Wrap together some coords `flow` with a `size`.

Decorator for pixel tfm funcs.

Manage all of the registered `callbacks` and `metrics`, smoothing loss by momentum `beta`.

Trainer for `model` using `data` to minimize `loss_func` with optimizer `opt_func`.

Merge a shortcut with the result of the module by adding them or concatenating thme if `dense=True`.

Basic `ItemList` for text data.

Decorator for coord tfm funcs.

Support applying transforms to segmentation masks data in `px`.

A parameter in a function used in `anno_parser` or `call_parse`

Register a lone parameter `p` in a module.

Upsample by `scale` from `ni` filters to `nf` (default `ni`), using `nn.PixelShuffle`, `icnr` init, and `weight_norm`.

Decorator for affine tfm funcs.

InceptionC block patched for FID computation

Same as `func`, but flattens input and target.

Support applying transforms to image data in `px`.

`PreProcessor` that numericalizes the tokens in `ds`.

Create a smooth moving average for a value (loss, etc) using `beta`.

Basic class for single classification labels.

Bind a `DataLoader` to a `torch.device`.

InceptionA block patched for FID computation

Create several hooks on the modules in `ms` with `hook_func`.

An `ItemList` for each of `train` and `valid` (optional `test`).

Dropout with probability `p` that is consistent on the seq_len dimension.

Used to \"step\" from start,end (`vals`) over `n_iter` iterations on a schedule defined by `func`

Like `nn.Sequential`, but with ModuleList semantics, and can access module input

Sigmoid module with range `(low,x_max)`

Basic item for <code>text</code> data in numericalized `ids`.

`PreProcessor` that tokenizes the texts in `ds`.

Base item for collaborative filtering, subclasses `TabularLine`.

Creates a default config file 'config.yml' in $FASTAI_HOME (default `~/.fastai/`)

Upsample by `scale` from `ni` filters to `nf` (default `ni`), using `nn.PixelShuffle`, `icnr` init, and `weight_norm`.

Basic block of a Transformer model.

Should be used for a dummy label.

Flatten `x` to a single dimension, often used at the end of a model. `full` for rank-1 tensor

Basic class for float items.

Create a hook on `m` with `hook_func`.

`ItemList` suitable for computer vision.

Support applying transforms to a `flow` of points.

Writes model image output to Tensorboard.

Create a layer that simply calls `func` with `x`

Basic class for multi-classification labels.

An random `ItemBase` of size `noise_sz`.

A context manager to evaluate `loss_func` with none reduce and weights on the CPU inside.

`PreProcessor` that opens the filenames and read the texts.

Encode the position with a sinusoid.

Apply a single layer Quasi-Recurrent Neural Network (QRNN) to an input sequence.

Self attention layer for nd.

A sequential module that passes the reset call to its children.

Decorator for lighting tfm funcs.

Contain the correspondence between numbers and tokens and numericalize.

A module that warps another layer in which some weights will be replaced by 0 during training.

Layer that concats `AdaptiveAvgPool2d` and `AdaptiveMaxPool2d`.

Expand the `target` to match the `output` size before applying `crit`.

Callback for GANLearners that writes to Tensorboard. Extends LearnerTensorboardWriter and adds output image writes.

Wrap a `func` in a callback for metrics computation.

`PreProcessor` that create `classes` from `ds.items` and handle the mapping.

`Learner` suitable for collaborative filtering.

Create a U-Net from a given architecture.

Create a U-Net from a given architecture.

Base dot model for collaborative filtering.

Apply dropout with probabily `embed_p` to an embedding layer `emb`.

Subclass `TabularModel` to create a NN suitable for collaborative filtering.

Basic `ItemList` for dummy labels.

First InceptionE block patched for FID computation

Second InceptionE block patched for FID computation

Wrapper around `loss_funcC` (for the critic) and `loss_funcG` (for the generator).

Wrapper around a `generator` and a `critic` to create a GAN.

Handles GAN Training.

Trace allocated and peaked GPU memory usage (deltas).

Request object for model histogram writes to Tensorboard.

Request object for model histogram writes to Tensorboard.

Writes model histograms to Tensorboard.

Request object for model image output writes to Tensorboard.

Pretrained InceptionV3 network returning feature maps

Causes `learn` to go on a mock training from `start_lr` to `end_lr` for `num_it` iterations.

A `LabelList` for each of `train` and `valid` (optional `test`).

Subclass of RNNLearner for predictions.

Transforms the tokens in `dataset` to a stream of contiguous batches for language modelling.

To go on top of a RNNCore module and create a Language Model.

Adapt the loss function `crit` to go with mixup.

Convenience object that holds the original, real(target) and generated versions of a single image fed to a model.

A context manager to evaluate `model` on the CPU inside.