Build pretrained InceptionV3 Parameters ---------- output_blocks : list of int Indices of blocks to return featur

Transform `coords` with `coeffs`.

one_batch function is extremely slow with large datasets. This is caching the result as an optimization.

Add `val` to calculate updated smoothed value.

Change `name` from camel to snake style.

Get children of `m`.

Compose `funcs`

Create and initialize `nn.Conv2d` layer. `padding` defaults to `ks//2`.

Create a seuence Conv2d->BatchNorm2d->LeakyReLu layer.

Cut off the body of a typically pretrained `model` at `cut` (int) or cut the model as specified by `cut(model)` (function).

Create an `ImageDataBunch` from `LabelLists` `lls` with potential `ds_tfms`.

Pass a `dummy_batch` in evaluation mode in `m` with `size`.

Code cell with the command to show the doc of `func_name`.

Return the model maybe wrapped inside `model`.

get total, used and free memory (in MBs) for gpu `id`. if `id` is not passed, currently selected torch device is used

get used memory (in MBs) for the currently selected gpu id, after emptying the cache

Convert from torch style `image` to numpy/matplotlib style.

Return the shape of the first weight layer in `m`.

Return the slice of `a` corresponding to `idxs`.

Initialize `m` weights with `func` and set `bias` to 0.

A special labelling method for language models.

Build differential learning rates from `lr`.

Pass a dummy input through the model `m` to get the various sizes of activations.

Queues up an asynchronous write request to Tensorboard.

Resnet block of `nf` features. `conv_kwargs` are passed to `conv_layer`.

Reset the hidden states.

Split the data according to the indexes in `valid_idx`.

Convert a list of `nums` to their tokens.

Recursively detach lists of tensors in `b `; put them on the CPU if `cpu=True`.

Recursively put `b` on `device`.

Return list of trainable params in `m`.

Set `tfms` to be applied to the xs of the train and validation set.

Try to convert `o` to int, default to `o` if not possible.

Same as `nn.MSELoss`, but flattens input and target.

Return one hidden state.

Writes list of images as tensors to Tensorboard.

Sequence of batchnorm (if `bn`), dropout (with `p`) and linear (`n_in`,`n_out`) layers followed by `actn`.

Concatenate the `arrs` along the batch dimension.

Create and initialize a `nn.Conv1d` layer with spectral normalization.

Creates a `Config` from `fpath`.

Create optimizer with `lr` learning rate and `wd` weight decay.

To throw a clear error message when the data wasn't labeled.

Return the column indexes of `names` in `df`.

Returns appropriate `Dataset` for validation, training, or test (`ds_type`).

Download `url` to `dest` unless it exists and not `overwrite`.

Return a dropout mask of the same type as `x`, size `sz`, with probability `p` to cancel an element.

Only keep elements for which `func` returns `True`.

Recursively get the first element of `x`.

Get the filenames in `cols` of `df` with `folder` in front of them, `suffix` at the end.

Return the arguments of `func`.

Code cell containing `code` that may be `hidden`.

Empty cell of type `ctype`.

Return all the functions of module `mod`.

Return list of files in `c` that are images. `check_ext` will filter to `image_extensions`.

Pick the monitored value.

Get the `path` in the config file.

Return a `Hook` that stores activations of `module` in `self.stored`