramlinks_pl_bolts.models.rl.common.networks.DuelingCNN.input_shape: Tuple :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.DuelingCNN.input_shape: observation shape of the environment :type _sphinx_paramlinks_pl_bolts.models.rl.common.networks.DuelingCNN.n_actions: int :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.DuelingCNN.n_actions: number of discrete actions available in the environment :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.DuelingCNN.hidden_size: size of hidden layers

_get_conv_out(shape)[source]¶

Calculates the output size of the last conv layer :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.DuelingCNN._get_conv_out.shape: input dimensions

Return type: int
Returns: size of the conv output

adv_val(input_x)[source]¶

Gets the advantage and value by passing out of the base network through the value and advantage heads :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.DuelingCNN.adv_val.input_x: input to network

Returns: advantage, value

forward(input_x)[source]¶

Forward pass through network. Calculates the Q using the value and advantage :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.DuelingCNN.forward.input_x: input to network

Returns: Q value

class pl_bolts.models.rl.common.networks.DuelingMLP(input_shape, n_actions, hidden_size=128)[source]¶

Bases: torch.nn.Module

MLP network with duel heads for val and advantage :type _sphinx_paramlinks_pl_bolts.models.rl.common.networks.DuelingMLP.input_shape: Tuple :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.DuelingMLP.input_shape: observation shape of the environment :type _sphinx_paramlinks_pl_bolts.models.rl.common.networks.DuelingMLP.n_actions: int :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.DuelingMLP.n_actions: number of discrete actions available in the environment :type _sphinx_paramlinks_pl_bolts.models.rl.common.networks.DuelingMLP.hidden_size: int :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.DuelingMLP.hidden_size: size of hidden layers

adv_val(input_x)[source]¶

Gets the advantage and value by passing out of the base network through the value and advantage heads :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.DuelingMLP.adv_val.input_x: input to network

Return type: Tuple[Tensor, Tensor]
Returns: advantage, value

forward(input_x)[source]¶

Forward pass through network. Calculates the Q using the value and advantage :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.DuelingMLP.forward.x: input to network

Returns: Q value

class pl_bolts.models.rl.common.networks.MLP(input_shape, n_actions, hidden_size=128)[source]¶

Bases: torch.nn.Module

Simple MLP network :type _sphinx_paramlinks_pl_bolts.models.rl.common.networks.MLP.input_shape: Tuple :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.MLP.input_shape: observation shape of the environment :type _sphinx_paramlinks_pl_bolts.models.rl.common.networks.MLP.n_actions: int :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.MLP.n_actions: number of discrete actions available in the environment :type _sphinx_paramlinks_pl_bolts.models.rl.common.networks.MLP.hidden_size: int :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.MLP.hidden_size: size of hidden layers

forward(input_x)[source]¶

Forward pass through network :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.MLP.forward.x: input to network

Returns: output of network

class pl_bolts.models.rl.common.networks.NoisyCNN(input_shape, n_actions)[source]¶

Bases: torch.nn.Module

CNN with Noisy Linear layers for exploration :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.NoisyCNN.input_shape: observation shape of the environment :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.NoisyCNN.n_actions: number of discrete actions available in the environment

_get_conv_out(shape)[source]¶

Calculates the output size of the last conv layer :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.NoisyCNN._get_conv_out.shape: input dimensions

Return type: int
Returns: size of the conv output

forward(input_x)[source]¶

Forward pass through network :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.NoisyCNN.forward.x: input to network

Return type: Tensor
Returns: output of network

class pl_bolts.models.rl.common.networks.NoisyLinear(in_features, out_features, sigma_init=0.017, bias=True)[source]¶

Bases: torch.nn.Linear

Noisy Layer using Independent Gaussian Noise. based on https://github.com/PacktPublishing/Deep-Reinforcement-Learning-Hands-On-Second-Edition/blob/master/ Chapter08/lib/dqn_extra.py#L19 :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.NoisyLinear.in_features: number of inputs :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.NoisyLinear.out_features: number of outputs :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.NoisyLinear.sigma_init: initial fill value of noisy weights :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.NoisyLinear.bias: flag to include bias to linear layer

forward(input_x)[source]¶

Forward pass of the layer :type _sphinx_paramlinks_pl_bolts.models.rl.common.networks.NoisyLinear.forward.input_x: Tensor :param _sphinx_paramlinks_pl_bolts.models.rl.common.networks.NoisyLinear.forward.input_x: input tensor

Return type: Tensor
Returns: output of the layer

reset_parameters()[source]¶

initializes or resets the paramseter of the layer

Return type: None

Submodules¶

pl_bolts.models.rl.double_dqn_model module¶

Double DQN

class pl_bolts.models.rl.double_dqn_model.DoubleDQN(env, eps_start=1.0, eps_end=0.02, eps_last_frame=150000, sync_rate=1000, gamma=0.99, learning_rate=0.0001, batch_size=32, replay_size=100000, warm_start_size=10000, avg_reward_len=100, min_episode_reward=-21, seed=123, batches_per_epoch=1000, n_steps=1, **kwargs)[source]¶

Bases: pl_bolts.models.rl.dqn_model.DQN

Double Deep Q-network (DDQN) PyTorch Lightning implementation of Double DQN

Paper authors: Hado van Hasselt, Arthur Guez, David Silver

Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.double_dqn_model import DoubleDQN
...
>>> model = DoubleDQN("PongNoFrameskip-v4")

Train:

trainer = Trainer()
trainer.fit(model)

Parameters

env (str) – gym environment tag
gpus – number of gpus being used
eps_start (float) – starting value of epsilon for the epsilon-greedy exploration
eps_end (float) – final value of epsilon for the epsilon-greedy exploration
eps_last_frame (int) – the final frame in for the decrease of epsilon. At this frame espilon = eps_end
sync_rate (int) – the number of iterations between syncing up the target network with the train network
gamma (float) – discount factor
lr – learning rate
batch_size (int) – size of minibatch pulled from the DataLoader
replay_size (int) – total capacity of the replay buffer
warm_start_size (int) – how many random steps through the environment to be carried out at the start of training to fill the buffer with a starting point
sample_len – the number of samples to pull from the dataset iterator and feed to the DataLoader

Note

This example is based on https://github.com/PacktPublishing/Deep-Reinforcement-Learning-Hands-On-Second-Edition/blob/master/Chapter08/03_dqn_double.py

Note

Currently only supports CPU and single GPU training with distributed_backend=dp

PyTorch Lightning implementation of DQN Paper authors: Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, Martin Riedmiller. Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.dqn_model import DQN
...
>>> model = DQN("PongNoFrameskip-v4")

Train:

trainer = Trainer()
trainer.fit(model)

Parameters

env (str) – gym environment tag
eps_start (float) – starting value of epsilon for the epsilon-greedy exploration
eps_end (float) – final value of epsilon for the epsilon-greedy exploration
eps_last_frame (int) – the final frame in for the decrease of epsilon. At this frame espilon = eps_end
sync_rate (int) – the number of iterations between syncing up the target network with the train network
gamma (float) – discount factor
learning_rate (float) – learning rate
batch_size (int) – size of minibatch pulled from the DataLoader
replay_size (int) – total capacity of the replay buffer
warm_start_size (int) – how many random steps through the environment to be carried out at the start of training to fill the buffer with a starting point
avg_reward_len (int) – how many episodes to take into account when calculating the avg reward
min_episode_reward (int) – the minimum score that can be achieved in an episode. Used for filling the avg buffer before training begins
seed (int) – seed value for all RNG used
batches_per_epoch (int) – number of batches per epoch
n_steps (int) – size of n step look ahead

Note

This example is based on: https://github.com/PacktPublishing/Deep-Reinforcement-Learning-Hands-On-Second-Edition/blob/master/Chapter06/02_dqn_pong.py

Note

Currently only supports CPU and single GPU training with distributed_backend=dp

training_step(batch, _)[source]¶

Carries out a single step through the environment to update the replay buffer. Then calculates loss based on the minibatch recieved :type _sphinx_paramlinks_pl_bolts.models.rl.double_dqn_model.DoubleDQN.training_step.batch: Tuple[Tensor, Tensor] :param _sphinx_paramlinks_pl_bolts.models.rl.double_dqn_model.DoubleDQN.training_step.batch: current mini batch of replay data :param _sphinx_paramlinks_pl_bolts.models.rl.double_dqn_model.DoubleDQN.training_step._: batch number, not used

Return type: OrderedDict
Returns: Training loss and log metrics

pl_bolts.models.rl.double_dqn_model.cli_main()[source]¶

pl_bolts.models.rl.dqn_model module¶

Deep Q Network

class pl_bolts.models.rl.dqn_model.DQN(env, eps_start=1.0, eps_end=0.02, eps_last_frame=150000, sync_rate=1000, gamma=0.99, learning_rate=0.0001, batch_size=32, replay_size=100000, warm_start_size=10000, avg_reward_len=100, min_episode_reward=-21, seed=123, batches_per_epoch=1000, n_steps=1, **kwargs)[source]¶

Bases: pytorch_lightning.LightningModule

Basic DQN Model

PyTorch Lightning implementation of DQN Paper authors: Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, Martin Riedmiller. Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.dqn_model import DQN
...
>>> model = DQN("PongNoFrameskip-v4")

Train:

trainer = Trainer()
trainer.fit(model)

Parameters

env (str) – gym environment tag
eps_start (float) – starting value of epsilon for the epsilon-greedy exploration
eps_end (float) – final value of epsilon for the epsilon-greedy exploration
eps_last_frame (int) – the final frame in for the decrease of epsilon. At this frame espilon = eps_end
sync_rate (int) – the number of iterations between syncing up the target network with the train network
gamma (float) – discount factor
learning_rate (float) – learning rate
batch_size (int) – size of minibatch pulled from the DataLoader
replay_size (int) – total capacity of the replay buffer
warm_start_size (int) – how many random steps through the environment to be carried out at the start of training to fill the buffer with a starting point
avg_reward_len (int) – how many episodes to take into account when calculating the avg reward
min_episode_reward (int) – the minimum score that can be achieved in an episode. Used for filling the avg buffer before training begins
seed (int) – seed value for all RNG used
batches_per_epoch (int) – number of batches per epoch
n_steps (int) – size of n step look ahead

Note

This example is based on: https://github.com/PacktPublishing/Deep-Reinforcement-Learning-Hands-On-Second-Edition/blob/master/Chapter06/02_dqn_pong.py

Note

Currently only supports CPU and single GPU training with distributed_backend=dp

_dataloader()[source]¶

Initialize the Replay Buffer dataset used for retrieving experiences

Return type: DataLoader

static add_model_specific_args(arg_parser)[source]¶

Adds arguments for DQN model Note: these params are fine tuned for Pong env :type _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.add_model_specific_args.arg_parser: ArgumentParser :param _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.add_model_specific_args.arg_parser: parent parser

Return type: ArgumentParser

build_networks()[source]¶

Initializes the DQN train and target networks

Return type: None

configure_optimizers()[source]¶

Initialize Adam optimizer

Return type: List[Optimizer]

forward(x)[source]¶

Passes in a state x through the network and gets the q_values of each action as an output :type _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.forward.x: Tensor :param _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.forward.x: environment state

Return type: Tensor
Returns: q values

static make_environment(env_name, seed=None)[source]¶

Initialise gym environment :type _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.make_environment.env_name: str :param _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.make_environment.env_name: environment name or tag :type _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.make_environment.seed: Optional[int] :param _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.make_environment.seed: value to seed the environment RNG for reproducibility

Return type: Env
Returns: gym environment

populate(warm_start)[source]¶

Populates the buffer with initial experience

Return type: None

run_n_episodes(env, n_epsiodes=1, epsilon=1.0)[source]¶

Carries out N episodes of the environment with the current agent :param _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.run_n_episodes.env: environment to use, either train environment or test environment :type _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.run_n_episodes.n_epsiodes: int :param _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.run_n_episodes.n_epsiodes: number of episodes to run :type _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.run_n_episodes.epsilon: float :param _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.run_n_episodes.epsilon: epsilon value for DQN agent

Return type: List[int]

test_dataloader()[source]¶

Get test loader

Return type: DataLoader

test_epoch_end(outputs)[source]¶

Log the avg of the test results

Return type: Dict[str, Tensor]

test_step(*args, **kwargs)[source]¶

Evaluate the agent for 10 episodes

Return type: Dict[str, Tensor]

train_batch()[source]¶: Contains the logic for generating a new batch of data to be passed to the DataLoader :rtype: Tuple[Tensor, Tensor, Tensor, Tensor, Tensor] :returns: yields a Experience tuple containing the state, action, reward, done and next_state.

train_dataloader()[source]¶

Get train loader

Return type: DataLoader

training_step(batch, _)[source]¶

Carries out a single step through the environment to update the replay buffer. Then calculates loss based on the minibatch recieved :type _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.training_step.batch: Tuple[Tensor, Tensor] :param _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.training_step.batch: current mini batch of replay data :param _sphinx_paramlinks_pl_bolts.models.rl.dqn_model.DQN.training_step._: batch number, not used

Return type: OrderedDict
Returns: Training loss and log metrics

pl_bolts.models.rl.dqn_model.cli_main()[source]¶

pl_bolts.models.rl.dueling_dqn_model module¶

Dueling DQN

class pl_bolts.models.rl.dueling_dqn_model.DuelingDQN(env, eps_start=1.0, eps_end=0.02, eps_last_frame=150000, sync_rate=1000, gamma=0.99, learning_rate=0.0001, batch_size=32, replay_size=100000, warm_start_size=10000, avg_reward_len=100, min_episode_reward=-21, seed=123, batches_per_epoch=1000, n_steps=1, **kwargs)[source]¶

Bases: pl_bolts.models.rl.dqn_model.DQN

PyTorch Lightning implementation of Dueling DQN

Paper authors: Ziyu Wang, Tom Schaul, Matteo Hessel, Hado van Hasselt, Marc Lanctot, Nando de Freitas

Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.dueling_dqn_model import DuelingDQN
...
>>> model = DuelingDQN("PongNoFrameskip-v4")

Train:

trainer = Trainer()
trainer.fit(model)

Parameters

env (str) – gym environment tag
gpus – number of gpus being used
eps_start (float) – starting value of epsilon for the epsilon-greedy exploration
eps_end (float) – final value of epsilon for the epsilon-greedy exploration
eps_last_frame (int) – the final frame in for the decrease of epsilon. At this frame espilon = eps_end
sync_rate (int) – the number of iterations between syncing up the target network with the train network
gamma (float) – discount factor
lr – learning rate
batch_size (int) – size of minibatch pulled from the DataLoader
replay_size (int) – total capacity of the replay buffer
warm_start_size (int) – how many random steps through the environment to be carried out at the start of training to fill the buffer with a starting point
sample_len – the number of samples to pull from the dataset iterator and feed to the DataLoader

Note

Currently only supports CPU and single GPU training with distributed_backend=dp

PyTorch Lightning implementation of DQN Paper authors: Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, Martin Riedmiller. Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.dqn_model import DQN
...
>>> model = DQN("PongNoFrameskip-v4")

Train:

trainer = Trainer()
trainer.fit(model)

Parameters

env (str) – gym environment tag
eps_start (float) – starting value of epsilon for the epsilon-greedy exploration
eps_end (float) – final value of epsilon for the epsilon-greedy exploration
eps_last_frame (int) – the final frame in for the decrease of epsilon. At this frame espilon = eps_end
sync_rate (int) – the number of iterations between syncing up the target network with the train network
gamma (float) – discount factor
learning_rate (float) – learning rate
batch_size (int) – size of minibatch pulled from the DataLoader
replay_size (int) – total capacity of the replay buffer
warm_start_size (int) – how many random steps through the environment to be carried out at the start of training to fill the buffer with a starting point
avg_reward_len (int) – how many episodes to take into account when calculating the avg reward
min_episode_reward (int) – the minimum score that can be achieved in an episode. Used for filling the avg buffer before training begins
seed (int) – seed value for all RNG used
batches_per_epoch (int) – number of batches per epoch
n_steps (int) – size of n step look ahead

Note

This example is based on: https://github.com/PacktPublishing/Deep-Reinforcement-Learning-Hands-On-Second-Edition/blob/master/Chapter06/02_dqn_pong.py

Note

Currently only supports CPU and single GPU training with distributed_backend=dp

build_networks()[source]¶

Initializes the Dueling DQN train and target networks

Return type: None

pl_bolts.models.rl.dueling_dqn_model.cli_main()[source]¶

pl_bolts.models.rl.noisy_dqn_model module¶

Noisy DQN

class pl_bolts.models.rl.noisy_dqn_model.NoisyDQN(env, eps_start=1.0, eps_end=0.02, eps_last_frame=150000, sync_rate=1000, gamma=0.99, learning_rate=0.0001, batch_size=32, replay_size=100000, warm_start_size=10000, avg_reward_len=100, min_episode_reward=-21, seed=123, batches_per_epoch=1000, n_steps=1, **kwargs)[source]¶

Bases: pl_bolts.models.rl.dqn_model.DQN

PyTorch Lightning implementation of Noisy DQN

Paper authors: Meire Fortunato, Mohammad Gheshlaghi Azar, Bilal Piot, Jacob Menick, Ian Osband, Alex Graves, Vlad Mnih, Remi Munos, Demis Hassabis, Olivier Pietquin, Charles Blundell, Shane Legg

Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.noisy_dqn_model import NoisyDQN
...
>>> model = NoisyDQN("PongNoFrameskip-v4")

Train:

trainer = Trainer()
trainer.fit(model)

Parameters

env (str) – gym environment tag
gpus – number of gpus being used
eps_start (float) – starting value of epsilon for the epsilon-greedy exploration
eps_end (float) – final value of epsilon for the epsilon-greedy exploration
eps_last_frame (int) – the final frame in for the decrease of epsilon. At this frame espilon = eps_end
sync_rate (int) – the number of iterations between syncing up the target network with the train network
gamma (float) – discount factor
lr – learning rate
batch_size (int) – size of minibatch pulled from the DataLoader
replay_size (int) – total capacity of the replay buffer
warm_start_size (int) – how many random steps through the environment to be carried out at the start of
to fill the buffer with a starting point (training) –
sample_len – the number of samples to pull from the dataset iterator and feed to the DataLoader

Note

Currently only supports CPU and single GPU training with distributed_backend=dp

PyTorch Lightning implementation of DQN Paper authors: Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, Martin Riedmiller. Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.dqn_model import DQN
...
>>> model = DQN("PongNoFrameskip-v4")

Train:

trainer = Trainer()
trainer.fit(model)

Parameters

env (str) – gym environment tag
eps_start (float) – starting value of epsilon for the epsilon-greedy exploration
eps_end (float) – final value of epsilon for the epsilon-greedy exploration
eps_last_frame (int) – the final frame in for the decrease of epsilon. At this frame espilon = eps_end
sync_rate (int) – the number of iterations between syncing up the target network with the train network
gamma (float) – discount factor
learning_rate (float) – learning rate
batch_size (int) – size of minibatch pulled from the DataLoader
replay_size (int) – total capacity of the replay buffer
warm_start_size (int) – how many random steps through the environment to be carried out at the start of training to fill the buffer with a starting point
avg_reward_len (int) – how many episodes to take into account when calculating the avg reward
min_episode_reward (int) – the minimum score that can be achieved in an episode. Used for filling the avg buffer before training begins
seed (int) – seed value for all RNG used
batches_per_epoch (int) – number of batches per epoch
n_steps (int) – size of n step look ahead

Note

This example is based on: https://github.com/PacktPublishing/Deep-Reinforcement-Learning-Hands-On-Second-Edition/blob/master/Chapter06/02_dqn_pong.py

Note

Currently only supports CPU and single GPU training with distributed_backend=dp

build_networks()[source]¶

Initializes the Noisy DQN train and target networks

Return type: None

on_train_start()[source]¶

Set the agents epsilon to 0 as the exploration comes from the network

Return type: None

train_batch()[source]¶: Contains the logic for generating a new batch of data to be passed to the DataLoader. This is the same function as the standard DQN except that we dont update epsilon as it is always 0. The exploration comes from the noisy network. :rtype: Tuple[Tensor, Tensor, Tensor, Tensor, Tensor] :returns: yields a Experience tuple containing the state, action, reward, done and next_state.

pl_bolts.models.rl.noisy_dqn_model.cli_main()[source]¶

pl_bolts.models.rl.per_dqn_model module¶

Prioritized Experience Replay DQN

class pl_bolts.models.rl.per_dqn_model.PERDQN(env, eps_start=1.0, eps_end=0.02, eps_last_frame=150000, sync_rate=1000, gamma=0.99, learning_rate=0.0001, batch_size=32, replay_size=100000, warm_start_size=10000, avg_reward_len=100, min_episode_reward=-21, seed=123, batches_per_epoch=1000, n_steps=1, **kwargs)[source]¶

Bases: pl_bolts.models.rl.dqn_model.DQN

PyTorch Lightning implementation of DQN With Prioritized Experience Replay

Paper authors: Tom Schaul, John Quan, Ioannis Antonoglou, David Silver

Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.per_dqn_model import PERDQN
...
>>> model = PERDQN("PongNoFrameskip-v4")

Train:

trainer = Trainer()
trainer.fit(model)

Args:
    env: gym environment tag
    gpus: number of gpus being used
    eps_start: starting value of epsilon for the epsilon-greedy exploration
    eps_end: final value of epsilon for the epsilon-greedy exploration
    eps_last_frame: the final frame in for the decrease of epsilon. At this frame espilon = eps_end
    sync_rate: the number of iterations between syncing up the target network with the train network
    gamma: discount factor
    learning_rate: learning rate
    batch_size: size of minibatch pulled from the DataLoader
    replay_size: total capacity of the replay buffer
    warm_start_size: how many random steps through the environment to be carried out at the start of
        training to fill the buffer with a starting point
    num_samples: the number of samples to pull from the dataset iterator and feed to the DataLoader

.. note::
    This example is based on:
     https://github.com/PacktPublishing/Deep-Reinforcement-Learning-Hands-On-Second-Edition/blob/master/Chapter08/05_dqn_prio_replay.py

.. note:: Currently only supports CPU and single GPU training with `distributed_backend=dp`

PyTorch Lightning implementation of DQN Paper authors: Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, Martin Riedmiller. Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.dqn_model import DQN
...
>>> model = DQN("PongNoFrameskip-v4")

Train:

trainer = Trainer()
trainer.fit(model)

Parameters

env (str) – gym environment tag
eps_start (float) – starting value of epsilon for the epsilon-greedy exploration
eps_end (float) – final value of epsilon for the epsilon-greedy exploration
eps_last_frame (int) – the final frame in for the decrease of epsilon. At this frame espilon = eps_end
sync_rate (int) – the number of iterations between syncing up the target network with the train network
gamma (float) – discount factor
learning_rate (float) – learning rate
batch_size (int) – size of minibatch pulled from the DataLoader
replay_size (int) – total capacity of the replay buffer
warm_start_size (int) – how many random steps through the environment to be carried out at the start of training to fill the buffer with a starting point
avg_reward_len (int) – how many episodes to take into account when calculating the avg reward
min_episode_reward (int) – the minimum score that can be achieved in an episode. Used for filling the avg buffer before training begins
seed (int) – seed value for all RNG used
batches_per_epoch (int) – number of batches per epoch
n_steps (int) – size of n step look ahead

Note

This example is based on: https://github.com/PacktPublishing/Deep-Reinforcement-Learning-Hands-On-Second-Edition/blob/master/Chapter06/02_dqn_pong.py

Note

Currently only supports CPU and single GPU training with distributed_backend=dp

_dataloader()[source]¶

Initialize the Replay Buffer dataset used for retrieving experiences

Return type: DataLoader

train_batch()[source]¶: Contains the logic for generating a new batch of data to be passed to the DataLoader :rtype: Tuple[Tensor, Tensor, Tensor, Tensor, Tensor] :returns: yields a Experience tuple containing the state, action, reward, done and next_state.

training_step(batch, _)[source]¶

Carries out a single step through the environment to update the replay buffer. Then calculates loss based on the minibatch recieved :param _sphinx_paramlinks_pl_bolts.models.rl.per_dqn_model.PERDQN.training_step.batch: current mini batch of replay data :param _sphinx_paramlinks_pl_bolts.models.rl.per_dqn_model.PERDQN.training_step._: batch number, not used

Return type: OrderedDict
Returns: Training loss and log metrics

pl_bolts.models.rl.per_dqn_model.cli_main()[source]¶

pl_bolts.models.rl.reinforce_model module¶

class pl_bolts.models.rl.reinforce_model.Reinforce(env, gamma=0.99, lr=0.01, batch_size=8, n_steps=10, avg_reward_len=100, entropy_beta=0.01, epoch_len=1000, num_batch_episodes=4, **kwargs)[source]¶

Bases: pytorch_lightning.LightningModule

PyTorch Lightning implementation of REINFORCE Paper authors: Richard S. Sutton, David McAllester, Satinder Singh, Yishay Mansour Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.reinforce_model import Reinforce
...
>>> model = Reinforce("CartPole-v0")

Train:

trainer = Trainer()
trainer.fit(model)

Parameters

env (str) – gym environment tag
gamma (float) – discount factor
lr (float) – learning rate
batch_size (int) – size of minibatch pulled from the DataLoader
n_steps (int) – number of stakes per discounted experience
entropy_beta (float) – entropy coefficient
epoch_len (int) – how many batches before pseudo epoch
num_batch_episodes (int) – how many episodes to rollout for each batch of training
avg_reward_len (int) – how many episodes to take into account when calculating the avg reward

Note

This example is based on: https://github.com/PacktPublishing/Deep-Reinforcement-Learning-Hands-On-Second-Edition/blob/master/Chapter11/02_cartpole_reinforce.py

Note

Currently only supports CPU and single GPU training with distributed_backend=dp

_dataloader()[source]¶

Initialize the Replay Buffer dataset used for retrieving experiences

Return type: DataLoader

static add_model_specific_args(arg_parser)[source]¶

Adds arguments for DQN model Note: these params are fine tuned for Pong env :param _sphinx_paramlinks_pl_bolts.models.rl.reinforce_model.Reinforce.add_model_specific_args.arg_parser: the current argument parser to add to

Return type: ArgumentParser
Returns: arg_parser with model specific cargs added

calc_qvals(rewards)[source]¶

Calculate the discounted rewards of all rewards in list :type _sphinx_paramlinks_pl_bolts.models.rl.reinforce_model.Reinforce.calc_qvals.rewards: List[float] :param _sphinx_paramlinks_pl_bolts.models.rl.reinforce_model.Reinforce.calc_qvals.rewards: list of rewards from latest batch

Return type: List[float]
Returns: list of discounted rewards

configure_optimizers()[source]¶

Initialize Adam optimizer

Return type: List[Optimizer]

discount_rewards(experiences)[source]¶

Calculates the discounted reward over N experiences :type _sphinx_paramlinks_pl_bolts.models.rl.reinforce_model.Reinforce.discount_rewards.experiences: Tuple[Experience] :param _sphinx_paramlinks_pl_bolts.models.rl.reinforce_model.Reinforce.discount_rewards.experiences: Tuple of Experience

Return type: float
Returns: total discounted reward

forward(x)[source]¶

Passes in a state x through the network and gets the q_values of each action as an output :type _sphinx_paramlinks_pl_bolts.models.rl.reinforce_model.Reinforce.forward.x: Tensor :param _sphinx_paramlinks_pl_bolts.models.rl.reinforce_model.Reinforce.forward.x: environment state

Return type: Tensor
Returns: q values

get_device(batch)[source]¶

Retrieve device currently being used by minibatch

Return type: str

loss(states, actions, scaled_rewards)[source]¶

Return type: Tensor

train_batch()[source]¶

Contains the logic for generating a new batch of data to be passed to the DataLoader :Yields: yields a tuple of Lists containing tensors for states, actions and rewards of the batch.

Return type: Tuple[List[Tensor], List[Tensor], List[Tensor]]

train_dataloader()[source]¶

Get train loader

Return type: DataLoader

training_step(batch, _)[source]¶

Carries out a single step through the environment to update the replay buffer. Then calculates loss based on the minibatch recieved :type _sphinx_paramlinks_pl_bolts.models.rl.reinforce_model.Reinforce.training_step.batch: Tuple[Tensor, Tensor] :param _sphinx_paramlinks_pl_bolts.models.rl.reinforce_model.Reinforce.training_step.batch: current mini batch of replay data :param _sphinx_paramlinks_pl_bolts.models.rl.reinforce_model.Reinforce.training_step._: batch number, not used

Return type: OrderedDict
Returns: Training loss and log metrics

pl_bolts.models.rl.reinforce_model.cli_main()[source]¶

pl_bolts.models.rl.vanilla_policy_gradient_model module¶

class pl_bolts.models.rl.vanilla_policy_gradient_model.VanillaPolicyGradient(env, gamma=0.99, lr=0.01, batch_size=8, n_steps=10, avg_reward_len=100, entropy_beta=0.01, epoch_len=1000, **kwargs)[source]¶

Bases: pytorch_lightning.LightningModule

PyTorch Lightning implementation of Vanilla Policy Gradient Paper authors: Richard S. Sutton, David McAllester, Satinder Singh, Yishay Mansour Model implemented by:

Donal Byrne <https://github.com/djbyrne>

Example

>>> from pl_bolts.models.rl.vanilla_policy_gradient_model import VanillaPolicyGradient
...
>>> model = VanillaPolicyGradient("CartPole-v0")

Train::: trainer = Trainer() trainer.fit(model)

Parameters

env (str) – gym environment tag
gamma (float) – discount factor
lr (float) – learning rate
batch_size (int) – size of minibatch pulled from the DataLoader
batch_episodes – how many episodes to rollout for each batch of training
entropy_beta (float) – dictates the level of entropy per batch
avg_reward_len (int) – how many episodes to take into account when calculating the avg reward

Note

This example is based on: https://github.com/PacktPublishing/Deep-Reinforcement-Learning-Hands-On-Second-Edition/blob/master/Chapter11/04_cartpole_pg.py

Note

Currently only supports CPU and single GPU training with distributed_backend=dp

_dataloader()[source]¶

Initialize the Replay Buffer dataset used for retrieving experiences

Return type: DataLoader

static add_model_specific_args(arg_parser)[source]¶

Adds arguments for DQN model Note: these params are fine tuned for Pong env :param _sphinx_paramlinks_pl_bolts.models.rl.vanilla_policy_gradient_model.VanillaPolicyGradient.add_model_specific_args.arg_parser: the current argument parser to add to

Return type: ArgumentParser
Returns: arg_parser with model specific cargs added

compute_returns(rewards)[source]¶

Calculate the discounted rewards of the batched rewards

Parameters: rewards – list of batched rewards
Returns: list of discounted rewards

configure_optimizers()[source]¶

Initialize Adam optimizer

Return type: List[Optimizer]

forward(x)[source]¶

Passes in a state x through the network and gets the q_values of each action as an output :type _sphinx_paramlinks_pl_bolts.models.rl.vanilla_policy_gradient_model.VanillaPolicyGradient.forward.x: Tensor :param _sphinx_paramlinks_pl_bolts.models.rl.vanilla_policy_gradient_model.VanillaPolicyGradient.forward.x: environment state

Return type: Tensor
Returns: q values

get_device(batch)[source]¶

Retrieve device currently being used by minibatch

Return type: str

loss(states, actions, scaled_rewards)[source]¶

Calculates the loss for VPG

Parameters

states – batched states
actions – batch actions
scaled_rewards – batche Q values

Return type

Tensor

Returns

loss for the current batch

train_batch()[source]¶: Contains the logic for generating a new batch of data to be passed to the DataLoader :rtype: Tuple[List[Tensor], List[Tensor], List[Tensor]] :returns: yields a tuple of Lists containing tensors for states, actions and rewards of the batch.

train_dataloader()[source]¶

Get train loader

Return type: DataLoader

training_step(batch, _)[source]¶

Carries out a single step through the environment to update the replay buffer. Then calculates loss based on the minibatch recieved :type _sphinx_paramlinks_pl_bolts.models.rl.vanilla_policy_gradient_model.VanillaPolicyGradient.training_step.batch: Tuple[Tensor, Tensor] :param _sphinx_paramlinks_pl_bolts.models.rl.vanilla_policy_gradient_model.VanillaPolicyGradient.training_step.batch: current mini batch of replay data :param _sphinx_paramlinks_pl_bolts.models.rl.vanilla_policy_gradient_model.VanillaPolicyGradient.training_step._: batch number, not used

Return type: OrderedDict
Returns: Training loss and log metrics

pl_bolts.models.rl.vanilla_policy_gradient_model.cli_main()[source]¶

pl_bolts.models.self_supervised package¶

These models have been pre-trained using self-supervised learning. The models can also be used without pre-training and overwritten for your own research.

Here’s an example for using these as pretrained models.

from pl_bolts.models.self_supervised import CPCV2

images = get_imagenet_batch()

# extract unsupervised representations
pretrained = CPCV2(pretrained=True)
representations = pretrained(images)

# use these in classification or any downstream task
classifications = classifier(representations)

Subpackages¶

pl_bolts.models.self_supervised.amdim package¶

Submodules¶

pl_bolts.models.self_supervised.amdim.amdim_module module¶

class pl_bolts.models.self_supervised.amdim.amdim_module.AMDIM(datamodule='cifar10', encoder='amdim_encoder', contrastive_task=torch.nn.Module, image_channels=3, image_height=32, encoder_feature_dim=320, embedding_fx_dim=1280, conv_block_depth=10, use_bn=False, tclip=20.0, learning_rate=0.0002, data_dir='', num_classes=10, batch_size=200, **kwargs)[source]¶

Bases: pytorch_lightning.LightningModule

PyTorch Lightning implementation of Augmented Multiscale Deep InfoMax (AMDIM)

Paper authors: Philip Bachman, R Devon Hjelm, William Buchwalter.

Model implemented by: William Falcon

This code is adapted to Lightning using the original author repo (the original repo).

Example

>>> from pl_bolts.models.self_supervised import AMDIM
...
>>> model = AMDIM(encoder='resnet18')

Train:

trainer = Trainer()
trainer.fit(model)

Parameters

datamodule (Union[str, LightningDataModule]) – A LightningDatamodule
encoder (Union[str, Module, LightningModule]) – an encoder string or model
image_channels (int) – 3
image_height (int) – pixels
encoder_feature_dim (int) – Called ndf in the paper, this is the representation size for the encoder.
embedding_fx_dim (int) – Output dim of the embedding function (nrkhs in the paper) (Reproducing Kernel Hilbert Spaces).
conv_block_depth (int) – Depth of each encoder block,
use_bn (bool) – If true will use batchnorm.
tclip (int) – soft clipping non-linearity to the scores after computing the regularization term and before computing the log-softmax. This is the ‘second trick’ used in the paper
learning_rate (int) – The learning rate
data_dir (str) – Where to store data
num_classes (int) – How many classes in the dataset
batch_size (int) – The batch size

static add_model_specific_args(parent_parser)[source]¶

configure_optimizers()[source]¶

forward(img_1, img_2)[source]¶

init_encoder()[source]¶

train_dataloader()[source]¶

training_step(batch, batch_nb)[source]¶

training_step_end(outputs)[source]¶

val_dataloader()[source]¶

validation_epoch_end(outputs)[source]¶

validation_step(batch, batch_nb)[source]¶

pl_bolts.models.self_supervised.amdim.amdim_module.cli_main()[source]¶

pl_bolts.models.self_supervised.amdim.datasets module¶

class pl_bolts.models.self_supervised.amdim.datasets.AMDIMPatchesPretraining[source]¶

Bases: object

” For pretraining we use the train transform for both train and val.

static cifar10(dataset_root, patch_size, patch_overlap, split='train')[source]¶

static imagenet(dataset_root, nb_classes, patch_size, patch_overlap, split='train')[source]¶

static stl(dataset_root, patch_size, patch_overlap, split=None)[source]¶

class pl_bolts.models.self_supervised.amdim.datasets.AMDIMPretraining[source]¶

Bases: object

” For pretraining we use the train transform for both train and val.

static cifar10(dataset_root, split='train')[source]¶

static cifar10_tiny(dataset_root, split='train')[source]¶

static get_dataset(datamodule, data_dir, split='train', **kwargs)[source]¶

static imagenet(dataset_root, nb_classes, split='train')[source]¶

static stl(dataset_root, split=None)[source]¶

pl_bolts.models.self_supervised.amdim.networks module¶

class pl_bolts.models.self_supervised.amdim.networks.AMDIMEncoder(dummy_batch, num_channels=3, encoder_feature_dim=64, embedding_fx_dim=512, conv_block_depth=3, encoder_size=32, use_bn=False)[source]¶

Bases: torch.nn.Module

_config_modules(x, output_widths, n_rkhs, use_bn)[source]¶: Configure the modules for extracting fake rkhs embeddings for infomax.

_forward_acts(x)[source]¶: Return activations from all layers.

forward(x)[source]¶

init_weights(init_scale=1.0)[source]¶: Run custom weight init for modules…

class pl_bolts.models.self_supervised.amdim.networks.Conv3x3(n_in, n_out, n_kern, n_stride, n_pad, use_bn=True, pad_mode='constant')[source]¶

Bases: torch.nn.Module

forward(x)[source]¶

class pl_bolts.models.self_supervised.amdim.networks.ConvResBlock(n_in, n_out, width, stride, pad, depth, use_bn)[source]¶

Bases: torch.nn.Module

forward(x)[source]¶

init_weights(init_scale=1.0)[source]¶: Do a fixup-ish init for each ConvResNxN in this block.

class pl_bolts.models.self_supervised.amdim.networks.ConvResNxN(n_in, n_out, width, stride, pad, use_bn=False)[source]¶

Bases: torch.nn.Module

forward(x)[source]¶

init_weights(init_scale=1.0)[source]¶

class pl_bolts.models.self_supervised.amdim.networks.FakeRKHSConvNet(n_input, n_output, use_bn=False)[source]¶

Bases: torch.nn.Module

forward(x)[source]¶

init_weights(init_scale=1.0)[source]¶

class pl_bolts.models.self_supervised.amdim.networks.MaybeBatchNorm2d(n_ftr, affine, use_bn)[source]¶

Bases: torch.nn.Module

forward(x)[source]¶

class pl_bolts.models.self_supervised.amdim.networks.NopNet(norm_dim=None)[source]¶

Bases: torch.nn.Module

forward(x)[source]¶

pl_bolts.models.self_supervised.amdim.transforms module¶

class pl_bolts.models.self_supervised.amdim.transforms.AMDIMEvalTransformsCIFAR10[source]¶

Bases: object

Transforms applied to AMDIM

Transforms:

transforms.ToTensor(),
normalize

Example:

x = torch.rand(5, 3, 32, 32)

transform = AMDIMEvalTransformsCIFAR10()
(view1, view2) = transform(x)

__call__(inp)[source]¶: Call self as a function.

class pl_bolts.models.self_supervised.amdim.transforms.AMDIMEvalTransformsImageNet128(height=128)[source]¶

Bases: object

Transforms applied to AMDIM

Transforms:

transforms.Resize(height + 6, interpolation=3),
transforms.CenterCrop(height),
transforms.ToTensor(),
normalize

Example:

x = torch.rand(5, 3, 128, 128)

transform = AMDIMEvalTransformsImageNet128()
view1 = transform(x)

__call__(inp)[source]¶: Call self as a function.

class pl_bolts.models.self_supervised.amdim.transforms.AMDIMEvalTransformsSTL10(height=64)[source]¶

Bases: object

Transforms applied to AMDIM

Transforms:

transforms.Resize(height + 6, interpolation=3),
transforms.CenterCrop(height),
transforms.ToTensor(),
normalize

Example:

x = torch.rand(5, 3, 64, 64)

transform = AMDIMTrainTransformsSTL10()
view1 = transform(x)

__call__(inp)[source]¶: Call self as a function.

class pl_bolts.models.self_supervised.amdim.transforms.AMDIMTrainTransformsCIFAR10[source]¶

Bases: object

Transforms applied to AMDIM

Transforms:

img_jitter,
col_jitter,
rnd_gray,
transforms.ToTensor(),
normalize

Example:

x = torch.rand(5, 3, 32, 32)

transform = AMDIMTrainTransformsCIFAR10()
(view1, view2) = transform(x)

__call__(inp)[source]¶: Call self as a function.

class pl_bolts.models.self_supervised.amdim.transforms.AMDIMTrainTransformsImageNet128(height=128)[source]¶

Bases: object

Transforms applied to AMDIM

Transforms:

img_jitter,
col_jitter,
rnd_gray,
transforms.ToTensor(),
normalize

Example:

x = torch.rand(5, 3, 128, 128)

transform = AMDIMTrainTransformsSTL10()
(view1, view2) = transform(x)

__call__(inp)[source]¶: Call self as a function.

class pl_bolts.models.self_supervised.amdim.transforms.AMDIMTrainTransformsSTL10(height=64)[source]¶

Bases: object

Transforms applied to AMDIM

Transforms:

img_jitter,
col_jitter,
rnd_gray,
transforms.ToTensor(),
normalize

Example:

x = torch.rand(5, 3, 64, 64)

transform = AMDIMTrainTransformsSTL10()
(view1, view2) = transform(x)

__call__(inp)[source]¶: Call self as a function.

pl_bolts.models.self_supervised.byol package¶

Submodules¶

pl_bolts.models.self_supervised.byol.byol_module module¶

class pl_bolts.models.self_supervised.byol.byol_module.BYOL(num_classes, learning_rate=0.2, weight_decay=1.5e-06, input_height=32, batch_size=32, num_workers=0, warmup_epochs=10, max_epochs=1000, **kwargs)[source]¶

Bases: pytorch_lightning.LightningModule

PyTorch Lightning implementation of Bootstrap Your Own Latent (BYOL)

Paper authors: Jean-Bastien Grill, Florian Strub, Florent Altché, Corentin Tallec, Pierre H. Richemond, Elena Buchatskaya, Carl Doersch, Bernardo Avila Pires, Zhaohan Daniel Guo, Mohammad Gheshlaghi Azar, Bilal Piot, Koray Kavukcuoglu, Rémi Munos, Michal Valko.

Model implemented by:

Annika Brundyn

Warning

Work in progress. This implementation is still being verified.

TODOs:

verify on CIFAR-10
verify on STL-10
pre-train on imagenet

Example:

import pytorch_lightning as pl
from pl_bolts.models.self_supervised import BYOL
from pl_bolts.datamodules import CIFAR10DataModule
from pl_bolts.models.self_supervised.simclr.transforms import (
    SimCLREvalDataTransform, SimCLRTrainDataTransform)

# model
model = BYOL(num_classes=10)

# data
dm = CIFAR10DataModule(num_workers=0)
dm.train_transforms = SimCLRTrainDataTransform(32)
dm.val_transforms = SimCLREvalDataTransform(32)

trainer = pl.Trainer()
trainer.fit(model, dm)

Train:

trainer = Trainer()
trainer.fit(model)

CLI command:

# cifar10
python byol_module.py --gpus 1

# imagenet
python byol_module.py
    --gpus 8
    --dataset imagenet2012
    --data_dir /path/to/imagenet/
    --meta_dir /path/to/folder/with/meta.bin/
    --batch_size 32

Parameters

datamodule – The datamodule
learning_rate (float) – the learning rate
weight_decay (float) – optimizer weight decay
input_height (int) – image input height
batch_size (int) – the batch size
num_workers (int) – number of workers
warmup_epochs (int) – num of epochs for scheduler warm up
max_epochs (int) – max epochs for scheduler

static add_model_specific_args(parent_parser)[source]¶

configure_optimizers()[source]¶

cosine_similarity(a, b)[source]¶

forward(x)[source]¶

on_train_batch_end(batch, batch_idx, dataloader_idx)[source]¶

Return type: None

shared_step(batch, batch_idx)[source]¶

training_step(batch, batch_idx)[source]¶

validation_step(batch, batch_idx)[source]¶

pl_bolts.models.self_supervised.byol.byol_module.cli_main()[source]¶

pl_bolts.models.self_supervised.byol.models module¶

class pl_bolts.models.self_supervised.byol.models.MLP(input_dim=2048, hidden_size=4096, output_dim=256)[source]¶

Bases: torch.nn.Module

forward(x)[source]¶

class pl_bolts.models.self_supervised.byol.models.SiameseArm(encoder=None)[source]¶

Bases: torch.nn.Module

forward(x)[source]¶

pl_bolts.models.self_supervised.cpc package¶

Submodules¶

pl_bolts.models.self_supervised.cpc.cpc_finetuner module¶

pl_bolts.models.self_supervised.cpc.cpc_finetuner.cli_main()[source]¶

pl_bolts.models.self_supervised.cpc.cpc_module module¶

CPC V2¶

class pl_bolts.models.self_supervised.cpc.cpc_module.CPCV2(datamodule=None, encoder_name='cpc_encoder', patch_size=8, patch_overlap=4, online_ft=True, task='cpc', num_workers=4, learning_rate=0.0001, data_dir='', batch_size=32, pretrained=None, **kwargs)[source]¶

Bases: pytorch_lightning.LightningModule

Parameters

datamodule (Optional[LightningDataModule]) – A Datamodule (optional). Otherwise set the dataloaders directly
encoder_name (str) – A string for any of the resnets in torchvision, or the original CPC encoder, or a custon nn.Module encoder
patch_size (int) – How big to make the image patches
patch_overlap (int) – How much overlap should each patch have.
online_ft (int) – Enable a 1024-unit MLP to fine-tune online
task (str) – Which self-supervised task to use (‘cpc’, ‘amdim’, etc…)
num_workers (int) – num dataloader worksers
learning_rate (int) – what learning rate to use
data_dir (str) – where to store data
batch_size (int) – batch size
pretrained (Optional[str]) – If true, will use the weights pretrained (using CPC) on Imagenet

_CPCV2__compute_final_nb_c(patch_size)[source]¶

_CPCV2__recover_z_shape(Z, b)[source]¶

static add_model_specific_args(parent_parser)[source]¶

configure_optimizers()[source]¶

forward(img_1)[source]¶

init_encoder()[source]¶

load_pretrained(encoder_name)[source]¶

shared_step(batch)[source]¶

training_step(batch, batch_nb)[source]¶

validation_step(batch, batch_nb)[source]¶

pl_bolts.models.self_supervised.cpc.networks module¶

class pl_bolts.models.self_supervised.cpc.networks.CPCResNet(sample_batch, block, layers, zero_init_residual=False, groups=1, width_per_group=64, replace_stride_with_dilation=None, norm_layer=None)[source]¶

Bases: torch.nn.Module

_make_layer(sample_batch, block, planes, blocks, stride=1, dilate=False, expansion=4)[source]¶

flatten(x)[source]¶

forward(x)[source]¶

class pl_bolts.models.self_supervised.cpc.networks.LNBottleneck(sample_batch, inplanes, planes, stride=1, downsample_conv=None, groups=1, base_width=64, dilation=1, norm_layer=None, expansion=4)[source]¶

Bases: torch.nn.Module

_LNBottleneck__init_layer_norms(x, conv1, conv2, conv3, downsample_conv)[source]¶

forward(x)[source]¶

pl_bolts.models.self_supervised.cpc.networks.conv1x1(in_planes, out_planes, stride=1)[source]¶: 1x1 convolution

pl_bolts.models.self_supervised.cpc.networks.conv3x3(in_planes, out_planes, stride=1, groups=1, dilation=1)[source]¶: 3x3 convolution with padding

pl_bolts.models.self_supervised.cpc.networks.cpc_resnet101(sample_batch, **kwargs)[source]¶

pl_bolts.models.self_supervised.cpc.networks.cpc_resnet50(sample_batch, **kwargs)[source]¶

pl_bolts.models.self_supervised.cpc.transforms module¶

class pl_bolts.models.self_supervised.cpc.transforms.CPCEvalTransformsCIFAR10(patch_size=8, overlap=4)[source]¶

Bases: object

Transforms used for CPC:

Parameters

patch_size – size of patches when cutting up the image into overlapping patches
overlap – how much to overlap patches

Transforms:

random_flip
transforms.ToTensor()
normalize
Patchify(patch_size=patch_size, overlap_size=overlap)

Example:

# in a regular dataset
CIFAR10(..., transforms=CPCEvalTransformsCIFAR10())

# in a DataModule
module = CIFAR10DataModule(PATH)
train_loader = module.train_dataloader(batch_size=32, transforms=CPCEvalTransformsCIFAR10())

__call__(inp)[source]¶: Call self as a function.

class pl_bolts.models.self_supervised.cpc.transforms.CPCEvalTransformsImageNet128(patch_size=32, overlap=16)[source]¶

Bases: object

Transforms used for CPC:

Parameters

patch_size – size of patches when cutting up the image into overlapping patches
overlap – how much to overlap patches

Transforms:

random_flip
transforms.ToTensor()
normalize
Patchify(patch_size=patch_size, overlap_size=patch_size // 2)

Example:

# in a regular dataset
Imagenet(..., transforms=CPCEvalTransformsImageNet128())

# in a DataModule
module = ImagenetDataModule(PATH)
train_loader = module.train_dataloader(batch_size=32, transforms=CPCEvalTransformsImageNet128())

__call__(inp)[source]¶: Call self as a function.

class pl_bolts.models.self_supervised.cpc.transforms.CPCEvalTransformsSTL10(patch_size=16, overlap=8)[source]¶

Bases: object

Transforms used for CPC:

Parameters

patch_size – size of patches when cutting up the image into overlapping patches
overlap – how much to overlap patches

Transforms:

random_flip
transforms.ToTensor()
normalize
Patchify(patch_size=patch_size, overlap_size=patch_size // 2)

Example:

# in a regular dataset
STL10(..., transforms=CPCEvalTransformsSTL10())

# in a DataModule
module = STL10DataModule(PATH)
train_loader = module.train_dataloader(batch_size=32, transforms=CPCEvalTransformsSTL10())

__call__(inp)[source]¶: Call self as a function.

class pl_bolts.models.self_supervised.cpc.transforms.CPCTrainTransformsCIFAR10(patch_size=8, overlap=4)[source]¶

Bases: object

Transforms used for CPC:

Parameters

patch_size – size of patches when cutting up the image into overlapping patches
overlap – how much to overlap patches

Transforms:

random_flip
img_jitter
col_jitter
rnd_gray
transforms.ToTensor()
normalize
Patchify(patch_size=patch_size, overlap_size=patch_size // 2)

Example:

# in a regular dataset
CIFAR10(..., transforms=CPCTrainTransformsCIFAR10())

# in a DataModule
module = CIFAR10DataModule(PATH)
train_loader = module.train_dataloader(batch_size=32, transforms=CPCTrainTransformsCIFAR10())

__call__(inp)[source]¶: Call self as a function.

class pl_bolts.models.self_supervised.cpc.transforms.CPCTrainTransformsImageNet128(patch_size=32, overlap=16)[source]¶

Bases: object

Transforms used for CPC:

Parameters

patch_size – size of patches when cutting up the image into overlapping patches
overlap – how much to overlap patches

Transforms:

random_flip
transforms.ToTensor()
normalize
Patchify(patch_size=patch_size, overlap_size=patch_size // 2)

Example:

# in a regular dataset
Imagenet(..., transforms=CPCTrainTransformsImageNet128())

# in a DataModule
module = ImagenetDataModule(PATH)
train_loader = module.train_dataloader(batch_size=32, transforms=CPCTrainTransformsImageNet128())

__call__(inp)[source]¶: Call self as a function.

class pl_bolts.models.self_supervised.cpc.transforms.CPCTrainTransformsSTL10(patch_size=16, overlap=8)[source]¶

Bases: object

Transforms used for CPC:

Parameters

patch_size – size of patches when cutting up the image into overlapping patches
overlap – how much to overlap patches

Transforms:

random_flip
img_jitter
col_jitter
rnd_gray
transforms.ToTensor()
normalize
Patchify(patch_size=patch_size, overlap_size=patch_size // 2)

Example:

# in a regular dataset
STL10(..., transforms=CPCTrainTransformsSTL10())

# in a DataModule
module = STL10DataModule(PATH)
train_loader = module.train_dataloader(batch_size=32, transforms=CPCTrainTransformsSTL10())

__call__(inp)[source]¶: Call self as a function.

pl_bolts.models.self_supervised.moco package¶

Submodules¶

pl_bolts.models.self_supervised.moco.callbacks module¶

class pl_bolts.models.self_supervised.moco.callbacks.MocoLRScheduler(initial_lr=0.03, use_cosine_scheduler=False, schedule=(120, 160), max_epochs=200)[source]¶

Bases: pytorch_lightning.Callback

on_epoch_start(trainer, pl_module)[source]¶

pl_bolts.models.self_supervised.moco.moco2_module module¶

Adapted from: https://github.com/facebookresearch/moco

class pl_bolts.models.self_supervised.moco.moco2_module.MocoV2(base_encoder='resnet18', emb_dim=128, num_negatives=65536, encoder_momentum=0.999, softmax_temperature=0.07, learning_rate=0.03, momentum=0.9, weight_decay=0.0001, datamodule=None, data_dir='./', batch_size=256, use_mlp=False, num_workers=8, *args, **kwargs)[source]¶

Bases: pytorch_lightning.LightningModule

PyTorch Lightning implementation of Moco

Paper authors: Xinlei Chen, Haoqi Fan, Ross Girshick, Kaiming He.

Code adapted from facebookresearch/moco to Lightning by:

William Falcon

Example::: from pl_bolts.models.self_supervised import MocoV2 model = MocoV2() trainer = Trainer() trainer.fit(model)

CLI command:

# cifar10
python moco2_module.py --gpus 1

# imagenet
python moco2_module.py
    --gpus 8
    --dataset imagenet2012
    --data_dir /path/to/imagenet/
    --meta_dir /path/to/folder/with/meta.bin/
    --batch_size 32

Parameters

base_encoder (Union[str, Module]) – torchvision model name or torch.nn.Module
emb_dim (int) – feature dimension (default: 128)
num_negatives (int) – queue size; number of negative keys (default: 65536)
encoder_momentum (float) – moco momentum of updating key encoder (default: 0.999)
softmax_temperature (float) – softmax temperature (default: 0.07)
learning_rate (float) – the learning rate
momentum (float) – optimizer momentum
weight_decay (float) – optimizer weight decay
datamodule (Optional[LightningDataModule]) – the DataModule (train, val, test dataloaders)
data_dir (str) – the directory to store data
batch_size (int) – batch size
use_mlp (bool) – add an mlp to the encoders
num_workers (int) – workers for the loaders

_batch_shuffle_ddp(x)[source]¶: Batch shuffle, for making use of BatchNorm. * Only support DistributedDataParallel (DDP) model. *

_batch_unshuffle_ddp(x, idx_unshuffle)[source]¶: Undo batch shuffle. * Only support DistributedDataParallel (DDP) model. *

_dequeue_and_enqueue(keys)[source]¶

_momentum_update_key_encoder()[source]¶: Momentum update of the key encoder

static add_model_specific_args(parent_parser)[source]¶

configure_optimizers()[source]¶

forward(img_q, img_k)[source]¶

Input:: im_q: a batch of query images im_k: a batch of key images
Output:: logits, targets

init_encoders(base_encoder)[source]¶: Override to add your own encoders

training_step(batch, batch_idx)[source]¶

validation_epoch_end(outputs)[source]¶

validation_step(batch, batch_idx)[source]¶

pl_bolts.models.self_supervised.moco.moco2_module.cli_main()[source]¶

pl_bolts.models.self_supervised.moco.moco2_module.concat_all_gather(tensor)[source]¶: Performs all_gather operation on the provided tensors. * Warning *: torch.distributed.all_gather has no gradient.

pl_bolts.models.self_supervised.moco.transforms module¶

class pl_bolts.models.self_supervised.moco.transforms.GaussianBlur(sigma=(0.1, 2.0))[source]¶

Bases: object

Gaussian blur augmentation in SimCLR https://arxiv.org/abs/2002.05709

__call__(x)[source]¶: Call self as a function.

class pl_bolts.models.self_supervised.moco.transforms.Moco2EvalCIFAR10Transforms(height=32)[source]¶

Bases: object

Moco 2 augmentation: https://arxiv.org/pdf/2003.04297.pdf

__call__(inp)[source]¶: Call self as a function.

class pl_bolts.models.self_supervised.moco.transforms.Moco2EvalImagenetTransforms(height=128)[source]¶

Bases: object

Moco 2 augmentation: https://arxiv.org/pdf/2003.04297.pdf

__call__(inp)[source]¶: Call self as a function.

class pl_bolts.models.self_supervised.moco.transforms.Moco2EvalSTL10Transforms(height=64)[source]¶

Bases: object

Moco 2 augmentation: https://arxiv.org/pdf/2003.04297.pdf

__call__(inp)[source]¶: Call self as a function.

class pl_bolts.models.self_supervised.moco.transforms.Moco2TrainCIFAR10Transforms(height=32)[source]¶

Bases: object

Moco 2 augmentation: https://arxiv.org/pdf/2003.04297.pdf

__call__(inp)[source]¶: Call self as a function.

class pl_bolts.models.self_supervised.moco.transforms.Moco2TrainImagenetTransforms(height=128)[source]¶

Bases: object

Moco 2 augmentation: https://arxiv.org/pdf/2003.04297.pdf

__call__(inp)[source]¶: Call self as a function.

class pl_bolts.models.self_supervised.moco.transforms.Moco2TrainSTL10Transforms(height=64)[source]¶

Bases: object

Moco 2 augmentation: https://arxiv.org/pdf/2003.04297.pdf

__call__(inp)[source]¶: Call self as a function.

pl_bolts.models.self_supervised.simclr package¶

Submodules¶

pl_bolts.models.self_supervised.simclr.simclr_finetuner module¶

pl_bolts.models.self_supervised.simclr.simclr_finetuner.cli_main()[source]¶

pl_bolts.models.self_supervised.simclr.simclr_module module¶

class pl_bolts.models.self_supervised.simclr.simclr_module.DensenetEncoder[source]¶

Bases: torch.nn.Module

forward(x)[source]¶

class pl_bolts.models.self_supervised.simclr.simclr_module.Projection(input_dim=2048, hidden_dim=2048, output_dim=128)[source]¶

Bases: torch.nn.Module

forward(x)[source]¶

class pl_bolts.models.self_supervised.simclr.simclr_module.SimCLR(batch_size, num_samples, warmup_epochs=10, lr=0.0001, opt_weight_decay=1e-06, loss_temperature=0.5, **kwargs)[source]¶

Bases: pytorch_lightning.LightningModule

Parameters

batch_size – the batch size
num_samples – num samples in the dataset
warmup_epochs – epochs to warmup the lr for
lr – the optimizer learning rate
opt_weight_decay – the optimizer weight decay
loss_temperature – the loss temperature

static add_model_specific_args(parent_parser)[source]¶

configure_optimizers()[source]¶

exclude_from_wt_decay(named_params, weight_decay, skip_list=['bias', 'bn'])[source]¶

forward(x)[source]¶

init_encoder()[source]¶

setup(stage)[source]¶

shared_step(batch, batch_idx)[source]¶

training_step(batch, batch_idx)[source]¶

validation_step(batch, batch_idx)[source]¶

pl_bolts.models.self_supervised.simclr.simclr_module.cli_main()[source]¶

pl_bolts.models.self_supervised.simclr.transforms module¶

class pl_bolts.models.self_supervised.simclr.transforms.GaussianBlur(kernel_size, min=0.1, max=2.0)[source]¶

Bases: object

__call__(sample)[source]¶: Call self as a function.

class pl_bolts.models.self_supervised.simclr.transforms.SimCLREvalDataTransform(input_height, s=1)[source]¶

Bases: object

Transforms for SimCLR

Transform:

Resize(input_height + 10, interpolation=3)
transforms.CenterCrop(input_height),
transforms.ToTensor()

Example:

from pl_bolts.models.self_supervised.simclr.transforms import SimCLREvalDataTransform

transform = SimCLREvalDataTransform(input_height=32)
x = sample()
(xi, xj) = transform(x)

__call__(sample)[source]¶: Call self as a function.

class pl_bolts.models.self_supervised.simclr.transforms.SimCLRTrainDataTransform(input_height, s=1)[source]¶

Bases: object

Transforms for SimCLR

Transform:

RandomResizedCrop(size=self.input_height)
RandomHorizontalFlip()
RandomApply([color_jitter], p=0.8)
RandomGrayscale(p=0.2)
GaussianBlur(kernel_size=int(0.1 * self.input_height))
transforms.ToTensor()

Example:

from pl_bolts.models.self_supervised.simclr.transforms import SimCLRTrainDataTransform

transform = SimCLRTrainDataTransform(input_height=32)
x = sample()
(xi, xj) = transform(x)

__call__(sample)[source]¶: Call self as a function.

Submodules¶

pl_bolts.models.self_supervised.evaluator module¶

class pl_bolts.models.self_supervised.evaluator.Flatten[source]¶

Bases: torch.nn.Module

forward(input_tensor)[source]¶

class pl_bolts.models.self_supervised.evaluator.SSLEvaluator(n_input, n_classes, n_hidden=512, p=0.1)[source]¶

Bases: torch.nn.Module

forward(x)[source]¶

pl_bolts.models.self_supervised.resnets module¶

class pl_bolts.models.self_supervised.resnets.ResNet(block, layers, num_classes=1000, zero_init_residual=False, groups=1, width_per_group=64, replace_stride_with_dilation=None, norm_layer=None, return_all_feature_maps=False)[source]¶

Bases: torch.nn.Module

_make_layer(block, planes, blocks, stride=1, dilate=False)[source]¶

forward(x)[source]¶

pl_bolts.models.self_supervised.resnets.resnet18(pretrained=False, progress=True, **kwargs)[source]¶: ResNet-18 model from “Deep Residual Learning for Image Recognition” :param _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet18.pretrained: If True, returns a model pre-trained on ImageNet :type _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet18.pretrained: bool :param _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet18.progress: If True, displays a progress bar of the download to stderr :type _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet18.progress: bool

pl_bolts.models.self_supervised.resnets.resnet34(pretrained=False, progress=True, **kwargs)[source]¶: ResNet-34 model from “Deep Residual Learning for Image Recognition” :param _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet34.pretrained: If True, returns a model pre-trained on ImageNet :type _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet34.pretrained: bool :param _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet34.progress: If True, displays a progress bar of the download to stderr :type _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet34.progress: bool

pl_bolts.models.self_supervised.resnets.resnet50(pretrained=False, progress=True, **kwargs)[source]¶: ResNet-50 model from “Deep Residual Learning for Image Recognition” :param _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet50.pretrained: If True, returns a model pre-trained on ImageNet :type _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet50.pretrained: bool :param _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet50.progress: If True, displays a progress bar of the download to stderr :type _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet50.progress: bool

pl_bolts.models.self_supervised.resnets.resnet50_bn(pretrained=False, progress=True, **kwargs)[source]¶: ResNet-50 model from “Deep Residual Learning for Image Recognition” :param _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet50_bn.pretrained: If True, returns a model pre-trained on ImageNet :type _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet50_bn.pretrained: bool :param _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet50_bn.progress: If True, displays a progress bar of the download to stderr :type _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet50_bn.progress: bool

pl_bolts.models.self_supervised.resnets.resnet101(pretrained=False, progress=True, **kwargs)[source]¶: ResNet-101 model from “Deep Residual Learning for Image Recognition” :param _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet101.pretrained: If True, returns a model pre-trained on ImageNet :type _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet101.pretrained: bool :param _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet101.progress: If True, displays a progress bar of the download to stderr :type _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet101.progress: bool

pl_bolts.models.self_supervised.resnets.resnet152(pretrained=False, progress=True, **kwargs)[source]¶: ResNet-152 model from “Deep Residual Learning for Image Recognition” :param _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet152.pretrained: If True, returns a model pre-trained on ImageNet :type _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet152.pretrained: bool :param _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet152.progress: If True, displays a progress bar of the download to stderr :type _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnet152.progress: bool

pl_bolts.models.self_supervised.resnets.resnext50_32x4d(pretrained=False, progress=True, **kwargs)[source]¶: ResNeXt-50 32x4d model from “Aggregated Residual Transformation for Deep Neural Networks” :param _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnext50_32x4d.pretrained: If True, returns a model pre-trained on ImageNet :type _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnext50_32x4d.pretrained: bool :param _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnext50_32x4d.progress: If True, displays a progress bar of the download to stderr :type _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnext50_32x4d.progress: bool

pl_bolts.models.self_supervised.resnets.resnext101_32x8d(pretrained=False, progress=True, **kwargs)[source]¶: ResNeXt-101 32x8d model from “Aggregated Residual Transformation for Deep Neural Networks” :param _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnext101_32x8d.pretrained: If True, returns a model pre-trained on ImageNet :type _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnext101_32x8d.pretrained: bool :param _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnext101_32x8d.progress: If True, displays a progress bar of the download to stderr :type _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.resnext101_32x8d.progress: bool

pl_bolts.models.self_supervised.resnets.wide_resnet50_2(pretrained=False, progress=True, **kwargs)[source]¶: Wide ResNet-50-2 model from “Wide Residual Networks” The model is the same as ResNet except for the bottleneck number of channels which is twice larger in every block. The number of channels in outer 1x1 convolutions is the same, e.g. last block in ResNet-50 has 2048-512-2048 channels, and in Wide ResNet-50-2 has 2048-1024-2048. :param _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.wide_resnet50_2.pretrained: If True, returns a model pre-trained on ImageNet :type _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.wide_resnet50_2.pretrained: bool :param _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.wide_resnet50_2.progress: If True, displays a progress bar of the download to stderr :type _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.wide_resnet50_2.progress: bool

pl_bolts.models.self_supervised.resnets.wide_resnet101_2(pretrained=False, progress=True, **kwargs)[source]¶: Wide ResNet-101-2 model from “Wide Residual Networks” The model is the same as ResNet except for the bottleneck number of channels which is twice larger in every block. The number of channels in outer 1x1 convolutions is the same, e.g. last block in ResNet-50 has 2048-512-2048 channels, and in Wide ResNet-50-2 has 2048-1024-2048. :param _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.wide_resnet101_2.pretrained: If True, returns a model pre-trained on ImageNet :type _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.wide_resnet101_2.pretrained: bool :param _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.wide_resnet101_2.progress: If True, displays a progress bar of the download to stderr :type _sphinx_paramlinks_pl_bolts.models.self_supervised.resnets.wide_resnet101_2.progress: bool

pl_bolts.models.self_supervised.ssl_finetuner module¶

class pl_bolts.models.self_supervised.ssl_finetuner.SSLFineTuner(backbone, in_features, num_classes, hidden_dim=1024)[source]¶

Bases: pytorch_lightning.LightningModule

Finetunes a self-supervised learning backbone using the standard evaluation protocol of a singler layer MLP with 1024 units

Example:

from pl_bolts.utils.self_supervised import SSLFineTuner
from pl_bolts.models.self_supervised import CPCV2
from pl_bolts.datamodules import CIFAR10DataModule
from pl_bolts.models.self_supervised.cpc.transforms import CPCEvalTransformsCIFAR10,
                                                            CPCTrainTransformsCIFAR10

# pretrained model
backbone = CPCV2.load_from_checkpoint(PATH, strict=False)

# dataset + transforms
dm = CIFAR10DataModule(data_dir='.')
dm.train_transforms = CPCTrainTransformsCIFAR10()
dm.val_transforms = CPCEvalTransformsCIFAR10()

# finetuner
finetuner = SSLFineTuner(backbone, in_features=backbone.z_dim, num_classes=backbone.num_classes)

# train
trainer = pl.Trainer()
trainer.fit(finetuner, dm)

# test
trainer.test(datamodule=dm)

Parameters

backbone – a pretrained model
in_features – feature dim of backbone outputs
num_classes – classes of the dataset
hidden_dim – dim of the MLP (1024 default used in self-supervised literature)

configure_optimizers()[source]¶

on_train_epoch_start()[source]¶

Return type: None

shared_step(batch)[source]¶

test_step(batch, batch_idx)[source]¶

training_step(batch, batch_idx)[source]¶

validation_step(batch, batch_idx)[source]¶

pl_bolts.models.vision package¶

Subpackages¶

pl_bolts.models.vision.image_gpt package¶

Submodules¶

pl_bolts.models.vision.image_gpt.gpt2 module¶

class pl_bolts.models.vision.image_gpt.gpt2.Block(embed_dim, heads)[source]¶

Bases: torch.nn.Module

forward(x)[source]¶

class pl_bolts.models.vision.image_gpt.gpt2.GPT2(embed_dim, heads, layers, num_positions, vocab_size, num_classes)[source]¶

Bases: pytorch_lightning.LightningModule

GPT-2 from language Models are Unsupervised Multitask Learners

Paper by: Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, Ilya Sutskever

Implementation contributed by:

Teddy Koker

Example:

from pl_bolts.models import GPT2

seq_len = 17
batch_size = 32
vocab_size = 16
x = torch.randint(0, vocab_size, (seq_len, batch_size))
model = GPT2(embed_dim=32, heads=2, layers=2, num_positions=seq_len, vocab_size=vocab_size, num_classes=4)
results = model(x)

_init_embeddings()[source]¶

_init_layers()[source]¶

_init_sos_token()[source]¶

forward(x, classify=False)[source]¶: Expect input as shape [sequence len, batch] If classify, return classification logits

pl_bolts.models.vision.image_gpt.igpt_module module¶

class pl_bolts.models.vision.image_gpt.igpt_module.ImageGPT(datamodule=None, embed_dim=16, heads=2, layers=2, pixels=28, vocab_size=16, num_classes=10, classify=False, batch_size=64, learning_rate=0.01, steps=25000, data_dir='.', num_workers=8, **kwargs)[source]¶

Bases: pytorch_lightning.LightningModule

Paper: Generative Pretraining from Pixels [original paper code].

Paper by: Mark Che, Alec Radford, Rewon Child, Jeff Wu, Heewoo Jun, Prafulla Dhariwal, David Luan, Ilya Sutskever

Implementation contributed by:

Teddy Koker

Original repo with results and more implementation details:

https://github.com/teddykoker/image-gpt

Example Results (Photo credits: Teddy Koker):

Default arguments:

Argument Defaults¶
Argument	Default	iGPT-S (Chen et al.)
–embed_dim	16	512
–heads	2	8
–layers	8	24
–pixels	28	32
–vocab_size	16	512
–num_classes	10	10
–batch_size	64	128
–learning_rate	0.01	0.01
–steps	25000	1000000

Example:

import pytorch_lightning as pl
from pl_bolts.models.vision import ImageGPT

dm = MNISTDataModule('.')
model = ImageGPT(dm)

pl.Trainer(gpu=4).fit(model)

As script:

cd pl_bolts/models/vision/image_gpt
python igpt_module.py --learning_rate 1e-2 --batch_size 32 --gpus 4

Parameters

datamodule (Optional[LightningDataModule]) – LightningDataModule
embed_dim (int) – the embedding dim
heads (int) – number of attention heads
layers (int) – number of layers
pixels (int) – number of input pixels
vocab_size (int) – vocab size
num_classes (int) – number of classes in the input
classify (bool) – true if should classify
batch_size (int) – the batch size
learning_rate (float) – learning rate
steps (int) – number of steps for cosine annealing
data_dir (str) – where to store data
num_workers (int) – num_data workers

static add_model_specific_args(parent_parser)[source]¶

configure_optimizers()[source]¶

forward(x, classify=False)[source]¶

test_epoch_end(outs)[source]¶

test_step(batch, batch_idx)[source]¶

training_step(batch, batch_idx)[source]¶

validation_epoch_end(outs)[source]¶

validation_step(batch, batch_idx)[source]¶

pl_bolts.models.vision.image_gpt.igpt_module._shape_input(x)[source]¶: shape batch of images for input into GPT2 model

pl_bolts.models.vision.image_gpt.igpt_module.cli_main()[source]¶

Submodules¶

pl_bolts.models.vision.pixel_cnn module¶

PixelCNN Implemented by: William Falcon Reference: https://arxiv.org/pdf/1905.09272.pdf (page 15) Accessed: May 14, 2020

class pl_bolts.models.vision.pixel_cnn.PixelCNN(input_channels, hidden_channels=256, num_blocks=5)[source]¶

Bases: torch.nn.Module

Implementation of Pixel CNN.

Paper authors: Aaron van den Oord, Nal Kalchbrenner, Oriol Vinyals, Lasse Espeholt, Alex Graves, Koray Kavukcuoglu

Implemented by:

William Falcon

Example:

>>> from pl_bolts.models.vision import PixelCNN
>>> import torch
...
>>> model = PixelCNN(input_channels=3)
>>> x = torch.rand(5, 3, 64, 64)
>>> out = model(x)
...
>>> out.shape
torch.Size([5, 3, 64, 64])

conv_block(input_channels)[source]¶

forward(z)[source]¶

pl_bolts.models.vision.unet module¶

class pl_bolts.models.vision.unet.DoubleConv(in_ch, out_ch)[source]¶

Bases: torch.nn.Module

[ Conv2d => BatchNorm (optional) => ReLU ] x 2

forward(x)[source]¶

class pl_bolts.models.vision.unet.Down(in_ch, out_ch)[source]¶

Bases: torch.nn.Module

Downscale with MaxPool => DoubleConvolution block

forward(x)[source]¶

class pl_bolts.models.vision.unet.UNet(num_classes, num_layers=5, features_start=64, bilinear=False)[source]¶

Bases: torch.nn.Module

PyTorch Lightning implementation of U-Net: Convolutional Networks for Biomedical Image Segmentation

Paper authors: Olaf Ronneberger, Philipp Fischer, Thomas Brox

Model implemented by:

Warning

Work in progress. This implementation is still being verified.

Parameters

num_classes (int) – Number of output classes required
num_layers (int) – Number of layers in each side of U-net (default 5)
features_start (int) – Number of features in first layer (default 64)
bilinear (bool) – Whether to use bilinear interpolation or transposed convolutions (default) for upsampling.

forward(x)[source]¶

class pl_bolts.models.vision.unet.Up(in_ch, out_ch, bilinear=False)[source]¶

Bases: torch.nn.Module

Upsampling (by either bilinear interpolation or transpose convolutions) followed by concatenation of feature map from contracting path, followed by DoubleConv.

forward(x1, x2)[source]¶

Submodules¶

pl_bolts.models.mnist_module module¶

class pl_bolts.models.mnist_module.LitMNIST(hidden_dim=128, learning_rate=0.001, batch_size=32, num_workers=4, data_dir='', **kwargs)[source]¶

Bases: pytorch_lightning.LightningModule

static add_model_specific_args(parent_parser)[source]¶

configure_optimizers()[source]¶

forward(x)[source]¶

prepare_data()[source]¶

test_dataloader()[source]¶

test_step(batch, batch_idx)[source]¶

train_dataloader()[source]¶

training_step(batch, batch_idx)[source]¶

val_dataloader()[source]¶

validation_step(batch, batch_idx)[source]¶

pl_bolts.models.mnist_module.cli_main()[source]¶

pl_bolts.losses package¶

Submodules¶

pl_bolts.losses.rl module¶

Loss functions for the RL models

pl_bolts.losses.rl.double_dqn_loss(batch, net, target_net, gamma=0.99)[source]¶

Calculates the mse loss using a mini batch from the replay buffer. This uses an improvement to the original DQN loss by using the double dqn. This is shown by using the actions of the train network to pick the value from the target network. This code is heavily commented in order to explain the process clearly :type _sphinx_paramlinks_pl_bolts.losses.rl.double_dqn_loss.batch: Tuple[Tensor, Tensor] :param _sphinx_paramlinks_pl_bolts.losses.rl.double_dqn_loss.batch: current mini batch of replay data :type _sphinx_paramlinks_pl_bolts.losses.rl.double_dqn_loss.net: Module :param _sphinx_paramlinks_pl_bolts.losses.rl.double_dqn_loss.net: main training network :type _sphinx_paramlinks_pl_bolts.losses.rl.double_dqn_loss.target_net: Module :param _sphinx_paramlinks_pl_bolts.losses.rl.double_dqn_loss.target_net: target network of the main training network :type _sphinx_paramlinks_pl_bolts.losses.rl.double_dqn_loss.gamma: float :param _sphinx_paramlinks_pl_bolts.losses.rl.double_dqn_loss.gamma: discount factor

Return type: Tensor
Returns: loss

pl_bolts.losses.rl.dqn_loss(batch, net, target_net, gamma=0.99)[source]¶

Calculates the mse loss using a mini batch from the replay buffer :type _sphinx_paramlinks_pl_bolts.losses.rl.dqn_loss.batch: Tuple[Tensor, Tensor] :param _sphinx_paramlinks_pl_bolts.losses.rl.dqn_loss.batch: current mini batch of replay data :type _sphinx_paramlinks_pl_bolts.losses.rl.dqn_loss.net: Module :param _sphinx_paramlinks_pl_bolts.losses.rl.dqn_loss.net: main training network :type _sphinx_paramlinks_pl_bolts.losses.rl.dqn_loss.target_net: Module :param _sphinx_paramlinks_pl_bolts.losses.rl.dqn_loss.target_net: target network of the main training network :type _sphinx_paramlinks_pl_bolts.losses.rl.dqn_loss.gamma: float :param _sphinx_paramlinks_pl_bolts.losses.rl.dqn_loss.gamma: discount factor

Return type: Tensor
Returns: loss

pl_bolts.losses.rl.per_dqn_loss(batch, batch_weights, net, target_net, gamma=0.99)[source]¶

Calculates the mse loss with the priority weights of the batch from the PER buffer :type _sphinx_paramlinks_pl_bolts.losses.rl.per_dqn_loss.batch: Tuple[Tensor, Tensor] :param _sphinx_paramlinks_pl_bolts.losses.rl.per_dqn_loss.batch: current mini batch of replay data :type _sphinx_paramlinks_pl_bolts.losses.rl.per_dqn_loss.batch_weights: List :param _sphinx_paramlinks_pl_bolts.losses.rl.per_dqn_loss.batch_weights: how each of these samples are weighted in terms of priority :type _sphinx_paramlinks_pl_bolts.losses.rl.per_dqn_loss.net: Module :param _sphinx_paramlinks_pl_bolts.losses.rl.per_dqn_loss.net: main training network :type _sphinx_paramlinks_pl_bolts.losses.rl.per_dqn_loss.target_net: Module :param _sphinx_paramlinks_pl_bolts.losses.rl.per_dqn_loss.target_net: target network of the main training network :type _sphinx_paramlinks_pl_bolts.losses.rl.per_dqn_loss.gamma: float :param _sphinx_paramlinks_pl_bolts.losses.rl.per_dqn_loss.gamma: discount factor

Return type: Tuple[Tensor, ndarray]
Returns: loss and batch_weights

pl_bolts.losses.self_supervised_learning module¶

class pl_bolts.losses.self_supervised_learning.AmdimNCELoss(tclip)[source]¶

Bases: torch.nn.Module

forward(anchor_representations, positive_representations, mask_mat)[source]¶

Compute the NCE scores for predicting r_src->r_trg. :param _sphinx_paramlinks_pl_bolts.losses.self_supervised_learning.AmdimNCELoss.forward.anchor_representations: (batch_size, emb_dim) :param _sphinx_paramlinks_pl_bolts.losses.self_supervised_learning.AmdimNCELoss.forward.positive_representations: (emb_dim, n_batch * w* h) (ie: nb_feat_vectors x embedding_dim) :param _sphinx_paramlinks_pl_bolts.losses.self_supervised_learning.AmdimNCELoss.forward.mask_mat: (n_batch_gpu, n_batch)

Output:: raw_scores : (n_batch_gpu, n_locs) nce_scores : (n_batch_gpu, n_locs) lgt_reg : scalar

class pl_bolts.losses.self_supervised_learning.CPCTask(num_input_channels, target_dim=64, embed_scale=0.1)[source]¶

Bases: torch.nn.Module

Loss used in CPC

compute_loss_h(targets, preds, i)[source]¶

forward(Z)[source]¶

class pl_bolts.losses.self_supervised_learning.FeatureMapContrastiveTask(comparisons='00, 11', tclip=10.0, bidirectional=True)[source]¶

Bases: torch.nn.Module

Performs an anchor, positive negative pair comparison for each each tuple of feature maps passed.

# extract feature maps
pos_0, pos_1, pos_2 = encoder(x_pos)
anc_0, anc_1, anc_2 = encoder(x_anchor)

# compare only the 0th feature maps
task = FeatureMapContrastiveTask('00')
loss, regularizer = task((pos_0), (anc_0))

# compare (pos_0 to anc_1) and (pos_0, anc_2)
task = FeatureMapContrastiveTask('01, 02')
losses, regularizer = task((pos_0, pos_1, pos_2), (anc_0, anc_1, anc_2))
loss = losses.sum()

# compare (pos_1 vs a anc_random)
task = FeatureMapContrastiveTask('0r')
loss, regularizer = task((pos_0, pos_1, pos_2), (anc_0, anc_1, anc_2))

Parameters

comparisons (str) – groupings of feature map indices to compare (zero indexed, ‘r’ means random) ex: ‘00, 1r’
tclip (float) – stability clipping value
bidirectional (bool) – if true, does the comparison both ways

# with bidirectional the comparisons are done both ways
task = FeatureMapContrastiveTask('01, 02')

# will compare the following:
# 01: (pos_0, anc_1), (anc_0, pos_1)
# 02: (pos_0, anc_2), (anc_0, pos_2)

_FeatureMapContrastiveTask__cache_dimension_masks(*args)[source]¶

_FeatureMapContrastiveTask__compare_maps(m1, m2)[source]¶

_sample_src_ftr(r_cnv, masks)[source]¶

feat_size_w_mask(w, feature_map)[source]¶

forward(anchor_maps, positive_maps)[source]¶

Takes in a set of tuples, each tuple has two feature maps with all matching dimensions

Example

>>> import torch
>>> from pytorch_lightning import seed_everything
>>> seed_everything(0)
0
>>> a1 = torch.rand(3, 5, 2, 2)
>>> a2 = torch.rand(3, 5, 2, 2)
>>> b1 = torch.rand(3, 5, 2, 2)
>>> b2 = torch.rand(3, 5, 2, 2)
...
>>> task = FeatureMapContrastiveTask('01, 11')
...
>>> losses, regularizer = task((a1, a2), (b1, b2))
>>> losses
tensor([2.2351, 2.1902])
>>> regularizer
tensor(0.0324)

static parse_map_indexes(comparisons)[source]¶

Example:

>>> FeatureMapContrastiveTask.parse_map_indexes('11')
[(1, 1)]
>>> FeatureMapContrastiveTask.parse_map_indexes('11,59')
[(1, 1), (5, 9)]
>>> FeatureMapContrastiveTask.parse_map_indexes('11,59, 2r')
[(1, 1), (5, 9), (2, -1)]

pl_bolts.losses.self_supervised_learning.nt_xent_loss(out_1, out_2, temperature)[source]¶: Loss used in SimCLR

pl_bolts.losses.self_supervised_learning.tanh_clip(x, clip_val=10.0)[source]¶: soft clip values to the range [-clip_val, +clip_val]

pl_bolts.optimizers package¶

Submodules¶

pl_bolts.optimizers.lars_scheduling module¶

References

class pl_bolts.optimizers.lars_scheduling.LARSWrapper(optimizer, eta=0.02, clip=True, eps=1e-08)[source]¶

Bases: object

Wrapper that adds LARS scheduling to any optimizer. This helps stability with huge batch sizes.

Parameters

optimizer – torch optimizer
eta – LARS coefficient (trust)
clip – True to clip LR
eps – adaptive_lr stability coefficient

step()[source]¶

update_p(p, group, weight_decay)[source]¶

property param_groups[source]¶

property state[source]¶

pl_bolts.optimizers.lr_scheduler module¶

class pl_bolts.optimizers.lr_scheduler.LinearWarmupCosineAnnealingLR(optimizer, warmup_epochs, max_epochs, warmup_start_lr=0.0, eta_min=0.0, last_epoch=-1)[source]¶

Bases: torch.optim.lr_scheduler._LRScheduler

Sets the learning rate of each parameter group to follow a linear warmup schedule between warmup_start_lr and base_lr followed by a cosine annealing schedule between base_lr and eta_min.

Warning

It is recommended to call step() for LinearWarmupCosineAnnealingLR after each iteration as calling it after each epoch will keep the starting lr at warmup_start_lr for the first epoch which is 0 in most cases.

Warning

passing epoch to step() is being deprecated and comes with an EPOCH_DEPRECATION_WARNING. It calls the _get_closed_form_lr() method for this scheduler instead of get_lr(). Though this does not change the behavior of the scheduler, when passing epoch param to step(), the user should call the step() function before calling train and validation methods.

Parameters

optimizer (Optimizer) – Wrapped optimizer.
warmup_epochs (int) – Maximum number of iterations for linear warmup
max_epochs (int) – Maximum number of iterations
warmup_start_lr (float) – Learning rate to start the linear warmup. Default: 0.
eta_min (float) – Minimum learning rate. Default: 0.
last_epoch (int) – The index of last epoch. Default: -1.

Example

>>> layer = nn.Linear(10, 1)
>>> optimizer = Adam(layer.parameters(), lr=0.02)
>>> scheduler = LinearWarmupCosineAnnealingLR(optimizer, warmup_epochs=10, max_epochs=40)
>>> #
>>> # the default case
>>> for epoch in range(40):
...     # train(...)
...     # validate(...)
...     scheduler.step()
>>> #
>>> # passing epoch param case
>>> for epoch in range(40):
...     scheduler.step(epoch)
...     # train(...)
...     # validate(...)

_get_closed_form_lr()[source]¶

Called when epoch is passed as a param to the step function of the scheduler.

Return type: List[float]

get_lr()[source]¶

Compute learning rate using chainable form of the scheduler

Return type: List[float]

pl_bolts.transforms package¶

Subpackages¶

pl_bolts.transforms.self_supervised package¶

Submodules¶

pl_bolts.transforms.self_supervised.ssl_transforms module¶

class pl_bolts.transforms.self_supervised.ssl_transforms.Patchify(patch_size, overlap_size)[source]¶

Bases: object

__call__(x)[source]¶: Call self as a function.

class pl_bolts.transforms.self_supervised.ssl_transforms.RandomTranslateWithReflect(max_translation)[source]¶

Bases: object

Translate image randomly Translate vertically and horizontally by n pixels where n is integer drawn uniformly independently for each axis from [-max_translation, max_translation]. Fill the uncovered blank area with reflect padding.

__call__(old_image)[source]¶: Call self as a function.

Submodules¶

pl_bolts.transforms.dataset_normalizations module¶

pl_bolts.transforms.dataset_normalizations.cifar10_normalization()[source]¶

pl_bolts.transforms.dataset_normalizations.imagenet_normalization()[source]¶

pl_bolts.transforms.dataset_normalizations.stl10_normalization()[source]¶