Torch Core¶

This module contains all the basic functions we need in other modules of the fastai library (split with core that contains the ones not requiring pytorch). Its documentation can easily be skipped at a first read, unless you want to know what a given fuction does.

In [ ]:

from fastai.gen_doc.nbdoc import *
from fastai.torch_core import * 

Global constants¶

AdamW = partial(optim.Adam, betas=(0.9,0.99))

[source]

bn_types = (nn.BatchNorm1d, nn.BatchNorm2d, nn.BatchNorm3d)

[source]

default_device = torch.device('cuda') if torch.cuda.is_available() else torch.device('cpu')

[source]

Functions that operate conversions¶

In [ ]:

show_doc(flatten_model, full_name='flatten')

`flatten`[source]

flatten(m)

Flattens all the layers of m.

In [ ]:

show_doc(model2half)

`model2half`[source]

model2half(model:Module) → Module

Convert model to half precision except the batchnorm layers.

In [ ]:

show_doc(np2model_tensor)

`np2model_tensor`[source]

np2model_tensor(a)

Tranform numpy array a to a tensor of the same type.

In [ ]:

show_doc(requires_grad, doc_string=False)

`requires_grad`[source]

requires_grad(m:Module, b:Optional[bool]=None) → Optional[bool]

If b is None, returns the requires_grad state of the first layer of m. Otherwise, sets requires_grad=b in all children of m.

In [ ]:

show_doc(tensor)

`tensor`[source]

tensor(x:Any, rest) → Tensor

Like torch.as_tensor, but handle lists too, and can pass multiple vector elements directly

Ensures x is a torch Tensor.

In [ ]:

show_doc(to_data)

`to_data`[source]

to_data(b:ItemsList)

Recursively map lists of items in b to their wrapped data.

In [ ]:

show_doc(to_detach)

`to_detach`[source]

to_detach(b:Tensors, cpu:bool=True)

Recursively detach lists of tensors in b, puts them on the CPU if cpu=True.

In [ ]:

show_doc(to_device)

`to_device`[source]

to_device(b:Tensors, device:device)

Ensure b is on device.

In [ ]:

show_doc(to_half, doc_string=False)

`to_half`[source]

to_half(b:Collection[Tensor]) → Collection[Tensor]

Put the input of the batch b in half precision.

In [ ]:

show_doc(to_np)

`to_np`[source]

to_np(x)

Convert x to a numpy array.

Functions to deal with model initialization¶

In [ ]:

show_doc(apply_init)

`apply_init`[source]

apply_init(m, init_func:LayerFunc)

Initialize all non-batchnorm layers of m with init_func.

In [ ]:

show_doc(apply_leaf)

`apply_leaf`[source]

apply_leaf(m:Module, f:LayerFunc)

Apply f to children of m.

In [ ]:

show_doc(cond_init)

`cond_init`[source]

cond_init(m:Module, init_func:LayerFunc)

Initialize the non-batchnorm layers of m with init_func

In [ ]:

show_doc(in_channels)

`in_channels`[source]

in_channels(m:Module) → List[int]

Return the shape of the first weight layer in m.

Functions to get information of a model¶

In [ ]:

show_doc(children)

`children`[source]

children(m:Module) → ModuleList

Get children of m.

In [ ]:

show_doc(first_layer)

`first_layer`[source]

first_layer(m:Module) → Module

Retrieve first layer in a module m.

In [ ]:

show_doc(last_layer)

`last_layer`[source]

last_layer(m:Module) → Module

Retrieve last layer in a module m.

In [ ]:

show_doc(num_children)

`num_children`[source]

num_children(m:Module) → int

Get number of children modules in module m.

In [ ]:

show_doc(range_children)

`range_children`[source]

range_children(m:Module) → Iterator[int]

Return iterator of len of children of m.

In [ ]:

show_doc(trainable_params)

`trainable_params`[source]

trainable_params(m:Module) → ParamList

Return list of trainable params in m.

Functions to deal with BatchNorm layers¶

In [ ]:

show_doc(bn2float)

`bn2float`[source]

bn2float(module:Module) → Module

If module is batchnorm don't use half precision.

In [ ]:

show_doc(set_bn_eval)

`set_bn_eval`[source]

set_bn_eval(m:Module)

Set bn layers in eval mode for all recursive children of m.

In [ ]:

show_doc(split_bn_bias)

`split_bn_bias`[source]

split_bn_bias(layer_groups:ModuleList) → ModuleList

Sort each layer in layer_groups into batchnorm (bn_types) and non-batchnorm groups.

Other functions¶

In [ ]:

show_doc(calc_loss)

`calc_loss`[source]

calc_loss(y_pred:Tensor, y_true:Tensor, loss_func:LossFunction)

Calculate loss between y_pred and y_true using loss_class and bs.

In [ ]:

show_doc(data_collate)

`data_collate`[source]

data_collate(batch:ItemsList) → Tensor

Convert batch items to tensor data.

In [ ]:

show_doc(model_type)

`model_type`[source]

model_type(dtype)

Return the torch type corresponding to dtype.

In [ ]:

show_doc(np_address)

`np_address`[source]

np_address(x:ndarray) → int

Address of x in memory

In [ ]:

show_doc(split_model, doc_string=False)

`split_model`[source]

split_model(model:Module, splits:Collection[Union[Module, ModuleList]], want_idxs:bool=False)

Splits the model according to the layer in splits. If splits are layers, the model is split at those (not included) sequentially. If want_idxs is True, the corresponding indexes are returned. If splits are lists of layers, the model is split according to those.

In [ ]:

show_doc(split_model_idx)

`split_model_idx`[source]

split_model_idx(model:Module, idxs:Collection[int]) → ModuleList

Split model according to the indices in idxs.

In [ ]:

show_doc(trange_of)

`trange_of`[source]

trange_of(x)

Return a tensor from a range that has the same length as x.

Undocumented Methods - Methods moved below this line will intentionally be hidden¶

In [ ]:

show_doc(tensor__array__)

`tensorarray`[source]

tensor__array__(dtype=None)

New Methods - Please document or move to the undocumented section¶

In [ ]:

show_doc(init_default)

`init_default`[source]

init_default(m:Module, func:LayerFunc='kaiming_normal_')

In [ ]:

show_doc(log_uniform)

`log_uniform`[source]

log_uniform(low, high, size:Optional[List[int]]=None) → FloatOrTensor

Draw 1 or shape=size random floats from uniform dist: min=log(low), max=log(high).

In [ ]:

show_doc(grab_idx)

`grab_idx`[source]

grab_idx(x, i, batch_first:bool=True)

In [ ]:

show_doc(uniform_int)

`uniform_int`[source]

uniform_int(low:int, high:int, size:Optional[List[int]]=None) → IntOrTensor

Generate int or tensor size of ints between low and high (included).

In [ ]:

show_doc(to_cpu)

`to_cpu`[source]

to_cpu(b:ItemsList)

Recursively map lists of tensors in b to the cpu.

In [ ]:

show_doc(logit)

`logit`[source]

logit(x:Tensor) → Tensor

Logit of x, clamped to avoid inf

In [ ]:

show_doc(FloatItem)

`class` `FloatItem`[source]

FloatItem(obj) :: ItemBase

All transformable dataset items use this type.

In [ ]:

show_doc(logit_)

`logit_`[source]

logit_(x:Tensor) → Tensor

Inplace logit of x, clamped to avoid inf

In [ ]:

show_doc(rand_bool)

`rand_bool`[source]

rand_bool(p:float, size:Optional[List[int]]=None) → BoolOrTensor

Draw 1 or shape=size random booleans (True occuring probability p).

In [ ]:

show_doc(one_param)

`one_param`[source]

one_param(m:Module) → Tensor

In [ ]:

show_doc(uniform)

`uniform`[source]

uniform(low:Number, high:Number=None, size:Optional[List[int]]=None) → FloatOrTensor

Draw 1 or shape=size random floats from uniform dist: min=low, max=high.

Torch Core¶

Global constants¶

Functions that operate conversions¶

flatten[source]

model2half[source]

np2model_tensor[source]

requires_grad[source]

tensor[source]

to_data[source]

to_detach[source]

to_device[source]

to_half[source]

to_np[source]

Functions to deal with model initialization¶

apply_init[source]

apply_leaf[source]

cond_init[source]

in_channels[source]

Functions to get information of a model¶

children[source]

first_layer[source]

last_layer[source]

num_children[source]

range_children[source]

trainable_params[source]

Functions to deal with BatchNorm layers¶

bn2float[source]

set_bn_eval[source]

split_bn_bias[source]

Other functions¶

calc_loss[source]

data_collate[source]

model_type[source]

np_address[source]

split_model[source]

split_model_idx[source]

trange_of[source]

Undocumented Methods - Methods moved below this line will intentionally be hidden¶

tensor__array__[source]

New Methods - Please document or move to the undocumented section¶

init_default[source]

log_uniform[source]

grab_idx[source]

uniform_int[source]

to_cpu[source]

logit[source]

class FloatItem[source]

logit_[source]

rand_bool[source]

one_param[source]

uniform[source]

`flatten`[source]

`model2half`[source]

`np2model_tensor`[source]

`requires_grad`[source]

`tensor`[source]

`to_data`[source]

`to_detach`[source]

`to_device`[source]

`to_half`[source]

`to_np`[source]

`apply_init`[source]

`apply_leaf`[source]

`cond_init`[source]

`in_channels`[source]

`children`[source]

`first_layer`[source]

`last_layer`[source]

`num_children`[source]

`range_children`[source]

`trainable_params`[source]

`bn2float`[source]

`set_bn_eval`[source]

`split_bn_bias`[source]

`calc_loss`[source]

`data_collate`[source]

`model_type`[source]

`np_address`[source]

`split_model`[source]

`split_model_idx`[source]

`trange_of`[source]

`tensorarray`[source]

`init_default`[source]

`log_uniform`[source]

`grab_idx`[source]

`uniform_int`[source]

`to_cpu`[source]

`logit`[source]

`class` `FloatItem`[source]

`logit_`[source]

`rand_bool`[source]

`one_param`[source]

`uniform`[source]