Formatted enum documentation.

Return env var value if it's defined and not an empty string, or return Unknown

Returns link to `ft` in source code.

Return the position to insert a given function doc in a notebook.

Get the path to `key` in the config file.

Create a language model from `arch` and its `config`, maybe `pretrained`.

Factory method to convert a batch of model images to a list of ModelImageSet.

Return two lists, one for the model parameters in FP16 and one for the master parameters in FP32.

Find module given a source path. Assume it belongs to fastai directory

Return predictions and targets on `ds_type` dataset.

Read the default class processors if none have been set.

creates a collapsible bootstrap card for `show_test`

Returns link to pytorch docs of `ft`.

Free traceback from references to locals() in each frame to avoid circular reference leading to gc.collect() unable to reclaim memory

Gets the indices for the most similar images in `ds_type` dataset

Return the minimal state for export.

Absolute path of `fastai/tests` directory

Create a text classifier from `arch` and its `config`, maybe `pretrained`.

Sorts `ds_type` dataset by top losses and returns dataset and sorted indices.

Utility func to easily create a list of flip, rotate, `zoom`, warp, lighting transforms.

Create and format widget set.

get total, used and free memory (in MBs) for each available gpu

Returns `True` if the current device is GPU, `False` otherwise.

Is the code running in Google Colaboratory?

Is the code running in the ipython environment (jupyter including)

Return `Path(path)/Path(fname)`, `path` defaults to current dir.

To resume training at `epoch` directly.

Returns the `n` largest indices from a numpy array `arr`.

Load the state in `fn` to create an empty `LabelList` for inference.

Inplace logit of `x`, clamped to avoid inf

Finds `this_test` entries from test_registry.json

Return a Dropdown widget with specified `handler`.

starts with conv layer - `ch_in` channels in - then has `num_blocks` `ResLayer`

Make a horizontal box with `children` and `layout`.

Returns an image widget for specified file name `img`.

Make a vertical box with `children` and `layout`.

Creates dictionary test format to match doctest api

Pool MultiBatchEncoder outputs into one vector [last_hidden, max_pool, avg_pool].

Copy `master_params` to `model_params`.

Mean squared error between `pred` and `targ`.

Convert `model` to half precision except the batchnorm layers.

Copy the `model_params` gradients to `master_params` for the optimizer step.

Get the path to fastai pretrained models in the config file.

Return the torch type corresponding to `dtype`.

Sorted descending list of largest non-diagonal entries of confusion matrix, presented as actual, predicted, number of occurrences.

Create a new `ItemList` from `items`, keeping the same attributes.

Create a new `OptimWrapper` from `self` with another `layer_groups` but the same hyper-parameters.

Handler for 'Next Batch' button click. Delete all flagged images and renders next batch.

Get number of cpus

Return the number of output features for `model`.

Convert a list of tokens `t` to their ids.

Record the loss before any other callback has a chance to modify it.

Called after backprop but before optimizer step. Useful for true weight decay in AdamW.

Record learning rate and momentum at beginning of batch.

Handle end of processing one batch with `loss`.

Take a step in lr,mom sched, start next stepper when the current one is complete.

Callback function that writes batch end appropriate data to Tensorboard.

Take one step forward on the annealing schedule for the optim params.

Called after forward pass but before loss has been computed.

Called after the step of the optimizer but before the gradients are zeroed.

About to start learning.

Register the `Hooks` on `self.modules`.

Handle end of training, `exception` is an `Exception` or False if no exceptions during training.

Remove the `Hooks`.

One-hot encode `x` with `c` classes.

Return `ImageSegment` object create from mask in file `fn`. If `div`, divides pixel values by 255.

Read the text in `fn`.

Pass a dummy input through the model to get the various sizes. Returns (res,x,hooks) if `full`

Split iterables `a` in equal parts of size `sz`