Machine Learning

Navigation

Verified details

These details have been verified by PyPI
Maintainers
Avatar for schulzr from gravatar.com schulzr

Unverified details

These details have not been verified by PyPI
Project links
Meta
  • License: MIT
  • Author: Robert Schulz
  • Tags python , Machine Learning
Classifiers
Report project as malware

Project description

RSProduction MachineLearning

This project provides some usefull machine learning functionality.

Table of Contents

1 dataset

TOC

1.1 HMDB51 : torch.utils.data.dataset.Dataset

TOC

Description

Dataset class for HMDB51.

Example

from rsp.ml.dataset import HMDB51
import rsp.ml.multi_transforms as multi_transforms
import cv2 as cv

transforms = multi_transforms.Compose([
    multi_transforms.Color(1.5, p=0.5),
    multi_transforms.Stack()
])
ds = HMDB51('train', fold=1, transforms=transforms)

for X, T in ds:
  for x in X.permute(0, 2, 3, 1):
    img_color = x[:, :, :3].numpy()
    img_depth = x[:, :, 3].numpy()

    cv.imshow('color', img_color)
    cv.imshow('depth', img_depth)

    cv.waitKey(30)

1.1.1 __init__

TOC

Description

Initializes a new instance.

Parameters

Name Type Description
split str Dataset split [train
fold int Fold number. The dataset is split into 3 folds. If fold is None, all folds will be loaded.
cache_dir str, default = None Directory to store the downloaded files. If set to None, the default cache directory will be used
force_reload bool, default = False If set to True, the dataset will be reloaded
target_size (int, int), default = (400, 400) Size of the frames. The frames will be resized to this size.
sequence_length int, default = 30 Length of the sequences
transforms rsp.ml.multi_transforms.Compose = default = rsp.ml.multi_transforms.Compose([]) Transformations, that will be applied to each input sequence. See documentation of rsp.ml.multi_transforms for more details.
verbose bool, default = False If set to True, the progress will be printed.

1.2 Kinetics : torch.utils.data.dataset.Dataset

TOC

Description

Dataset class for the Kinetics dataset.

Example

from rsp.ml.dataset import Kinetics

ds = Kinetics(split='train', type=400)

for X, T in ds:
    print(X)

1.2.1 __init__

TOC

Description

Initializes a new instance.

Parameters

Name Type Description
split str Dataset split [train
sequence_length int, default = 60 Length of the sequences
type int, default = 400 Type of the kineticts dataset. Currently only 400 is supported.
frame_size (int, int), default = (400, 400) Size of the frames. The frames will be resized to this size.
transforms rsp.ml.multi_transforms.Compose = default = rsp.ml.multi_transforms.Compose([]) Transformations, that will be applied to each input sequence. See documentation of rsp.ml.multi_transforms for more details.
cache_dir str, default = None Directory to store the downloaded files. If set to None, the default cache directory will be used
num_threads int, default = 0 Number of threads to use for downloading the files.
verbose bool, default = True If set to True, the progress and additional information will be printed.

1.3 TUCHRI : torch.utils.data.dataset.Dataset

TOC

Description

Dataset class for the Robot Interaction Dataset by University of Technology Chemnitz (TUCHRI).

1.3.1 __init__

TOC

Description

Initializes a new instance.

Parameters

Name Type Description
phase str Dataset phase [train
load_depth_data bool, default = True Load depth data
sequence_length int, default = 30 Length of the sequences
num_classes int, default = 10 Number of classes
transforms rsp.ml.multi_transforms.Compose = default = rsp.ml.multi_transforms.Compose([]) Transformations, that will be applied to each input sequence. See documentation of rsp.ml.multi_transforms for more details.

1.3.2 get_uniform_sampler

TOC

1.3.3 load_backgrounds

TOC

Description

Loads the background images.

Parameters

Name Type Description
load_depth_data bool, default = True If set to True, the depth images will be loaded as well.

1.4 TUCRID : torch.utils.data.dataset.Dataset

TOC

Description

Dataset class for the Robot Interaction Dataset by University of Technology Chemnitz (TUCRID).

Example

from rsp.ml.dataset import TUCRID
from rsp.ml.dataset import ReplaceBackgroundRGBD
import rsp.ml.multi_transforms as multi_transforms
import cv2 as cv

backgrounds = TUCRID.load_backgrounds_color()
transforms = multi_transforms.Compose([
    ReplaceBackgroundRGBD(backgrounds),
    multi_transforms.Stack()
])

ds = TUCRID('train', transforms=transforms)

for X, T in ds:
  for x in X.permute(0, 2, 3, 1):
    img_color = x[:, :, :3].numpy()
    img_depth = x[:, :, 3].numpy()

    cv.imshow('color', img_color)
    cv.imshow('depth', img_depth)

    cv.waitKey(30)

1.4.1 __init__

TOC

Description

Initializes a new instance.

Parameters

Name Type Description
phase str Dataset phase [train
load_depth_data bool, default = True Load depth data
sequence_length int, default = 30 Length of the sequences
num_classes int, default = 10 Number of classes
transforms rsp.ml.multi_transforms.Compose = default = rsp.ml.multi_transforms.Compose([]) Transformations, that will be applied to each input sequence. See documentation of rsp.ml.multi_transforms for more details.

1.4.2 get_uniform_sampler

TOC

1.4.3 load_backgrounds

TOC

Description

Loads the background images.

Parameters

Name Type Description
load_depth_data bool, default = True If set to True, the depth images will be loaded as well.

1.5 UCF101 : torch.utils.data.dataset.Dataset

TOC

Description

An abstract class representing a :class:Dataset.

All datasets that represent a map from keys to data samples should subclass it. All subclasses should overwrite :meth:__getitem__, supporting fetching a data sample for a given key. Subclasses could also optionally overwrite :meth:__len__, which is expected to return the size of the dataset by many :class:~torch.utils.data.Sampler implementations and the default options of :class:~torch.utils.data.DataLoader. Subclasses could also optionally implement :meth:__getitems__, for speedup batched samples loading. This method accepts list of indices of samples of batch and returns list of samples.

.. note:: :class:~torch.utils.data.DataLoader by default constructs an index sampler that yields integral indices. To make it work with a map-style dataset with non-integral indices/keys, a custom sampler must be provided.

Example

from rsp.ml.dataset import UCF101
import rsp.ml.multi_transforms as multi_transforms
import cv2 as cv

transforms = multi_transforms.Compose([
    multi_transforms.Color(1.5, p=0.5),
    multi_transforms.Stack()
])
ds = UCF101('train', fold=1, transforms=transforms)

for X, T in ds:
  for x in X.permute(0, 2, 3, 1):
    img_color = x[:, :, :3].numpy()
    img_depth = x[:, :, 3].numpy()

    cv.imshow('color', img_color)
    cv.imshow('depth', img_depth)

    cv.waitKey(30)

1.5.1 __init__

TOC

Description

Initializes a new instance.

Parameters

Name Type Description
split str Dataset split [train
fold int Fold number. The dataset is split into 3 folds. If fold is None, all folds will be loaded.
cache_dir str, default = None Directory to store the downloaded files. If set to None, the default cache directory will be used
force_reload bool, default = False If set to True, the dataset will be reloaded
target_size (int, int), default = (400, 400) Size of the frames. The frames will be resized to this size.
sequence_length int, default = 30 Length of the sequences
transforms rsp.ml.multi_transforms.Compose = default = rsp.ml.multi_transforms.Compose([]) Transformations, that will be applied to each input sequence. See documentation of rsp.ml.multi_transforms for more details.
verbose bool, default = False If set to True, the progress will be printed.

1.6 UTKinectAction3D : torch.utils.data.dataset.Dataset

TOC

Description

Dataset class for the UTKinectAction3D dataset.

Parameters

split : str Dataset split [train|val] cache_dir : str, default = None Directory to store the downloaded files. If set to None, the default cache directory will be used force_reload : bool, default = False If set to True, the dataset will be reloaded target_size : (int, int), default = (400, 400) Size of the frames. The frames will be resized to this size. sequence_length : int, default = 30 Length of the sequences transforms : rsp.ml.multi_transforms.Compose = default = rsp.ml.multi_transforms.Compose([]) Transformations, that will be applied to each input sequence. See documentation of rsp.ml.multi_transforms for more details. verbose : bool, default = False If set to True, the progress will be printed.

Example

from rsp.ml.dataset import UTKinectAction3D
import rsp.ml.multi_transforms as multi_transforms
import cv2 as cv

transforms = multi_transforms.Compose([
    multi_transforms.Color(1.5, p=0.5),
    multi_transforms.Stack()
])
ds = UTKinectAction3D('train', transforms=transforms)

for X, T in ds:
  for x in X.permute(0, 2, 3, 1):
    img_color = x[:, :, :3].numpy()
    img_depth = x[:, :, 3].numpy()

    cv.imshow('color', img_color)
    cv.imshow('depth', img_depth)

    cv.waitKey(30)

1.6.1 __init__

TOC

Description

Initialize self. See help(type(self)) for accurate signature.

2 metrics

TOC

The module rsp.ml.metrics provides some functionality to quantify the quality of predictions.

2.1 AUROC

TOC

Description

Calculates the Area under the Receiver Operation Chracteristic Curve.

Parameters

Name Type Description
Y torch.Tensor Prediction
T torch.Tensor True values
num_thresholds int, default = 100 Number of thresholds to compute.

Returns

Receiver Operation Chracteristic Area under the Curve : float

2.2 F1_Score

TOC

Description

F1 Score. Expected input shape: (batch_size, num_classes)

Parameters

Name Type Description
Y torch.Tensor Prediction
T torch.Tensor True values
threshold float All values that are greater than or equal to the threshold are considered a positive class.

Returns

F1 Score : float

Equations

$precision = \frac{TP}{TP + FP}$

$recall = \frac{TP}{TP + FN}$

$F_1 = \frac{2 \cdot precision \cdot recall}{precision + recall} = \frac{2 \cdot TP}{2 \cdot TP + FP + FN}$

Example

import rsp.ml.metrics as m

Y = torch.tensor([
  [0.1, 0.1, 0.8],
  [0.03, 0.95, 0.02],
  [0.05, 0.9, 0.05],
  [0.01, 0.87, 0.12],
  [0.04, 0.03, 0.93],
  [0.94, 0.02, 0.06]
])
T = torch.tensor([
  [0, 0, 1],
  [1, 0, 0],
  [0, 1, 0],
  [0, 1, 0],
  [0, 0, 1],
  [1, 0, 0]
])

f1score = m.F1_Score(Y, T)

print(f1score) --> 0.5

2.3 FN

TOC

Description

False negatives. Expected input shape: (batch_size, num_classes)

Parameters

Name Type Description
Y torch.Tensor Prediction
T torch.Tensor True values
threshold float All values that are greater than or equal to the threshold are considered a positive class.

Returns

False negatives : int

Example

import rsp.ml.metrics as m
import torch

Y = torch.tensor([
  [0.1, 0.1, 0.8],
  [0.03, 0.95, 0.02],
  [0.05, 0.9, 0.05],
  [0.01, 0.87, 0.12],
  [0.04, 0.03, 0.93],
  [0.94, 0.02, 0.06]
])
T = torch.tensor([
  [0, 0, 1],
  [1, 0, 0],
  [0, 1, 0],
  [0, 1, 0],
  [0, 0, 1],
  [1, 0, 0]
])

fn = m.FN(Y, T)
print(fn) -> 1

2.4 FP

TOC

Description

False positives. Expected input shape: (batch_size, num_classes)

Parameters

Name Type Description
Y torch.Tensor Prediction
T torch.Tensor True values
threshold float All values that are greater than or equal to the threshold are considered a positive class.

Returns

False positives : int

Example

import rsp.ml.metrics as m
import torch

Y = torch.tensor([
  [0.1, 0.1, 0.8],
  [0.03, 0.95, 0.02],
  [0.05, 0.9, 0.05],
  [0.01, 0.87, 0.12],
  [0.04, 0.03, 0.93],
  [0.94, 0.02, 0.06]
])
T = torch.tensor([
  [0, 0, 1],
  [1, 0, 0],
  [0, 1, 0],
  [0, 1, 0],
  [0, 0, 1],
  [1, 0, 0]
])

fp = m.FP(Y, T)
print(fp) -> 1

2.5 FPR

TOC

Description

False positive rate. Expected input shape: (batch_size, num_classes)

Parameters

Name Type Description
Y torch.Tensor Prediction
T torch.Tensor True values
threshold float All values that are greater than or equal to the threshold are considered a positive class.

Returns

False positive rate : float

Example

import rsp.ml.metrics as m
import torch

Y = torch.tensor([
  [0.1, 0.1, 0.8],
  [0.03, 0.95, 0.02],
  [0.05, 0.9, 0.05],
  [0.01, 0.87, 0.12],
  [0.04, 0.03, 0.93],
  [0.94, 0.02, 0.06]
])
T = torch.tensor([
  [0, 0, 1],
  [1, 0, 0],
  [0, 1, 0],
  [0, 1, 0],
  [0, 0, 1],
  [1, 0, 0]
])

fpr = m.FPR(Y, T)
print(fpr) -> 0.08333333333333333

2.6 ROC

TOC

Description

Calculates the receiver operating characteristic: computes False Positive Rates and True positive Rates for num_thresholds aligned between 0 and 1

Parameters

Name Type Description
Y torch.Tensor Prediction
T torch.Tensor True values
num_thresholds int, default = 100 Number of thresholds to compute.

Returns

(False Positive Rates, True Positive Rates) for 100 different thresholds : (List[float], List[float])

Example

import rsp.ml.metrics as m
import torch
import torch.nn.functional as F

num_elements = 100000
num_classes = 7

T = []
for i in range(num_elements):
  true_class = torch.randint(0, num_classes, (1,))
  t = F.one_hot(true_class, num_classes=num_classes)
  T.append(t)
T = torch.cat(T)

dist = torch.normal(T.float(), 1.5)
Y = F.softmax(dist, dim = 1)
FPRs, TPRs = m.ROC(Y, T)

2.7 TN

TOC

Description

True negatives. Expected input shape: (batch_size, num_classes)

Parameters

Name Type Description
Y torch.Tensor Prediction
T torch.Tensor True values
threshold float All values that are greater than or equal to the threshold are considered a positive class.

Returns

True negatives : int

Example

import rsp.ml.metrics as m
import torch

Y = torch.tensor([
  [0.1, 0.1, 0.8],
  [0.03, 0.95, 0.02],
  [0.05, 0.9, 0.05],
  [0.01, 0.87, 0.12],
  [0.04, 0.03, 0.93],
  [0.94, 0.02, 0.06]
])
T = torch.tensor([
  [0, 0, 1],
  [1, 0, 0],
  [0, 1, 0],
  [0, 1, 0],
  [0, 0, 1],
  [1, 0, 0]
])

tn = m.TN(Y, T)
print(tn) -> 11

2.8 TP

TOC

Description

True positives. Expected input shape: (batch_size, num_classes)

Parameters

Name Type Description
Y torch.Tensor Prediction
T torch.Tensor True values
threshold float All values that are greater than or equal to the threshold are considered a positive class.

Returns

True positives : int

Example

import rsp.ml.metrics as m
import torch

Y = torch.tensor([
  [0.1, 0.1, 0.8],
  [0.03, 0.95, 0.02],
  [0.05, 0.9, 0.05],
  [0.01, 0.87, 0.12],
  [0.04, 0.03, 0.93],
  [0.94, 0.02, 0.06]
])
T = torch.tensor([
  [0, 0, 1],
  [1, 0, 0],
  [0, 1, 0],
  [0, 1, 0],
  [0, 0, 1],
  [1, 0, 0]
])

tp = m.TP(Y, T)
print(tp) -> 5

2.9 TPR

TOC

Description

True positive rate. Expected input shape: (batch_size, num_classes)

Parameters

Name Type Description
Y torch.Tensor Prediction
T torch.Tensor True values
threshold float All values that are greater than or equal to the threshold are considered a positive class.

Returns

True positive rate : float

Example

import rsp.ml.metrics as m
import torch

Y = torch.tensor([
  [0.1, 0.1, 0.8],
  [0.03, 0.95, 0.02],
  [0.05, 0.9, 0.05],
  [0.01, 0.87, 0.12],
  [0.04, 0.03, 0.93],
  [0.94, 0.02, 0.06]
])
T = torch.tensor([
  [0, 0, 1],
  [1, 0, 0],
  [0, 1, 0],
  [0, 1, 0],
  [0, 0, 1],
  [1, 0, 0]
])

tpr = m.TPR(Y, T)
print(tpr) -> 0.8333333333333334

2.10 confusion_matrix

TOC

Description

Calculates the confusion matrix. Expected input shape: (batch_size, num_classes)

Parameters

Name Type Description
Y torch.Tensor Prediction
T torch.Tensor True values

Returns

Confusion matrix : torch.Tensor

Example

import rsp.ml.metrics as m
import torch

Y = torch.tensor([
  [0.1, 0.1, 0.8],
  [0.03, 0.95, 0.02],
  [0.05, 0.9, 0.05],
  [0.01, 0.87, 0.12],
  [0.04, 0.03, 0.93],
  [0.94, 0.02, 0.06]
])
T = torch.tensor([
  [0, 0, 1],
  [1, 0, 0],
  [0, 1, 0],
  [0, 1, 0],
  [0, 0, 1],
  [1, 0, 0]
])

conf_mat = m.confusion_matrix(Y, T)
print(conf_mat) -> tensor([
  [1, 1, 0],
  [0, 2, 0],
  [0, 0, 2]
])

2.11 plot_ROC

TOC

Description

Plot the receiver operating characteristic.

Parameters

Name Type Description
Y torch.Tensor Prediction
T torch.Tensor True values
num_thresholds int, default = 100 Number of thresholds to compute.
title str, optional, default = 'Confusion Matrix' Title of the plot
class_curves bool, default = False Plot ROC curve for each class
labels str, optional, default = None Class labels -> automatic labeling C000, ..., CXXX if labels is None
plt_show bool, optional, default = False Set to True to show the plot
save_file_name str, optional, default = None If not None, the plot is saved under the specified save_file_name.

Returns

Image of the confusion matrix : np.array

2.12 plot_confusion_matrix

TOC

Description

Plot the confusion matrix

Parameters

Name Type Description
confusion_matrix torch.Tensor Confusion matrix
labels str, optional, default = None Class labels -> automatic labeling C000, ..., CXXX if labels is None
cmap str, optional, default = 'Blues' Seaborn cmap, see https://r02b.github.io/seaborn_palettes/
xlabel str, optional, default = 'Predicted label' X-Axis label
ylabel str, optional, default = 'True label' Y-Axis label
title str, optional, default = 'Confusion Matrix' Title of the plot
plt_show bool, optional, default = False Set to True to show the plot
save_file_name str, optional, default = None If not None, the plot is saved under the specified save_file_name.

Returns

Image of the confusion matrix : np.array

2.13 precision

TOC

Description

Precision. Expected input shape: (batch_size, num_classes)

Parameters

Name Type Description
Y torch.Tensor Prediction
T torch.Tensor True values
threshold float All values that are greater than or equal to the threshold are considered a positive class.

Returns

Precision : float

Equations

$precision = \frac{TP}{TP + FP}$

Example

import rsp.ml.metrics as m
import torch

Y = torch.tensor([
  [0.1, 0.1, 0.8],
  [0.03, 0.95, 0.02],
  [0.05, 0.9, 0.05],
  [0.01, 0.87, 0.12],
  [0.04, 0.03, 0.93],
  [0.94, 0.02, 0.06]
])
T = torch.tensor([
  [0, 0, 1],
  [1, 0, 0],
  [0, 1, 0],
  [0, 1, 0],
  [0, 0, 1],
  [1, 0, 0]
])

precision = m.precision(Y, T)
print(precision) -> 0.8333333333333334

2.14 recall

TOC

Description

Recall. Expected input shape: (batch_size, num_classes)

Parameters

Name Type Description
Y torch.Tensor Prediction
T torch.Tensor True values
threshold float All values that are greater than or equal to the threshold are considered a positive class.

Returns

Recall : float

Equations

$recall = \frac{TP}{TP + FN}$

Example

import rsp.ml.metrics as m
import torch

Y = torch.tensor([
  [0.1, 0.1, 0.8],
  [0.03, 0.95, 0.02],
  [0.05, 0.9, 0.05],
  [0.01, 0.87, 0.12],
  [0.04, 0.03, 0.93],
  [0.94, 0.02, 0.06]
])
T = torch.tensor([
  [0, 0, 1],
  [1, 0, 0],
  [0, 1, 0],
  [0, 1, 0],
  [0, 0, 1],
  [1, 0, 0]
])

recall = m.recall(Y, T)
print(recall) -> 0.8333333333333334

2.15 top_10_accuracy

TOC

Description

Top 10 accuracy. Expected input shape: (batch_size, num_classes)

Parameters

Name Type Description
Y torch.Tensor Prediction
T torch.Tensor True values

Returns

Top 10 accuracy -> top k accuracy | k = 10 : float

Example

import rsp.ml.metrics as m

Y = torch.tensor([
  [0.1, 0.1, 0.8],
  [0.03, 0.95, 0.02],
  [0.05, 0.9, 0.05],
  [0.01, 0.87, 0.12],
  [0.04, 0.03, 0.93],
  [0.94, 0.02, 0.06]
])
T = torch.tensor([
  [0, 0, 1],
  [1, 0, 0],
  [0, 1, 0],
  [0, 1, 0],
  [0, 0, 1],
  [1, 0, 0]
])

top_10_accuracy = m.top_10_accuracy(Y, T, k = 3)

print(top_10_accuracy) --> 1.0

2.16 top_1_accuracy

TOC

Description

Top 1 accuracy. Expected input shape: (batch_size, num_classes)

Parameters

Name Type Description
Y torch.Tensor Prediction
T torch.Tensor True values

Returns

Top 1 accuracy -> top k accuracy | k = 1 : float

Example

import rsp.ml.metrics as m

Y = torch.tensor([
  [0.1, 0.1, 0.8],
  [0.03, 0.95, 0.02],
  [0.05, 0.9, 0.05],
  [0.01, 0.87, 0.12],
  [0.04, 0.03, 0.93],
  [0.94, 0.02, 0.06]
])
T = torch.tensor([
  [0, 0, 1],
  [1, 0, 0],
  [0, 1, 0],
  [0, 1, 0],
  [0, 0, 1],
  [1, 0, 0]
])

top_1_accuracy = m.top_1_accuracy(Y, T, k = 3)

print(top_1_accuracy) --> 0.8333333333333334

2.17 top_2_accuracy

TOC

Description

Top 2 accuracy. Expected input shape: (batch_size, num_classes)

Parameters

Name Type Description
Y torch.Tensor Prediction
T torch.Tensor True values

Returns

Top 2 accuracy -> top k accuracy | k = 2 : float

Example

import rsp.ml.metrics as m

Y = torch.tensor([
  [0.1, 0.1, 0.8],
  [0.03, 0.95, 0.02],
  [0.05, 0.9, 0.05],
  [0.01, 0.87, 0.12],
  [0.04, 0.03, 0.93],
  [0.94, 0.02, 0.06]
])
T = torch.tensor([
  [0, 0, 1],
  [1, 0, 0],
  [0, 1, 0],
  [0, 1, 0],
  [0, 0, 1],
  [1, 0, 0]
])

top_2_accuracy = m.top_2_accuracy(Y, T, k = 3)

print(top_2_accuracy) --> 1.0

2.18 top_3_accuracy

TOC

Description

Top 3 accuracy. Expected input shape: (batch_size, num_classes)

Parameters

Name Type Description
Y torch.Tensor Prediction
T torch.Tensor True values

Returns

Top 3 accuracy -> top k accuracy | k = 3 : float

Example

import rsp.ml.metrics as m

Y = torch.tensor([
  [0.1, 0.1, 0.8],
  [0.03, 0.95, 0.02],
  [0.05, 0.9, 0.05],
  [0.01, 0.87, 0.12],
  [0.04, 0.03, 0.93],
  [0.94, 0.02, 0.06]
])
T = torch.tensor([
  [0, 0, 1],
  [1, 0, 0],
  [0, 1, 0],
  [0, 1, 0],
  [0, 0, 1],
  [1, 0, 0]
])

top_3_accuracy = m.top_3_accuracy(Y, T, k = 3)

print(top_3_accuracy) --> 1.0

2.19 top_5_accuracy

TOC

Description

Top 5 accuracy. Expected input shape: (batch_size, num_classes)

Parameters

Name Type Description
Y torch.Tensor Prediction
T torch.Tensor True values

Returns

Top 5 accuracy -> top k accuracy | k = 5 : float

Example

import rsp.ml.metrics as m

Y = torch.tensor([
  [0.1, 0.1, 0.8],
  [0.03, 0.95, 0.02],
  [0.05, 0.9, 0.05],
  [0.01, 0.87, 0.12],
  [0.04, 0.03, 0.93],
  [0.94, 0.02, 0.06]
])
T = torch.tensor([
  [0, 0, 1],
  [1, 0, 0],
  [0, 1, 0],
  [0, 1, 0],
  [0, 0, 1],
  [1, 0, 0]
])

top_5_accuracy = m.top_5_accuracy(Y, T, k = 3)

print(top_5_accuracy) --> 1.0

2.20 top_k_accuracy

TOC

Description

Top k accuracy. Expected input shape: (batch_size, num_classes)

Parameters

Name Type Description
Y torch.Tensor Prediction
T torch.Tensor True values

Returns

Top k accuracy : float

Example

import rsp.ml.metrics as m

Y = torch.tensor([
  [0.1, 0.1, 0.8],
  [0.03, 0.95, 0.02],
  [0.05, 0.9, 0.05],
  [0.01, 0.87, 0.12],
  [0.04, 0.03, 0.93],
  [0.94, 0.02, 0.06]
])
T = torch.tensor([
  [0, 0, 1],
  [1, 0, 0],
  [0, 1, 0],
  [0, 1, 0],
  [0, 0, 1],
  [1, 0, 0]
])

top_k_accuracy = m.top_k_accuracy(Y, T, k = 3)

print(top_k_accuracy) --> 1.0

3 model

TOC

The module rsp.ml.model provides some usefull functionality to store and load pytorch models.

3.1 MODELS : enum.Enum

TOC

Description

Create a collection of name/value pairs.

Example enumeration:

class Color(Enum): ... RED = 1 ... BLUE = 2 ... GREEN = 3

Access them by:

  • attribute access::

Color.RED <Color.RED: 1>

  • value lookup:

Color(1) <Color.RED: 1>

  • name lookup:

Color['RED'] <Color.RED: 1>

Enumerations can be iterated over, and know how many members they have:

len(Color) 3

list(Color) [<Color.RED: 1>, <Color.BLUE: 2>, <Color.GREEN: 3>]

Methods can be added to enumerations, and members can have their own attributes -- see the documentation for details.

3.2 WEIGHTS : enum.Enum

TOC

Description

Create a collection of name/value pairs.

Example enumeration:

class Color(Enum): ... RED = 1 ... BLUE = 2 ... GREEN = 3

Access them by:

  • attribute access::

Color.RED <Color.RED: 1>

  • value lookup:

Color(1) <Color.RED: 1>

  • name lookup:

Color['RED'] <Color.RED: 1>

Enumerations can be iterated over, and know how many members they have:

len(Color) 3

list(Color) [<Color.RED: 1>, <Color.BLUE: 2>, <Color.GREEN: 3>]

Methods can be added to enumerations, and members can have their own attributes -- see the documentation for details.

3.3 list_model_weights

TOC

Description

Lists all available weight files.

Returns

List of (MODEL:str, WEIGHT:str) : List[Tuple(str, str)]

Example

import rsp.ml.model as model

model_weight_files = model.list_model_weights()

3.4 load_model

TOC

Description

Loads a pretrained PyTorch model from HuggingFace.

Parameters

Name Type Description
model MODELS ID of the model
weights WEIGHTS ID of the weights

Returns

Pretrained PyTorch model : torch.nn.Module

Example

import rsp.ml.model as model

action_recognition_model = model.load_model(MODEL.TUCARC3D, WEIGHTS.TUCAR)

3.5 publish_model

TOC

4 module

TOC

4.1 MultiHeadSelfAttention : torch.nn.modules.module.Module

TOC

Description

Base class for all neural network modules.

Your models should also subclass this class.

Modules can also contain other Modules, allowing them to be nested in a tree structure. You can assign the submodules as regular attributes::

import torch.nn as nn
import torch.nn.functional as F

class Model(nn.Module):
    def __init__(self) -> None:
        super().__init__()
        self.conv1 = nn.Conv2d(1, 20, 5)
        self.conv2 = nn.Conv2d(20, 20, 5)

    def forward(self, x):
        x = F.relu(self.conv1(x))
        return F.relu(self.conv2(x))

Submodules assigned in this way will be registered, and will also have their parameters converted when you call :meth:to, etc.

.. note:: As per the example above, an __init__() call to the parent class must be made before assignment on the child.

:ivar training: Boolean represents whether this module is in training or evaluation mode. :vartype training: bool

4.1.1 _wrapped_call_impl

TOC

4.1.2 __init__

TOC

Description

Initialize internal Module state, shared by both nn.Module and ScriptModule.

4.1.3 _apply

TOC

4.1.4 _call_impl

TOC

4.1.5 _get_backward_hooks

TOC

Description

Return the backward hooks for use in the call function.

It returns two lists, one with the full backward hooks and one with the non-full

backward hooks.

4.1.6 _get_backward_pre_hooks

TOC

4.1.7 _get_name

TOC

4.1.8 _load_from_state_dict

TOC

Description

Copy parameters and buffers from :attr:state_dict into only this module, but not its descendants.

This is called on every submodule

in :meth:~torch.nn.Module.load_state_dict. Metadata saved for this

module in input :attr:state_dict is provided as :attr:local_metadata.

For state dicts without metadata, :attr:local_metadata is empty.

Subclasses can achieve class-specific backward compatible loading using

the version number at local_metadata.get("version", None).

Additionally, :attr:local_metadata can also contain the key

assign_to_params_buffers that indicates whether keys should be

assigned their corresponding tensor in the state_dict.

.. note::

:attr:`state_dict` is not the same object as the input

:attr:`state_dict` to :meth:`~torch.nn.Module.load_state_dict`. So

it can be modified.

Args:

state_dict (dict): a dict containing parameters and

    persistent buffers.

prefix (str): the prefix for parameters and buffers used in this

    module

local_metadata (dict): a dict containing the metadata for this module.

    See

strict (bool): whether to strictly enforce that the keys in

    :attr:`state_dict` with :attr:`prefix` match the names of

    parameters and buffers in this module

missing_keys (list of str): if ``strict=True``, add missing keys to

    this list

unexpected_keys (list of str): if ``strict=True``, add unexpected

    keys to this list

error_msgs (list of str): error messages should be added to this

    list, and will be reported together in

    :meth:`~torch.nn.Module.load_state_dict`
4.1.9 _maybe_warn_non_full_backward_hook

TOC

4.1.10 _named_members

TOC

Description

Help yield various names + members of modules.

4.1.11 _register_load_state_dict_pre_hook

TOC

Description

See :meth:~torch.nn.Module.register_load_state_dict_pre_hook for details.

A subtle difference is that if with_module is set to False, then the

hook will not take the module as the first argument whereas

:meth:~torch.nn.Module.register_load_state_dict_pre_hook always takes the

module as the first argument.

Arguments:

hook (Callable): Callable hook that will be invoked before

    loading the state dict.

with_module (bool, optional): Whether or not to pass the module

    instance to the hook as the first parameter.
4.1.12 _register_state_dict_hook

TOC

Description

Register a post-hook for the :meth:~torch.nn.Module.state_dict method.

It should have the following signature::

hook(module, state_dict, prefix, local_metadata) -> None or state_dict

The registered hooks can modify the state_dict inplace or return a new one.

If a new state_dict is returned, it will only be respected if it is the root

module that :meth:~nn.Module.state_dict is called from.

4.1.13 _replicate_for_data_parallel

TOC

4.1.14 _save_to_state_dict

TOC

Description

Save module state to the destination dictionary.

The destination dictionary will contain the state

of the module, but not its descendants. This is called on every

submodule in :meth:~torch.nn.Module.state_dict.

In rare cases, subclasses can achieve class-specific behavior by

overriding this method with custom logic.

Args:

destination (dict): a dict where state will be stored

prefix (str): the prefix for parameters and buffers used in this

    module
4.1.15 _slow_forward

TOC

4.1.16 _wrapped_call_impl

TOC

4.1.17 add_module

TOC

Description

Add a child module to the current module.

The module can be accessed as an attribute using the given name.

Args:

name (str): name of the child module. The child module can be

    accessed from this module using the given name

module (Module): child module to be added to the module.
4.1.18 apply

TOC

Description

Apply fn recursively to every submodule (as returned by .children()) as well as self.

Typical use includes initializing the parameters of a model

(see also :ref:nn-init-doc).

Args:

fn (:class:`Module` -> None): function to be applied to each submodule

Returns:

Module: self

Example::

>>> @torch.no_grad()

>>> def init_weights(m):

>>>     print(m)

>>>     if type(m) == nn.Linear:

>>>         m.weight.fill_(1.0)

>>>         print(m.weight)

>>> net = nn.Sequential(nn.Linear(2, 2), nn.Linear(2, 2))

>>> net.apply(init_weights)

Linear(in_features=2, out_features=2, bias=True)

Parameter containing:

tensor([[1., 1.],

        [1., 1.]], requires_grad=True)

Linear(in_features=2, out_features=2, bias=True)

Parameter containing:

tensor([[1., 1.],

        [1., 1.]], requires_grad=True)

Sequential(

  (0): Linear(in_features=2, out_features=2, bias=True)

  (1): Linear(in_features=2, out_features=2, bias=True)

)
4.1.19 bfloat16

TOC

Description

Casts all floating point parameters and buffers to bfloat16 datatype.

.. note::

This method modifies the module in-place.

Returns:

Module: self
4.1.20 buffers

TOC

Description

Return an iterator over module buffers.

Args:

recurse (bool): if True, then yields buffers of this module

    and all submodules. Otherwise, yields only buffers that

    are direct members of this module.

Yields:

torch.Tensor: module buffer

Example::

>>> # xdoctest: +SKIP("undefined vars")

>>> for buf in model.buffers():

>>>     print(type(buf), buf.size())

<class 'torch.Tensor'> (20L,)

<class 'torch.Tensor'> (20L, 1L, 5L, 5L)
4.1.21 children

TOC

Description

Return an iterator over immediate children modules.

Yields:

Module: a child module
4.1.22 compile

TOC

Description

Compile this Module's forward using :func:torch.compile.

This Module's __call__ method is compiled and all arguments are passed as-is

to :func:torch.compile.

See :func:torch.compile for details on the arguments for this function.

4.1.23 cpu

TOC

Description

Move all model parameters and buffers to the CPU.

.. note::

This method modifies the module in-place.

Returns:

Module: self
4.1.24 cuda

TOC

Description

Move all model parameters and buffers to the GPU.

This also makes associated parameters and buffers different objects. So

it should be called before constructing the optimizer if the module will

live on GPU while being optimized.

.. note::

This method modifies the module in-place.

Args:

device (int, optional): if specified, all parameters will be

    copied to that device

Returns:

Module: self
4.1.25 double

TOC

Description

Casts all floating point parameters and buffers to double datatype.

.. note::

This method modifies the module in-place.

Returns:

Module: self
4.1.26 eval

TOC

Description

Set the module in evaluation mode.

This has an effect only on certain modules. See the documentation of

particular modules for details of their behaviors in training/evaluation

mode, i.e. whether they are affected, e.g. :class:Dropout, :class:BatchNorm,

etc.

This is equivalent with :meth:self.train(False) <torch.nn.Module.train>.

See :ref:locally-disable-grad-doc for a comparison between

.eval() and several similar mechanisms that may be confused with it.

Returns:

Module: self
4.1.27 extra_repr

TOC

Description

Return the extra representation of the module.

To print customized extra information, you should re-implement

this method in your own modules. Both single-line and multi-line

strings are acceptable.

4.1.28 float

TOC

Description

Casts all floating point parameters and buffers to float datatype.

.. note::

This method modifies the module in-place.

Returns:

Module: self

4.1.29 forward

TOC

Description

Define the computation performed at every call.

Should be overridden by all subclasses.

.. note::

Although the recipe for forward pass needs to be defined within

this function, one should call the :class:`Module` instance afterwards

instead of this since the former takes care of running the

registered hooks while the latter silently ignores them.
4.1.30 get_buffer

TOC

Description

Return the buffer given by target if it exists, otherwise throw an error.

See the docstring for get_submodule for a more detailed

explanation of this method's functionality as well as how to

correctly specify target.

Args:

target: The fully-qualified string name of the buffer

    to look for. (See ``get_submodule`` for how to specify a

    fully-qualified string.)

Returns:

torch.Tensor: The buffer referenced by ``target``

Raises:

AttributeError: If the target string references an invalid

    path or resolves to something that is not a

    buffer
4.1.31 get_extra_state

TOC

Description

Return any extra state to include in the module's state_dict.

Implement this and a corresponding :func:set_extra_state for your module

if you need to store extra state. This function is called when building the

module's state_dict().

Note that extra state should be picklable to ensure working serialization

of the state_dict. We only provide backwards compatibility guarantees

for serializing Tensors; other objects may break backwards compatibility if

their serialized pickled form changes.

Returns:

object: Any extra state to store in the module's state_dict
4.1.32 get_parameter

TOC

Description

Return the parameter given by target if it exists, otherwise throw an error.

See the docstring for get_submodule for a more detailed

explanation of this method's functionality as well as how to

correctly specify target.

Args:

target: The fully-qualified string name of the Parameter

    to look for. (See ``get_submodule`` for how to specify a

    fully-qualified string.)

Returns:

torch.nn.Parameter: The Parameter referenced by ``target``

Raises:

AttributeError: If the target string references an invalid

    path or resolves to something that is not an

    ``nn.Parameter``
4.1.33 get_submodule

TOC

Description

Return the submodule given by target if it exists, otherwise throw an error.

For example, let's say you have an nn.Module A that

looks like this:

.. code-block:: text

A(

    (net_b): Module(

        (net_c): Module(

            (conv): Conv2d(16, 33, kernel_size=(3, 3), stride=(2, 2))

        )

        (linear): Linear(in_features=100, out_features=200, bias=True)

    )

)

(The diagram shows an nn.Module A. A which has a nested

submodule net_b, which itself has two submodules net_c

and linear. net_c then has a submodule conv.)

To check whether or not we have the linear submodule, we

would call get_submodule("net_b.linear"). To check whether

we have the conv submodule, we would call

get_submodule("net_b.net_c.conv").

The runtime of get_submodule is bounded by the degree

of module nesting in target. A query against

named_modules achieves the same result, but it is O(N) in

the number of transitive modules. So, for a simple check to see

if some submodule exists, get_submodule should always be

used.

Args:

target: The fully-qualified string name of the submodule

    to look for. (See above example for how to specify a

    fully-qualified string.)

Returns:

torch.nn.Module: The submodule referenced by ``target``

Raises:

AttributeError: If the target string references an invalid

    path or resolves to something that is not an

    ``nn.Module``
4.1.34 half

TOC

Description

Casts all floating point parameters and buffers to half datatype.

.. note::

This method modifies the module in-place.

Returns:

Module: self
4.1.35 ipu

TOC

Description

Move all model parameters and buffers to the IPU.

This also makes associated parameters and buffers different objects. So

it should be called before constructing the optimizer if the module will

live on IPU while being optimized.

.. note::

This method modifies the module in-place.

Arguments:

device (int, optional): if specified, all parameters will be

    copied to that device

Returns:

Module: self
4.1.36 load_state_dict

TOC

Description

Copy parameters and buffers from :attr:state_dict into this module and its descendants.

If :attr:strict is True, then

the keys of :attr:state_dict must exactly match the keys returned

by this module's :meth:~torch.nn.Module.state_dict function.

.. warning::

If :attr:`assign` is ``True`` the optimizer must be created after

the call to :attr:`load_state_dict` unless

:func:`~torch.__future__.get_swap_module_params_on_conversion` is ``True``.

Args:

state_dict (dict): a dict containing parameters and

    persistent buffers.

strict (bool, optional): whether to strictly enforce that the keys

    in :attr:`state_dict` match the keys returned by this module's

    :meth:`~torch.nn.Module.state_dict` function. Default: ``True``

assign (bool, optional): When set to ``False``, the properties of the tensors

    in the current module are preserved whereas setting it to ``True`` preserves

    properties of the Tensors in the state dict. The only

    exception is the ``requires_grad`` field of :class:`~torch.nn.Parameter`s

    for which the value from the module is preserved.

    Default: ``False``

Returns:

``NamedTuple`` with ``missing_keys`` and ``unexpected_keys`` fields:

    * **missing_keys** is a list of str containing any keys that are expected

        by this module but missing from the provided ``state_dict``.

    * **unexpected_keys** is a list of str containing the keys that are not

        expected by this module but present in the provided ``state_dict``.

Note:

If a parameter or buffer is registered as ``None`` and its corresponding key

exists in :attr:`state_dict`, :meth:`load_state_dict` will raise a

``RuntimeError``.
4.1.37 modules

TOC

Description

Return an iterator over all modules in the network.

Yields:

Module: a module in the network

Note:

Duplicate modules are returned only once. In the following

example, ``l`` will be returned only once.

Example::

>>> l = nn.Linear(2, 2)

>>> net = nn.Sequential(l, l)

>>> for idx, m in enumerate(net.modules()):

...     print(idx, '->', m)

0 -> Sequential(

  (0): Linear(in_features=2, out_features=2, bias=True)

  (1): Linear(in_features=2, out_features=2, bias=True)

)

1 -> Linear(in_features=2, out_features=2, bias=True)
4.1.38 mtia

TOC

Description

Move all model parameters and buffers to the MTIA.

This also makes associated parameters and buffers different objects. So

it should be called before constructing the optimizer if the module will

live on MTIA while being optimized.

.. note::

This method modifies the module in-place.

Arguments:

device (int, optional): if specified, all parameters will be

    copied to that device

Returns:

Module: self
4.1.39 named_buffers

TOC

Description

Return an iterator over module buffers, yielding both the name of the buffer as well as the buffer itself.

Args:

prefix (str): prefix to prepend to all buffer names.

recurse (bool, optional): if True, then yields buffers of this module

    and all submodules. Otherwise, yields only buffers that

    are direct members of this module. Defaults to True.

remove_duplicate (bool, optional): whether to remove the duplicated buffers in the result. Defaults to True.

Yields:

(str, torch.Tensor): Tuple containing the name and buffer

Example::

>>> # xdoctest: +SKIP("undefined vars")

>>> for name, buf in self.named_buffers():

>>>     if name in ['running_var']:

>>>         print(buf.size())
4.1.40 named_children

TOC

Description

Return an iterator over immediate children modules, yielding both the name of the module as well as the module itself.

Yields:

(str, Module): Tuple containing a name and child module

Example::

>>> # xdoctest: +SKIP("undefined vars")

>>> for name, module in model.named_children():

>>>     if name in ['conv4', 'conv5']:

>>>         print(module)
4.1.41 named_modules

TOC

Description

Return an iterator over all modules in the network, yielding both the name of the module as well as the module itself.

Args:

memo: a memo to store the set of modules already added to the result

prefix: a prefix that will be added to the name of the module

remove_duplicate: whether to remove the duplicated module instances in the result

    or not

Yields:

(str, Module): Tuple of name and module

Note:

Duplicate modules are returned only once. In the following

example, ``l`` will be returned only once.

Example::

>>> l = nn.Linear(2, 2)

>>> net = nn.Sequential(l, l)

>>> for idx, m in enumerate(net.named_modules()):

...     print(idx, '->', m)

0 -> ('', Sequential(

  (0): Linear(in_features=2, out_features=2, bias=True)

  (1): Linear(in_features=2, out_features=2, bias=True)

))

1 -> ('0', Linear(in_features=2, out_features=2, bias=True))
4.1.42 named_parameters

TOC

Description

Return an iterator over module parameters, yielding both the name of the parameter as well as the parameter itself.

Args:

prefix (str): prefix to prepend to all parameter names.

recurse (bool): if True, then yields parameters of this module

    and all submodules. Otherwise, yields only parameters that

    are direct members of this module.

remove_duplicate (bool, optional): whether to remove the duplicated

    parameters in the result. Defaults to True.

Yields:

(str, Parameter): Tuple containing the name and parameter

Example::

>>> # xdoctest: +SKIP("undefined vars")

>>> for name, param in self.named_parameters():

>>>     if name in ['bias']:

>>>         print(param.size())
4.1.43 parameters

TOC

Description

Return an iterator over module parameters.

This is typically passed to an optimizer.

Args:

recurse (bool): if True, then yields parameters of this module

    and all submodules. Otherwise, yields only parameters that

    are direct members of this module.

Yields:

Parameter: module parameter

Example::

>>> # xdoctest: +SKIP("undefined vars")

>>> for param in model.parameters():

>>>     print(type(param), param.size())

<class 'torch.Tensor'> (20L,)

<class 'torch.Tensor'> (20L, 1L, 5L, 5L)
4.1.44 register_backward_hook

TOC

Description

Register a backward hook on the module.

This function is deprecated in favor of :meth:~torch.nn.Module.register_full_backward_hook and

the behavior of this function will change in future versions.

Returns:

:class:`torch.utils.hooks.RemovableHandle`:

    a handle that can be used to remove the added hook by calling

    ``handle.remove()``
4.1.45 register_buffer

TOC

Description

Add a buffer to the module.

This is typically used to register a buffer that should not to be

considered a model parameter. For example, BatchNorm's running_mean

is not a parameter, but is part of the module's state. Buffers, by

default, are persistent and will be saved alongside parameters. This

behavior can be changed by setting :attr:persistent to False. The

only difference between a persistent buffer and a non-persistent buffer

is that the latter will not be a part of this module's

:attr:state_dict.

Buffers can be accessed as attributes using given names.

Args:

name (str): name of the buffer. The buffer can be accessed

    from this module using the given name

tensor (Tensor or None): buffer to be registered. If ``None``, then operations

    that run on buffers, such as :attr:`cuda`, are ignored. If ``None``,

    the buffer is **not** included in the module's :attr:`state_dict`.

persistent (bool): whether the buffer is part of this module's

    :attr:`state_dict`.

Example::

>>> # xdoctest: +SKIP("undefined vars")

>>> self.register_buffer('running_mean', torch.zeros(num_features))
4.1.46 register_forward_hook

TOC

Description

Register a forward hook on the module.

The hook will be called every time after :func:forward has computed an output.

If with_kwargs is False or not specified, the input contains only

the positional arguments given to the module. Keyword arguments won't be

passed to the hooks and only to the forward. The hook can modify the

output. It can modify the input inplace but it will not have effect on

forward since this is called after :func:forward is called. The hook

should have the following signature::

hook(module, args, output) -> None or modified output

If with_kwargs is True, the forward hook will be passed the

kwargs given to the forward function and be expected to return the

output possibly modified. The hook should have the following signature::

hook(module, args, kwargs, output) -> None or modified output

Args:

hook (Callable): The user defined hook to be registered.

prepend (bool): If ``True``, the provided ``hook`` will be fired

    before all existing ``forward`` hooks on this

    :class:`torch.nn.modules.Module`. Otherwise, the provided

    ``hook`` will be fired after all existing ``forward`` hooks on

    this :class:`torch.nn.modules.Module`. Note that global

    ``forward`` hooks registered with

    :func:`register_module_forward_hook` will fire before all hooks

    registered by this method.

    Default: ``False``

with_kwargs (bool): If ``True``, the ``hook`` will be passed the

    kwargs given to the forward function.

    Default: ``False``

always_call (bool): If ``True`` the ``hook`` will be run regardless of

    whether an exception is raised while calling the Module.

    Default: ``False``

Returns:

:class:`torch.utils.hooks.RemovableHandle`:

    a handle that can be used to remove the added hook by calling

    ``handle.remove()``
4.1.47 register_forward_pre_hook

TOC

Description

Register a forward pre-hook on the module.

The hook will be called every time before :func:forward is invoked.

If with_kwargs is false or not specified, the input contains only

the positional arguments given to the module. Keyword arguments won't be

passed to the hooks and only to the forward. The hook can modify the

input. User can either return a tuple or a single modified value in the

hook. We will wrap the value into a tuple if a single value is returned

(unless that value is already a tuple). The hook should have the

following signature::

hook(module, args) -> None or modified input

If with_kwargs is true, the forward pre-hook will be passed the

kwargs given to the forward function. And if the hook modifies the

input, both the args and kwargs should be returned. The hook should have

the following signature::

hook(module, args, kwargs) -> None or a tuple of modified input and kwargs

Args:

hook (Callable): The user defined hook to be registered.

prepend (bool): If true, the provided ``hook`` will be fired before

    all existing ``forward_pre`` hooks on this

    :class:`torch.nn.modules.Module`. Otherwise, the provided

    ``hook`` will be fired after all existing ``forward_pre`` hooks

    on this :class:`torch.nn.modules.Module`. Note that global

    ``forward_pre`` hooks registered with

    :func:`register_module_forward_pre_hook` will fire before all

    hooks registered by this method.

    Default: ``False``

with_kwargs (bool): If true, the ``hook`` will be passed the kwargs

    given to the forward function.

    Default: ``False``

Returns:

:class:`torch.utils.hooks.RemovableHandle`:

    a handle that can be used to remove the added hook by calling

    ``handle.remove()``
4.1.48 register_full_backward_hook

TOC

Description

Register a backward hook on the module.

The hook will be called every time the gradients with respect to a module

are computed, i.e. the hook will execute if and only if the gradients with

respect to module outputs are computed. The hook should have the following

signature::

hook(module, grad_input, grad_output) -> tuple(Tensor) or None

The :attr:grad_input and :attr:grad_output are tuples that contain the gradients

with respect to the inputs and outputs respectively. The hook should

not modify its arguments, but it can optionally return a new gradient with

respect to the input that will be used in place of :attr:grad_input in

subsequent computations. :attr:grad_input will only correspond to the inputs given

as positional arguments and all kwarg arguments are ignored. Entries

in :attr:grad_input and :attr:grad_output will be None for all non-Tensor

arguments.

For technical reasons, when this hook is applied to a Module, its forward function will

receive a view of each Tensor passed to the Module. Similarly the caller will receive a view

of each Tensor returned by the Module's forward function.

.. warning ::

Modifying inputs or outputs inplace is not allowed when using backward hooks and

will raise an error.

Args:

hook (Callable): The user-defined hook to be registered.

prepend (bool): If true, the provided ``hook`` will be fired before

    all existing ``backward`` hooks on this

    :class:`torch.nn.modules.Module`. Otherwise, the provided

    ``hook`` will be fired after all existing ``backward`` hooks on

    this :class:`torch.nn.modules.Module`. Note that global

    ``backward`` hooks registered with

    :func:`register_module_full_backward_hook` will fire before

    all hooks registered by this method.

Returns:

:class:`torch.utils.hooks.RemovableHandle`:

    a handle that can be used to remove the added hook by calling

    ``handle.remove()``
4.1.49 register_full_backward_pre_hook

TOC

Description

Register a backward pre-hook on the module.

The hook will be called every time the gradients for the module are computed.

The hook should have the following signature::

hook(module, grad_output) -> tuple[Tensor] or None

The :attr:grad_output is a tuple. The hook should

not modify its arguments, but it can optionally return a new gradient with

respect to the output that will be used in place of :attr:grad_output in

subsequent computations. Entries in :attr:grad_output will be None for

all non-Tensor arguments.

For technical reasons, when this hook is applied to a Module, its forward function will

receive a view of each Tensor passed to the Module. Similarly the caller will receive a view

of each Tensor returned by the Module's forward function.

.. warning ::

Modifying inputs inplace is not allowed when using backward hooks and

will raise an error.

Args:

hook (Callable): The user-defined hook to be registered.

prepend (bool): If true, the provided ``hook`` will be fired before

    all existing ``backward_pre`` hooks on this

    :class:`torch.nn.modules.Module`. Otherwise, the provided

    ``hook`` will be fired after all existing ``backward_pre`` hooks

    on this :class:`torch.nn.modules.Module`. Note that global

    ``backward_pre`` hooks registered with

    :func:`register_module_full_backward_pre_hook` will fire before

    all hooks registered by this method.

Returns:

:class:`torch.utils.hooks.RemovableHandle`:

    a handle that can be used to remove the added hook by calling

    ``handle.remove()``
4.1.50 register_load_state_dict_post_hook

TOC

Description

Register a post-hook to be run after module's :meth:~nn.Module.load_state_dict is called.

It should have the following signature::

hook(module, incompatible_keys) -> None

The module argument is the current module that this hook is registered

on, and the incompatible_keys argument is a NamedTuple consisting

of attributes missing_keys and unexpected_keys. missing_keys

is a list of str containing the missing keys and

unexpected_keys is a list of str containing the unexpected keys.

The given incompatible_keys can be modified inplace if needed.

Note that the checks performed when calling :func:load_state_dict with

strict=True are affected by modifications the hook makes to

missing_keys or unexpected_keys, as expected. Additions to either

set of keys will result in an error being thrown when strict=True, and

clearing out both missing and unexpected keys will avoid an error.

Returns:

:class:`torch.utils.hooks.RemovableHandle`:

    a handle that can be used to remove the added hook by calling

    ``handle.remove()``
4.1.51 register_load_state_dict_pre_hook

TOC

Description

Register a pre-hook to be run before module's :meth:~nn.Module.load_state_dict is called.

It should have the following signature::

hook(module, state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs) -> None  # noqa: B950

Arguments:

hook (Callable): Callable hook that will be invoked before

    loading the state dict.
4.1.52 register_module

TOC

Description

Alias for :func:add_module.

4.1.53 register_parameter

TOC

Description

Add a parameter to the module.

The parameter can be accessed as an attribute using given name.

Args:

name (str): name of the parameter. The parameter can be accessed

    from this module using the given name

param (Parameter or None): parameter to be added to the module. If

    ``None``, then operations that run on parameters, such as :attr:`cuda`,

    are ignored. If ``None``, the parameter is **not** included in the

    module's :attr:`state_dict`.
4.1.54 register_state_dict_post_hook

TOC

Description

Register a post-hook for the :meth:~torch.nn.Module.state_dict method.

It should have the following signature::

hook(module, state_dict, prefix, local_metadata) -> None

The registered hooks can modify the state_dict inplace.

4.1.55 register_state_dict_pre_hook

TOC

Description

Register a pre-hook for the :meth:~torch.nn.Module.state_dict method.

It should have the following signature::

hook(module, prefix, keep_vars) -> None

The registered hooks can be used to perform pre-processing before the state_dict

call is made.

4.1.56 requires_grad_

TOC

Description

Change if autograd should record operations on parameters in this module.

This method sets the parameters' :attr:requires_grad attributes

in-place.

This method is helpful for freezing part of the module for finetuning

or training parts of a model individually (e.g., GAN training).

See :ref:locally-disable-grad-doc for a comparison between

.requires_grad_() and several similar mechanisms that may be confused with it.

Args:

requires_grad (bool): whether autograd should record operations on

                      parameters in this module. Default: ``True``.

Returns:

Module: self
4.1.57 set_extra_state

TOC

Description

Set extra state contained in the loaded state_dict.

This function is called from :func:load_state_dict to handle any extra state

found within the state_dict. Implement this function and a corresponding

:func:get_extra_state for your module if you need to store extra state within its

state_dict.

Args:

state (dict): Extra state from the `state_dict`
4.1.58 set_submodule

TOC

Description

Set the submodule given by target if it exists, otherwise throw an error.

For example, let's say you have an nn.Module A that

looks like this:

.. code-block:: text

A(

    (net_b): Module(

        (net_c): Module(

            (conv): Conv2d(16, 33, kernel_size=(3, 3), stride=(2, 2))

        )

        (linear): Linear(in_features=100, out_features=200, bias=True)

    )

)

(The diagram shows an nn.Module A. A has a nested

submodule net_b, which itself has two submodules net_c

and linear. net_c then has a submodule conv.)

To overide the Conv2d with a new submodule Linear, you

would call

set_submodule("net_b.net_c.conv", nn.Linear(33, 16)).

Args:

target: The fully-qualified string name of the submodule

    to look for. (See above example for how to specify a

    fully-qualified string.)

module: The module to set the submodule to.

Raises:

ValueError: If the target string is empty

AttributeError: If the target string references an invalid

    path or resolves to something that is not an

    ``nn.Module``
4.1.59 share_memory

TOC

Description

See :meth:torch.Tensor.share_memory_.

4.1.60 state_dict

TOC

Description

Return a dictionary containing references to the whole state of the module.

Both parameters and persistent buffers (e.g. running averages) are

included. Keys are corresponding parameter and buffer names.

Parameters and buffers set to None are not included.

.. note::

The returned object is a shallow copy. It contains references

to the module's parameters and buffers.

.. warning::

Currently ``state_dict()`` also accepts positional arguments for

``destination``, ``prefix`` and ``keep_vars`` in order. However,

this is being deprecated and keyword arguments will be enforced in

future releases.

.. warning::

Please avoid the use of argument ``destination`` as it is not

designed for end-users.

Args:

destination (dict, optional): If provided, the state of module will

    be updated into the dict and the same object is returned.

    Otherwise, an ``OrderedDict`` will be created and returned.

    Default: ``None``.

prefix (str, optional): a prefix added to parameter and buffer

    names to compose the keys in state_dict. Default: ``''``.

keep_vars (bool, optional): by default the :class:`~torch.Tensor` s

    returned in the state dict are detached from autograd. If it's

    set to ``True``, detaching will not be performed.

    Default: ``False``.

Returns:

dict:

    a dictionary containing a whole state of the module

Example::

>>> # xdoctest: +SKIP("undefined vars")

>>> module.state_dict().keys()

['bias', 'weight']
4.1.61 to

TOC

Description

Move and/or cast the parameters and buffers.

This can be called as

.. function:: to(device=None, dtype=None, non_blocking=False)

:noindex:

.. function:: to(dtype, non_blocking=False)

:noindex:

.. function:: to(tensor, non_blocking=False)

:noindex:

.. function:: to(memory_format=torch.channels_last)

:noindex:

Its signature is similar to :meth:torch.Tensor.to, but only accepts

floating point or complex :attr:dtype\ s. In addition, this method will

only cast the floating point or complex parameters and buffers to :attr:dtype

(if given). The integral parameters and buffers will be moved

:attr:device, if that is given, but with dtypes unchanged. When

:attr:non_blocking is set, it tries to convert/move asynchronously

with respect to the host if possible, e.g., moving CPU Tensors with

pinned memory to CUDA devices.

See below for examples.

.. note::

This method modifies the module in-place.

Args:

device (:class:`torch.device`): the desired device of the parameters

    and buffers in this module

dtype (:class:`torch.dtype`): the desired floating point or complex dtype of

    the parameters and buffers in this module

tensor (torch.Tensor): Tensor whose dtype and device are the desired

    dtype and device for all parameters and buffers in this module

memory_format (:class:`torch.memory_format`): the desired memory

    format for 4D parameters and buffers in this module (keyword

    only argument)

Returns:

Module: self

Examples::

>>> # xdoctest: +IGNORE_WANT("non-deterministic")

>>> linear = nn.Linear(2, 2)

>>> linear.weight

Parameter containing:

tensor([[ 0.1913, -0.3420],

        [-0.5113, -0.2325]])

>>> linear.to(torch.double)

Linear(in_features=2, out_features=2, bias=True)

>>> linear.weight

Parameter containing:

tensor([[ 0.1913, -0.3420],

        [-0.5113, -0.2325]], dtype=torch.float64)

>>> # xdoctest: +REQUIRES(env:TORCH_DOCTEST_CUDA1)

>>> gpu1 = torch.device("cuda:1")

>>> linear.to(gpu1, dtype=torch.half, non_blocking=True)

Linear(in_features=2, out_features=2, bias=True)

>>> linear.weight

Parameter containing:

tensor([[ 0.1914, -0.3420],

        [-0.5112, -0.2324]], dtype=torch.float16, device='cuda:1')

>>> cpu = torch.device("cpu")

>>> linear.to(cpu)

Linear(in_features=2, out_features=2, bias=True)

>>> linear.weight

Parameter containing:

tensor([[ 0.1914, -0.3420],

        [-0.5112, -0.2324]], dtype=torch.float16)

>>> linear = nn.Linear(2, 2, bias=None).to(torch.cdouble)

>>> linear.weight

Parameter containing:

tensor([[ 0.3741+0.j,  0.2382+0.j],

        [ 0.5593+0.j, -0.4443+0.j]], dtype=torch.complex128)

>>> linear(torch.ones(3, 2, dtype=torch.cdouble))

tensor([[0.6122+0.j, 0.1150+0.j],

        [0.6122+0.j, 0.1150+0.j],

        [0.6122+0.j, 0.1150+0.j]], dtype=torch.complex128)
4.1.62 to_empty

TOC

Description

Move the parameters and buffers to the specified device without copying storage.

Args:

device (:class:`torch.device`): The desired device of the parameters

    and buffers in this module.

recurse (bool): Whether parameters and buffers of submodules should

    be recursively moved to the specified device.

Returns:

Module: self
4.1.63 train

TOC

Description

Set the module in training mode.

This has an effect only on certain modules. See the documentation of

particular modules for details of their behaviors in training/evaluation

mode, i.e., whether they are affected, e.g. :class:Dropout, :class:BatchNorm,

etc.

Args:

mode (bool): whether to set training mode (``True``) or evaluation

             mode (``False``). Default: ``True``.

Returns:

Module: self
4.1.64 type

TOC

Description

Casts all parameters and buffers to :attr:dst_type.

.. note::

This method modifies the module in-place.

Args:

dst_type (type or string): the desired type

Returns:

Module: self
4.1.65 xpu

TOC

Description

Move all model parameters and buffers to the XPU.

This also makes associated parameters and buffers different objects. So

it should be called before constructing optimizer if the module will

live on XPU while being optimized.

.. note::

This method modifies the module in-place.

Arguments:

device (int, optional): if specified, all parameters will be

    copied to that device

Returns:

Module: self
4.1.66 zero_grad

TOC

Description

Reset gradients of all model parameters.

See similar function under :class:torch.optim.Optimizer for more context.

Args:

set_to_none (bool): instead of setting to zero, set the grads to None.

    See :meth:`torch.optim.Optimizer.zero_grad` for details.

4.2 SelfAttention : torch.nn.modules.module.Module

TOC

Description

Base class for all neural network modules.

Your models should also subclass this class.

Modules can also contain other Modules, allowing them to be nested in a tree structure. You can assign the submodules as regular attributes::

import torch.nn as nn
import torch.nn.functional as F

class Model(nn.Module):
    def __init__(self) -> None:
        super().__init__()
        self.conv1 = nn.Conv2d(1, 20, 5)
        self.conv2 = nn.Conv2d(20, 20, 5)

    def forward(self, x):
        x = F.relu(self.conv1(x))
        return F.relu(self.conv2(x))

Submodules assigned in this way will be registered, and will also have their parameters converted when you call :meth:to, etc.

.. note:: As per the example above, an __init__() call to the parent class must be made before assignment on the child.

:ivar training: Boolean represents whether this module is in training or evaluation mode. :vartype training: bool

4.2.1 _wrapped_call_impl

TOC

4.2.2 __init__

TOC

Description

Initialize internal Module state, shared by both nn.Module and ScriptModule.

4.2.3 _apply

TOC

4.2.4 _call_impl

TOC

4.2.5 _get_backward_hooks

TOC

Description

Return the backward hooks for use in the call function.

It returns two lists, one with the full backward hooks and one with the non-full

backward hooks.

4.2.6 _get_backward_pre_hooks

TOC

4.2.7 _get_name

TOC

4.2.8 _load_from_state_dict

TOC

Description

Copy parameters and buffers from :attr:state_dict into only this module, but not its descendants.

This is called on every submodule

in :meth:~torch.nn.Module.load_state_dict. Metadata saved for this

module in input :attr:state_dict is provided as :attr:local_metadata.

For state dicts without metadata, :attr:local_metadata is empty.

Subclasses can achieve class-specific backward compatible loading using

the version number at local_metadata.get("version", None).

Additionally, :attr:local_metadata can also contain the key

assign_to_params_buffers that indicates whether keys should be

assigned their corresponding tensor in the state_dict.

.. note::

:attr:`state_dict` is not the same object as the input

:attr:`state_dict` to :meth:`~torch.nn.Module.load_state_dict`. So

it can be modified.

Args:

state_dict (dict): a dict containing parameters and

    persistent buffers.

prefix (str): the prefix for parameters and buffers used in this

    module

local_metadata (dict): a dict containing the metadata for this module.

    See

strict (bool): whether to strictly enforce that the keys in

    :attr:`state_dict` with :attr:`prefix` match the names of

    parameters and buffers in this module

missing_keys (list of str): if ``strict=True``, add missing keys to

    this list

unexpected_keys (list of str): if ``strict=True``, add unexpected

    keys to this list

error_msgs (list of str): error messages should be added to this

    list, and will be reported together in

    :meth:`~torch.nn.Module.load_state_dict`
4.2.9 _maybe_warn_non_full_backward_hook

TOC

4.2.10 _named_members

TOC

Description

Help yield various names + members of modules.

4.2.11 _register_load_state_dict_pre_hook

TOC

Description

See :meth:~torch.nn.Module.register_load_state_dict_pre_hook for details.

A subtle difference is that if with_module is set to False, then the

hook will not take the module as the first argument whereas

:meth:~torch.nn.Module.register_load_state_dict_pre_hook always takes the

module as the first argument.

Arguments:

hook (Callable): Callable hook that will be invoked before

    loading the state dict.

with_module (bool, optional): Whether or not to pass the module

    instance to the hook as the first parameter.
4.2.12 _register_state_dict_hook

TOC

Description

Register a post-hook for the :meth:~torch.nn.Module.state_dict method.

It should have the following signature::

hook(module, state_dict, prefix, local_metadata) -> None or state_dict

The registered hooks can modify the state_dict inplace or return a new one.

If a new state_dict is returned, it will only be respected if it is the root

module that :meth:~nn.Module.state_dict is called from.

4.2.13 _replicate_for_data_parallel

TOC

4.2.14 _save_to_state_dict

TOC

Description

Save module state to the destination dictionary.

The destination dictionary will contain the state

of the module, but not its descendants. This is called on every

submodule in :meth:~torch.nn.Module.state_dict.

In rare cases, subclasses can achieve class-specific behavior by

overriding this method with custom logic.

Args:

destination (dict): a dict where state will be stored

prefix (str): the prefix for parameters and buffers used in this

    module
4.2.15 _slow_forward

TOC

4.2.16 _wrapped_call_impl

TOC

4.2.17 add_module

TOC

Description

Add a child module to the current module.

The module can be accessed as an attribute using the given name.

Args:

name (str): name of the child module. The child module can be

    accessed from this module using the given name

module (Module): child module to be added to the module.
4.2.18 apply

TOC

Description

Apply fn recursively to every submodule (as returned by .children()) as well as self.

Typical use includes initializing the parameters of a model

(see also :ref:nn-init-doc).

Args:

fn (:class:`Module` -> None): function to be applied to each submodule

Returns:

Module: self

Example::

>>> @torch.no_grad()

>>> def init_weights(m):

>>>     print(m)

>>>     if type(m) == nn.Linear:

>>>         m.weight.fill_(1.0)

>>>         print(m.weight)

>>> net = nn.Sequential(nn.Linear(2, 2), nn.Linear(2, 2))

>>> net.apply(init_weights)

Linear(in_features=2, out_features=2, bias=True)

Parameter containing:

tensor([[1., 1.],

        [1., 1.]], requires_grad=True)

Linear(in_features=2, out_features=2, bias=True)

Parameter containing:

tensor([[1., 1.],

        [1., 1.]], requires_grad=True)

Sequential(

  (0): Linear(in_features=2, out_features=2, bias=True)

  (1): Linear(in_features=2, out_features=2, bias=True)

)
4.2.19 bfloat16

TOC

Description

Casts all floating point parameters and buffers to bfloat16 datatype.

.. note::

This method modifies the module in-place.

Returns:

Module: self
4.2.20 buffers

TOC

Description

Return an iterator over module buffers.

Args:

recurse (bool): if True, then yields buffers of this module

    and all submodules. Otherwise, yields only buffers that

    are direct members of this module.

Yields:

torch.Tensor: module buffer

Example::

>>> # xdoctest: +SKIP("undefined vars")

>>> for buf in model.buffers():

>>>     print(type(buf), buf.size())

<class 'torch.Tensor'> (20L,)

<class 'torch.Tensor'> (20L, 1L, 5L, 5L)
4.2.21 children

TOC

Description

Return an iterator over immediate children modules.

Yields:

Module: a child module
4.2.22 compile

TOC

Description

Compile this Module's forward using :func:torch.compile.

This Module's __call__ method is compiled and all arguments are passed as-is

to :func:torch.compile.

See :func:torch.compile for details on the arguments for this function.

4.2.23 cpu

TOC

Description

Move all model parameters and buffers to the CPU.

.. note::

This method modifies the module in-place.

Returns:

Module: self
4.2.24 cuda

TOC

Description

Move all model parameters and buffers to the GPU.

This also makes associated parameters and buffers different objects. So

it should be called before constructing the optimizer if the module will

live on GPU while being optimized.

.. note::

This method modifies the module in-place.

Args:

device (int, optional): if specified, all parameters will be

    copied to that device

Returns:

Module: self
4.2.25 double

TOC

Description

Casts all floating point parameters and buffers to double datatype.

.. note::

This method modifies the module in-place.

Returns:

Module: self
4.2.26 eval

TOC

Description

Set the module in evaluation mode.

This has an effect only on certain modules. See the documentation of

particular modules for details of their behaviors in training/evaluation

mode, i.e. whether they are affected, e.g. :class:Dropout, :class:BatchNorm,

etc.

This is equivalent with :meth:self.train(False) <torch.nn.Module.train>.

See :ref:locally-disable-grad-doc for a comparison between

.eval() and several similar mechanisms that may be confused with it.

Returns:

Module: self
4.2.27 extra_repr

TOC

Description

Return the extra representation of the module.

To print customized extra information, you should re-implement

this method in your own modules. Both single-line and multi-line

strings are acceptable.

4.2.28 float

TOC

Description

Casts all floating point parameters and buffers to float datatype.

.. note::

This method modifies the module in-place.

Returns:

Module: self

4.2.29 forward

TOC

Description

Define the computation performed at every call.

Should be overridden by all subclasses.

.. note::

Although the recipe for forward pass needs to be defined within

this function, one should call the :class:`Module` instance afterwards

instead of this since the former takes care of running the

registered hooks while the latter silently ignores them.
4.2.30 get_buffer

TOC

Description

Return the buffer given by target if it exists, otherwise throw an error.

See the docstring for get_submodule for a more detailed

explanation of this method's functionality as well as how to

correctly specify target.

Args:

target: The fully-qualified string name of the buffer

    to look for. (See ``get_submodule`` for how to specify a

    fully-qualified string.)

Returns:

torch.Tensor: The buffer referenced by ``target``

Raises:

AttributeError: If the target string references an invalid

    path or resolves to something that is not a

    buffer
4.2.31 get_extra_state

TOC

Description

Return any extra state to include in the module's state_dict.

Implement this and a corresponding :func:set_extra_state for your module

if you need to store extra state. This function is called when building the

module's state_dict().

Note that extra state should be picklable to ensure working serialization

of the state_dict. We only provide backwards compatibility guarantees

for serializing Tensors; other objects may break backwards compatibility if

their serialized pickled form changes.

Returns:

object: Any extra state to store in the module's state_dict
4.2.32 get_parameter

TOC

Description

Return the parameter given by target if it exists, otherwise throw an error.

See the docstring for get_submodule for a more detailed

explanation of this method's functionality as well as how to

correctly specify target.

Args:

target: The fully-qualified string name of the Parameter

    to look for. (See ``get_submodule`` for how to specify a

    fully-qualified string.)

Returns:

torch.nn.Parameter: The Parameter referenced by ``target``

Raises:

AttributeError: If the target string references an invalid

    path or resolves to something that is not an

    ``nn.Parameter``
4.2.33 get_submodule

TOC

Description

Return the submodule given by target if it exists, otherwise throw an error.

For example, let's say you have an nn.Module A that

looks like this:

.. code-block:: text

A(

    (net_b): Module(

        (net_c): Module(

            (conv): Conv2d(16, 33, kernel_size=(3, 3), stride=(2, 2))

        )

        (linear): Linear(in_features=100, out_features=200, bias=True)

    )

)

(The diagram shows an nn.Module A. A which has a nested

submodule net_b, which itself has two submodules net_c

and linear. net_c then has a submodule conv.)

To check whether or not we have the linear submodule, we

would call get_submodule("net_b.linear"). To check whether

we have the conv submodule, we would call

get_submodule("net_b.net_c.conv").

The runtime of get_submodule is bounded by the degree

of module nesting in target. A query against

named_modules achieves the same result, but it is O(N) in

the number of transitive modules. So, for a simple check to see

if some submodule exists, get_submodule should always be

used.

Args:

target: The fully-qualified string name of the submodule

    to look for. (See above example for how to specify a

    fully-qualified string.)

Returns:

torch.nn.Module: The submodule referenced by ``target``

Raises:

AttributeError: If the target string references an invalid

    path or resolves to something that is not an

    ``nn.Module``
4.2.34 half

TOC

Description

Casts all floating point parameters and buffers to half datatype.

.. note::

This method modifies the module in-place.

Returns:

Module: self
4.2.35 ipu

TOC

Description

Move all model parameters and buffers to the IPU.

This also makes associated parameters and buffers different objects. So

it should be called before constructing the optimizer if the module will

live on IPU while being optimized.

.. note::

This method modifies the module in-place.

Arguments:

device (int, optional): if specified, all parameters will be

    copied to that device

Returns:

Module: self
4.2.36 load_state_dict

TOC

Description

Copy parameters and buffers from :attr:state_dict into this module and its descendants.

If :attr:strict is True, then

the keys of :attr:state_dict must exactly match the keys returned

by this module's :meth:~torch.nn.Module.state_dict function.

.. warning::

If :attr:`assign` is ``True`` the optimizer must be created after

the call to :attr:`load_state_dict` unless

:func:`~torch.__future__.get_swap_module_params_on_conversion` is ``True``.

Args:

state_dict (dict): a dict containing parameters and

    persistent buffers.

strict (bool, optional): whether to strictly enforce that the keys

    in :attr:`state_dict` match the keys returned by this module's

    :meth:`~torch.nn.Module.state_dict` function. Default: ``True``

assign (bool, optional): When set to ``False``, the properties of the tensors

    in the current module are preserved whereas setting it to ``True`` preserves

    properties of the Tensors in the state dict. The only

    exception is the ``requires_grad`` field of :class:`~torch.nn.Parameter`s

    for which the value from the module is preserved.

    Default: ``False``

Returns:

``NamedTuple`` with ``missing_keys`` and ``unexpected_keys`` fields:

    * **missing_keys** is a list of str containing any keys that are expected

        by this module but missing from the provided ``state_dict``.

    * **unexpected_keys** is a list of str containing the keys that are not

        expected by this module but present in the provided ``state_dict``.

Note:

If a parameter or buffer is registered as ``None`` and its corresponding key

exists in :attr:`state_dict`, :meth:`load_state_dict` will raise a

``RuntimeError``.
4.2.37 modules

TOC

Description

Return an iterator over all modules in the network.

Yields:

Module: a module in the network

Note:

Duplicate modules are returned only once. In the following

example, ``l`` will be returned only once.

Example::

>>> l = nn.Linear(2, 2)

>>> net = nn.Sequential(l, l)

>>> for idx, m in enumerate(net.modules()):

...     print(idx, '->', m)

0 -> Sequential(

  (0): Linear(in_features=2, out_features=2, bias=True)

  (1): Linear(in_features=2, out_features=2, bias=True)

)

1 -> Linear(in_features=2, out_features=2, bias=True)
4.2.38 mtia

TOC

Description

Move all model parameters and buffers to the MTIA.

This also makes associated parameters and buffers different objects. So

it should be called before constructing the optimizer if the module will

live on MTIA while being optimized.

.. note::

This method modifies the module in-place.

Arguments:

device (int, optional): if specified, all parameters will be

    copied to that device

Returns:

Module: self
4.2.39 named_buffers

TOC

Description

Return an iterator over module buffers, yielding both the name of the buffer as well as the buffer itself.

Args:

prefix (str): prefix to prepend to all buffer names.

recurse (bool, optional): if True, then yields buffers of this module

    and all submodules. Otherwise, yields only buffers that

    are direct members of this module. Defaults to True.

remove_duplicate (bool, optional): whether to remove the duplicated buffers in the result. Defaults to True.

Yields:

(str, torch.Tensor): Tuple containing the name and buffer

Example::

>>> # xdoctest: +SKIP("undefined vars")

>>> for name, buf in self.named_buffers():

>>>     if name in ['running_var']:

>>>         print(buf.size())
4.2.40 named_children

TOC

Description

Return an iterator over immediate children modules, yielding both the name of the module as well as the module itself.

Yields:

(str, Module): Tuple containing a name and child module

Example::

>>> # xdoctest: +SKIP("undefined vars")

>>> for name, module in model.named_children():

>>>     if name in ['conv4', 'conv5']:

>>>         print(module)
4.2.41 named_modules

TOC

Description

Return an iterator over all modules in the network, yielding both the name of the module as well as the module itself.

Args:

memo: a memo to store the set of modules already added to the result

prefix: a prefix that will be added to the name of the module

remove_duplicate: whether to remove the duplicated module instances in the result

    or not

Yields:

(str, Module): Tuple of name and module

Note:

Duplicate modules are returned only once. In the following

example, ``l`` will be returned only once.

Example::

>>> l = nn.Linear(2, 2)

>>> net = nn.Sequential(l, l)

>>> for idx, m in enumerate(net.named_modules()):

...     print(idx, '->', m)

0 -> ('', Sequential(

  (0): Linear(in_features=2, out_features=2, bias=True)

  (1): Linear(in_features=2, out_features=2, bias=True)

))

1 -> ('0', Linear(in_features=2, out_features=2, bias=True))
4.2.42 named_parameters

TOC

Description

Return an iterator over module parameters, yielding both the name of the parameter as well as the parameter itself.

Args:

prefix (str): prefix to prepend to all parameter names.

recurse (bool): if True, then yields parameters of this module

    and all submodules. Otherwise, yields only parameters that

    are direct members of this module.

remove_duplicate (bool, optional): whether to remove the duplicated

    parameters in the result. Defaults to True.

Yields:

(str, Parameter): Tuple containing the name and parameter

Example::

>>> # xdoctest: +SKIP("undefined vars")

>>> for name, param in self.named_parameters():

>>>     if name in ['bias']:

>>>         print(param.size())
4.2.43 parameters

TOC

Description

Return an iterator over module parameters.

This is typically passed to an optimizer.

Args:

recurse (bool): if True, then yields parameters of this module

    and all submodules. Otherwise, yields only parameters that

    are direct members of this module.

Yields:

Parameter: module parameter

Example::

>>> # xdoctest: +SKIP("undefined vars")

>>> for param in model.parameters():

>>>     print(type(param), param.size())

<class 'torch.Tensor'> (20L,)

<class 'torch.Tensor'> (20L, 1L, 5L, 5L)
4.2.44 register_backward_hook

TOC

Description

Register a backward hook on the module.

This function is deprecated in favor of :meth:~torch.nn.Module.register_full_backward_hook and

the behavior of this function will change in future versions.

Returns:

:class:`torch.utils.hooks.RemovableHandle`:

    a handle that can be used to remove the added hook by calling

    ``handle.remove()``
4.2.45 register_buffer

TOC

Description

Add a buffer to the module.

This is typically used to register a buffer that should not to be

considered a model parameter. For example, BatchNorm's running_mean

is not a parameter, but is part of the module's state. Buffers, by

default, are persistent and will be saved alongside parameters. This

behavior can be changed by setting :attr:persistent to False. The

only difference between a persistent buffer and a non-persistent buffer

is that the latter will not be a part of this module's

:attr:state_dict.

Buffers can be accessed as attributes using given names.

Args:

name (str): name of the buffer. The buffer can be accessed

    from this module using the given name

tensor (Tensor or None): buffer to be registered. If ``None``, then operations

    that run on buffers, such as :attr:`cuda`, are ignored. If ``None``,

    the buffer is **not** included in the module's :attr:`state_dict`.

persistent (bool): whether the buffer is part of this module's

    :attr:`state_dict`.

Example::

>>> # xdoctest: +SKIP("undefined vars")

>>> self.register_buffer('running_mean', torch.zeros(num_features))
4.2.46 register_forward_hook

TOC

Description

Register a forward hook on the module.

The hook will be called every time after :func:forward has computed an output.

If with_kwargs is False or not specified, the input contains only

the positional arguments given to the module. Keyword arguments won't be

passed to the hooks and only to the forward. The hook can modify the

output. It can modify the input inplace but it will not have effect on

forward since this is called after :func:forward is called. The hook

should have the following signature::

hook(module, args, output) -> None or modified output

If with_kwargs is True, the forward hook will be passed the

kwargs given to the forward function and be expected to return the

output possibly modified. The hook should have the following signature::

hook(module, args, kwargs, output) -> None or modified output

Args:

hook (Callable): The user defined hook to be registered.

prepend (bool): If ``True``, the provided ``hook`` will be fired

    before all existing ``forward`` hooks on this

    :class:`torch.nn.modules.Module`. Otherwise, the provided

    ``hook`` will be fired after all existing ``forward`` hooks on

    this :class:`torch.nn.modules.Module`. Note that global

    ``forward`` hooks registered with

    :func:`register_module_forward_hook` will fire before all hooks

    registered by this method.

    Default: ``False``

with_kwargs (bool): If ``True``, the ``hook`` will be passed the

    kwargs given to the forward function.

    Default: ``False``

always_call (bool): If ``True`` the ``hook`` will be run regardless of

    whether an exception is raised while calling the Module.

    Default: ``False``

Returns:

:class:`torch.utils.hooks.RemovableHandle`:

    a handle that can be used to remove the added hook by calling

    ``handle.remove()``
4.2.47 register_forward_pre_hook

TOC

Description

Register a forward pre-hook on the module.

The hook will be called every time before :func:forward is invoked.

If with_kwargs is false or not specified, the input contains only

the positional arguments given to the module. Keyword arguments won't be

passed to the hooks and only to the forward. The hook can modify the

input. User can either return a tuple or a single modified value in the

hook. We will wrap the value into a tuple if a single value is returned

(unless that value is already a tuple). The hook should have the

following signature::

hook(module, args) -> None or modified input

If with_kwargs is true, the forward pre-hook will be passed the

kwargs given to the forward function. And if the hook modifies the

input, both the args and kwargs should be returned. The hook should have

the following signature::

hook(module, args, kwargs) -> None or a tuple of modified input and kwargs

Args:

hook (Callable): The user defined hook to be registered.

prepend (bool): If true, the provided ``hook`` will be fired before

    all existing ``forward_pre`` hooks on this

    :class:`torch.nn.modules.Module`. Otherwise, the provided

    ``hook`` will be fired after all existing ``forward_pre`` hooks

    on this :class:`torch.nn.modules.Module`. Note that global

    ``forward_pre`` hooks registered with

    :func:`register_module_forward_pre_hook` will fire before all

    hooks registered by this method.

    Default: ``False``

with_kwargs (bool): If true, the ``hook`` will be passed the kwargs

    given to the forward function.

    Default: ``False``

Returns:

:class:`torch.utils.hooks.RemovableHandle`:

    a handle that can be used to remove the added hook by calling

    ``handle.remove()``
4.2.48 register_full_backward_hook

TOC

Description

Register a backward hook on the module.

The hook will be called every time the gradients with respect to a module

are computed, i.e. the hook will execute if and only if the gradients with

respect to module outputs are computed. The hook should have the following

signature::

hook(module, grad_input, grad_output) -> tuple(Tensor) or None

The :attr:grad_input and :attr:grad_output are tuples that contain the gradients

with respect to the inputs and outputs respectively. The hook should

not modify its arguments, but it can optionally return a new gradient with

respect to the input that will be used in place of :attr:grad_input in

subsequent computations. :attr:grad_input will only correspond to the inputs given

as positional arguments and all kwarg arguments are ignored. Entries

in :attr:grad_input and :attr:grad_output will be None for all non-Tensor

arguments.

For technical reasons, when this hook is applied to a Module, its forward function will

receive a view of each Tensor passed to the Module. Similarly the caller will receive a view

of each Tensor returned by the Module's forward function.

.. warning ::

Modifying inputs or outputs inplace is not allowed when using backward hooks and

will raise an error.

Args:

hook (Callable): The user-defined hook to be registered.

prepend (bool): If true, the provided ``hook`` will be fired before

    all existing ``backward`` hooks on this

    :class:`torch.nn.modules.Module`. Otherwise, the provided

    ``hook`` will be fired after all existing ``backward`` hooks on

    this :class:`torch.nn.modules.Module`. Note that global

    ``backward`` hooks registered with

    :func:`register_module_full_backward_hook` will fire before

    all hooks registered by this method.

Returns:

:class:`torch.utils.hooks.RemovableHandle`:

    a handle that can be used to remove the added hook by calling

    ``handle.remove()``
4.2.49 register_full_backward_pre_hook

TOC

Description

Register a backward pre-hook on the module.

The hook will be called every time the gradients for the module are computed.

The hook should have the following signature::

hook(module, grad_output) -> tuple[Tensor] or None

The :attr:grad_output is a tuple. The hook should

not modify its arguments, but it can optionally return a new gradient with

respect to the output that will be used in place of :attr:grad_output in

subsequent computations. Entries in :attr:grad_output will be None for

all non-Tensor arguments.

For technical reasons, when this hook is applied to a Module, its forward function will

receive a view of each Tensor passed to the Module. Similarly the caller will receive a view

of each Tensor returned by the Module's forward function.

.. warning ::

Modifying inputs inplace is not allowed when using backward hooks and

will raise an error.

Args:

hook (Callable): The user-defined hook to be registered.

prepend (bool): If true, the provided ``hook`` will be fired before

    all existing ``backward_pre`` hooks on this

    :class:`torch.nn.modules.Module`. Otherwise, the provided

    ``hook`` will be fired after all existing ``backward_pre`` hooks

    on this :class:`torch.nn.modules.Module`. Note that global

    ``backward_pre`` hooks registered with

    :func:`register_module_full_backward_pre_hook` will fire before

    all hooks registered by this method.

Returns:

:class:`torch.utils.hooks.RemovableHandle`:

    a handle that can be used to remove the added hook by calling

    ``handle.remove()``
4.2.50 register_load_state_dict_post_hook

TOC

Description

Register a post-hook to be run after module's :meth:~nn.Module.load_state_dict is called.

It should have the following signature::

hook(module, incompatible_keys) -> None

The module argument is the current module that this hook is registered

on, and the incompatible_keys argument is a NamedTuple consisting

of attributes missing_keys and unexpected_keys. missing_keys

is a list of str containing the missing keys and

unexpected_keys is a list of str containing the unexpected keys.

The given incompatible_keys can be modified inplace if needed.

Note that the checks performed when calling :func:load_state_dict with

strict=True are affected by modifications the hook makes to

missing_keys or unexpected_keys, as expected. Additions to either

set of keys will result in an error being thrown when strict=True, and

clearing out both missing and unexpected keys will avoid an error.

Returns:

:class:`torch.utils.hooks.RemovableHandle`:

    a handle that can be used to remove the added hook by calling

    ``handle.remove()``
4.2.51 register_load_state_dict_pre_hook

TOC

Description

Register a pre-hook to be run before module's :meth:~nn.Module.load_state_dict is called.

It should have the following signature::

hook(module, state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs) -> None  # noqa: B950

Arguments:

hook (Callable): Callable hook that will be invoked before

    loading the state dict.
4.2.52 register_module

TOC

Description

Alias for :func:add_module.

4.2.53 register_parameter

TOC

Description

Add a parameter to the module.

The parameter can be accessed as an attribute using given name.

Args:

name (str): name of the parameter. The parameter can be accessed

    from this module using the given name

param (Parameter or None): parameter to be added to the module. If

    ``None``, then operations that run on parameters, such as :attr:`cuda`,

    are ignored. If ``None``, the parameter is **not** included in the

    module's :attr:`state_dict`.
4.2.54 register_state_dict_post_hook

TOC

Description

Register a post-hook for the :meth:~torch.nn.Module.state_dict method.

It should have the following signature::

hook(module, state_dict, prefix, local_metadata) -> None

The registered hooks can modify the state_dict inplace.

4.2.55 register_state_dict_pre_hook

TOC

Description

Register a pre-hook for the :meth:~torch.nn.Module.state_dict method.

It should have the following signature::

hook(module, prefix, keep_vars) -> None

The registered hooks can be used to perform pre-processing before the state_dict

call is made.

4.2.56 requires_grad_

TOC

Description

Change if autograd should record operations on parameters in this module.

This method sets the parameters' :attr:requires_grad attributes

in-place.

This method is helpful for freezing part of the module for finetuning

or training parts of a model individually (e.g., GAN training).

See :ref:locally-disable-grad-doc for a comparison between

.requires_grad_() and several similar mechanisms that may be confused with it.

Args:

requires_grad (bool): whether autograd should record operations on

                      parameters in this module. Default: ``True``.

Returns:

Module: self
4.2.57 set_extra_state

TOC

Description

Set extra state contained in the loaded state_dict.

This function is called from :func:load_state_dict to handle any extra state

found within the state_dict. Implement this function and a corresponding

:func:get_extra_state for your module if you need to store extra state within its

state_dict.

Args:

state (dict): Extra state from the `state_dict`
4.2.58 set_submodule

TOC

Description

Set the submodule given by target if it exists, otherwise throw an error.

For example, let's say you have an nn.Module A that

looks like this:

.. code-block:: text

A(

    (net_b): Module(

        (net_c): Module(

            (conv): Conv2d(16, 33, kernel_size=(3, 3), stride=(2, 2))

        )

        (linear): Linear(in_features=100, out_features=200, bias=True)

    )

)

(The diagram shows an nn.Module A. A has a nested

submodule net_b, which itself has two submodules net_c

and linear. net_c then has a submodule conv.)

To overide the Conv2d with a new submodule Linear, you

would call

set_submodule("net_b.net_c.conv", nn.Linear(33, 16)).

Args:

target: The fully-qualified string name of the submodule

    to look for. (See above example for how to specify a

    fully-qualified string.)

module: The module to set the submodule to.

Raises:

ValueError: If the target string is empty

AttributeError: If the target string references an invalid

    path or resolves to something that is not an

    ``nn.Module``
4.2.59 share_memory

TOC

Description

See :meth:torch.Tensor.share_memory_.

4.2.60 state_dict

TOC

Description

Return a dictionary containing references to the whole state of the module.

Both parameters and persistent buffers (e.g. running averages) are

included. Keys are corresponding parameter and buffer names.

Parameters and buffers set to None are not included.

.. note::

The returned object is a shallow copy. It contains references

to the module's parameters and buffers.

.. warning::

Currently ``state_dict()`` also accepts positional arguments for

``destination``, ``prefix`` and ``keep_vars`` in order. However,

this is being deprecated and keyword arguments will be enforced in

future releases.

.. warning::

Please avoid the use of argument ``destination`` as it is not

designed for end-users.

Args:

destination (dict, optional): If provided, the state of module will

    be updated into the dict and the same object is returned.

    Otherwise, an ``OrderedDict`` will be created and returned.

    Default: ``None``.

prefix (str, optional): a prefix added to parameter and buffer

    names to compose the keys in state_dict. Default: ``''``.

keep_vars (bool, optional): by default the :class:`~torch.Tensor` s

    returned in the state dict are detached from autograd. If it's

    set to ``True``, detaching will not be performed.

    Default: ``False``.

Returns:

dict:

    a dictionary containing a whole state of the module

Example::

>>> # xdoctest: +SKIP("undefined vars")

>>> module.state_dict().keys()

['bias', 'weight']
4.2.61 to

TOC

Description

Move and/or cast the parameters and buffers.

This can be called as

.. function:: to(device=None, dtype=None, non_blocking=False)

:noindex:

.. function:: to(dtype, non_blocking=False)

:noindex:

.. function:: to(tensor, non_blocking=False)

:noindex:

.. function:: to(memory_format=torch.channels_last)

:noindex:

Its signature is similar to :meth:torch.Tensor.to, but only accepts

floating point or complex :attr:dtype\ s. In addition, this method will

only cast the floating point or complex parameters and buffers to :attr:dtype

(if given). The integral parameters and buffers will be moved

:attr:device, if that is given, but with dtypes unchanged. When

:attr:non_blocking is set, it tries to convert/move asynchronously

with respect to the host if possible, e.g., moving CPU Tensors with

pinned memory to CUDA devices.

See below for examples.

.. note::

This method modifies the module in-place.

Args:

device (:class:`torch.device`): the desired device of the parameters

    and buffers in this module

dtype (:class:`torch.dtype`): the desired floating point or complex dtype of

    the parameters and buffers in this module

tensor (torch.Tensor): Tensor whose dtype and device are the desired

    dtype and device for all parameters and buffers in this module

memory_format (:class:`torch.memory_format`): the desired memory

    format for 4D parameters and buffers in this module (keyword

    only argument)

Returns:

Module: self

Examples::

>>> # xdoctest: +IGNORE_WANT("non-deterministic")

>>> linear = nn.Linear(2, 2)

>>> linear.weight

Parameter containing:

tensor([[ 0.1913, -0.3420],

        [-0.5113, -0.2325]])

>>> linear.to(torch.double)

Linear(in_features=2, out_features=2, bias=True)

>>> linear.weight

Parameter containing:

tensor([[ 0.1913, -0.3420],

        [-0.5113, -0.2325]], dtype=torch.float64)

>>> # xdoctest: +REQUIRES(env:TORCH_DOCTEST_CUDA1)

>>> gpu1 = torch.device("cuda:1")

>>> linear.to(gpu1, dtype=torch.half, non_blocking=True)

Linear(in_features=2, out_features=2, bias=True)

>>> linear.weight

Parameter containing:

tensor([[ 0.1914, -0.3420],

        [-0.5112, -0.2324]], dtype=torch.float16, device='cuda:1')

>>> cpu = torch.device("cpu")

>>> linear.to(cpu)

Linear(in_features=2, out_features=2, bias=True)

>>> linear.weight

Parameter containing:

tensor([[ 0.1914, -0.3420],

        [-0.5112, -0.2324]], dtype=torch.float16)

>>> linear = nn.Linear(2, 2, bias=None).to(torch.cdouble)

>>> linear.weight

Parameter containing:

tensor([[ 0.3741+0.j,  0.2382+0.j],

        [ 0.5593+0.j, -0.4443+0.j]], dtype=torch.complex128)

>>> linear(torch.ones(3, 2, dtype=torch.cdouble))

tensor([[0.6122+0.j, 0.1150+0.j],

        [0.6122+0.j, 0.1150+0.j],

        [0.6122+0.j, 0.1150+0.j]], dtype=torch.complex128)
4.2.62 to_empty

TOC

Description

Move the parameters and buffers to the specified device without copying storage.

Args:

device (:class:`torch.device`): The desired device of the parameters

    and buffers in this module.

recurse (bool): Whether parameters and buffers of submodules should

    be recursively moved to the specified device.

Returns:

Module: self
4.2.63 train

TOC

Description

Set the module in training mode.

This has an effect only on certain modules. See the documentation of

particular modules for details of their behaviors in training/evaluation

mode, i.e., whether they are affected, e.g. :class:Dropout, :class:BatchNorm,

etc.

Args:

mode (bool): whether to set training mode (``True``) or evaluation

             mode (``False``). Default: ``True``.

Returns:

Module: self
4.2.64 type

TOC

Description

Casts all parameters and buffers to :attr:dst_type.

.. note::

This method modifies the module in-place.

Args:

dst_type (type or string): the desired type

Returns:

Module: self
4.2.65 xpu

TOC

Description

Move all model parameters and buffers to the XPU.

This also makes associated parameters and buffers different objects. So

it should be called before constructing optimizer if the module will

live on XPU while being optimized.

.. note::

This method modifies the module in-place.

Arguments:

device (int, optional): if specified, all parameters will be

    copied to that device

Returns:

Module: self
4.2.66 zero_grad

TOC

Description

Reset gradients of all model parameters.

See similar function under :class:torch.optim.Optimizer for more context.

Args:

set_to_none (bool): instead of setting to zero, set the grads to None.

    See :meth:`torch.optim.Optimizer.zero_grad` for details.

5 multi_transforms

TOC

The module rsp.ml.multi_transforms is based on torchvision.transforms, which is made for single images. rsp.ml.multi_transforms extends this functionality by providing transformations for sequences of images, which could be usefull for video augmentation.

5.1 BGR2GRAY : MultiTransform

TOC

Description

Converts a sequence of BGR images to grayscale images.

5.1.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.1.2 __init__

TOC

Description

Initializes a new instance.

5.2 BGR2RGB : MultiTransform

TOC

Description

Converts sequence of BGR images to RGB images.

5.2.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.2.2 __init__

TOC

Description

Initializes a new instance.

5.3 Brightness : MultiTransform

TOC

Description

MultiTransform is an extension to keep the same transformation over a sequence of images instead of initializing a new transformation for every single image. It is inspired by torchvision.transforms and could be used for video augmentation. Use rsp.ml.multi_transforms.Composeto combine multiple image sequence transformations.

Note rsp.ml.multi_transforms.MultiTransform is a base class and should be inherited.

5.3.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.3.2 __init__

TOC

Description

Initializes a new instance.

5.4 CenterCrop : MultiTransform

TOC

Description

Crops Images at the center after upscaling them. Dimensions kept the same.

5.4.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.4.2 __init__

TOC

Description

Initializes a new instance.

Parameters

Name Type Description
max_scale float Images are scaled randomly between 1. and max_scale before cropping to original size.

5.5 Color : MultiTransform

TOC

Description

MultiTransform is an extension to keep the same transformation over a sequence of images instead of initializing a new transformation for every single image. It is inspired by torchvision.transforms and could be used for video augmentation. Use rsp.ml.multi_transforms.Composeto combine multiple image sequence transformations.

Note rsp.ml.multi_transforms.MultiTransform is a base class and should be inherited.

5.5.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.5.2 __init__

TOC

Description

Initializes a new instance.

5.6 Compose : builtins.object

TOC

Description

Composes several MultiTransforms together.

Example

import rsp.ml.multi_transforms as t

transforms = t.Compose([
  t.BGR2GRAY(),
  t.Scale(0.5)
])

5.6.1 __call__

TOC

Description

Call self as a function.

5.6.2 __init__

TOC

Description

Initializes a new instance.

Parameters

Name Type Description
children List[MultiTransform] List of MultiTransforms to compose.

5.7 GaussianNoise : MultiTransform

TOC

Description

MultiTransform is an extension to keep the same transformation over a sequence of images instead of initializing a new transformation for every single image. It is inspired by torchvision.transforms and could be used for video augmentation. Use rsp.ml.multi_transforms.Composeto combine multiple image sequence transformations.

Note rsp.ml.multi_transforms.MultiTransform is a base class and should be inherited.

5.7.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.7.2 __init__

TOC

Description

Initializes a new instance.

5.8 MultiTransform : builtins.object

TOC

Description

MultiTransform is an extension to keep the same transformation over a sequence of images instead of initializing a new transformation for every single image. It is inspired by torchvision.transforms and could be used for video augmentation. Use rsp.ml.multi_transforms.Composeto combine multiple image sequence transformations.

Note rsp.ml.multi_transforms.MultiTransform is a base class and should be inherited.

5.8.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.8.2 __init__

TOC

Description

Initializes a new instance.

5.9 Normalize : MultiTransform

TOC

Description

Normalize images with mean and standard deviation. Given mean: (mean[1],...,mean[n]) and std: (std[1],..,std[n]) for n channels, this transform will normalize each channel of the input torch.*Tensor i.e., output[channel] = (input[channel] - mean[channel]) / std[channel]

Based on torchvision.transforms.Normalize

5.9.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.9.2 __init__

TOC

Description

Initializes a new instance.

Parameters

Name Type Description
mean List[float] Sequence of means for each channel.
std List[float] Sequence of standard deviations for each channel.
inplace bool Set to True make this operation in-place.

5.10 RGB2BGR : BGR2RGB

TOC

Description

Converts sequence of RGB images to BGR images.

5.10.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.10.2 __init__

TOC

Description

Initializes a new instance.

5.11 RandomCrop : MultiTransform

TOC

Description

Crops Images at a random location after upscaling them. Dimensions kept the same.

5.11.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.11.2 __init__

TOC

Description

Initializes a new instance.

Parameters

Name Type Description
max_scale float Images are scaled randomly between 1. and max_scale before cropping to original size.

5.12 RandomHorizontalFlip : MultiTransform

TOC

Description

MultiTransform is an extension to keep the same transformation over a sequence of images instead of initializing a new transformation for every single image. It is inspired by torchvision.transforms and could be used for video augmentation. Use rsp.ml.multi_transforms.Composeto combine multiple image sequence transformations.

Note rsp.ml.multi_transforms.MultiTransform is a base class and should be inherited.

5.12.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.12.2 __init__

TOC

Description

Initializes a new instance.

5.13 RandomVerticalFlip : MultiTransform

TOC

Description

MultiTransform is an extension to keep the same transformation over a sequence of images instead of initializing a new transformation for every single image. It is inspired by torchvision.transforms and could be used for video augmentation. Use rsp.ml.multi_transforms.Composeto combine multiple image sequence transformations.

Note rsp.ml.multi_transforms.MultiTransform is a base class and should be inherited.

5.13.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.13.2 __init__

TOC

Description

Initializes a new instance.

5.14 RemoveBackgroundAI : MultiTransform

TOC

Description

MultiTransform is an extension to keep the same transformation over a sequence of images instead of initializing a new transformation for every single image. It is inspired by torchvision.transforms and could be used for video augmentation. Use rsp.ml.multi_transforms.Composeto combine multiple image sequence transformations.

Note rsp.ml.multi_transforms.MultiTransform is a base class and should be inherited.

5.14.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.14.2 __init__

TOC

Description

Initializes a new instance.

5.15 ReplaceBackground : MultiTransform

TOC

Description

Transformation for background replacement based on HSV values. Supports depth background replacement. backgrounds have to be passed as list of tuples of rgb and depth images.

Example

from rsp.nl.dataset import TUCRID
import rsp.ml.multi_transforms as multi_transforms

USE_DEPTH_DATA = False
backgrounds = TUCRID.load_backgrounds(USE_DEPTH_DATA)
tranforms_train = multi_transforms.Compose([
    multi_transforms.ReplaceBackground(
        backgrounds = backgrounds,
        hsv_filter=[(69, 87, 139, 255, 52, 255)],
        p = 0.8
    ),
    multi_transforms.Stack()
])
tucrid = TUCRID('train', load_depth_data=USE_DEPTH_DATA, transforms=tranforms_train)

for X, T in tucrid:
    for x in X:
        img = x.permute(1, 2, 0).numpy()

        cv.imshow('img', img)
        cv.waitKey(30)

5.15.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.15.2 __init__

TOC

Description

Transformation for background replacement based on HSV values. Supports depth background replacement. backgrounds have to be passed as list of tuples of rgb and depth images.

Parameters

Name Type Description
backgrounds List[np.array] List of background images
hsv_filter List[tuple[int, int, int, int, int, int]] List of HSV filters
p float, default = 1. Probability of applying the transformation
rotate float, default = 5 Maximum rotation angle
max_scale float, default = 2 Maximum scaling factor
max_noise float, default = 0.002 Maximum noise level

5.16 Resize : MultiTransform

TOC

Description

MultiTransform is an extension to keep the same transformation over a sequence of images instead of initializing a new transformation for every single image. It is inspired by torchvision.transforms and could be used for video augmentation. Use rsp.ml.multi_transforms.Composeto combine multiple image sequence transformations.

Note rsp.ml.multi_transforms.MultiTransform is a base class and should be inherited.

5.16.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.16.2 __init__

TOC

Description

Initializes a new instance.

5.17 Rotate : MultiTransform

TOC

Description

Randomly rotates images.

Equations

$angle = -max_angle + 2 \cdot random() \cdot max_angle$

5.17.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.17.2 __init__

TOC

Description

Iitializes a new instance.

Parameters

Name Type Description
max_angle float Maximal rotation in degrees
auto_scale bool, default = True Image will be resized when auto scale is activated to avoid black margins.

5.18 Satturation : MultiTransform

TOC

Description

MultiTransform is an extension to keep the same transformation over a sequence of images instead of initializing a new transformation for every single image. It is inspired by torchvision.transforms and could be used for video augmentation. Use rsp.ml.multi_transforms.Composeto combine multiple image sequence transformations.

Note rsp.ml.multi_transforms.MultiTransform is a base class and should be inherited.

5.18.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.18.2 __init__

TOC

Description

Initializes a new instance.

5.19 Scale : MultiTransform

TOC

Description

MultiTransform is an extension to keep the same transformation over a sequence of images instead of initializing a new transformation for every single image. It is inspired by torchvision.transforms and could be used for video augmentation. Use rsp.ml.multi_transforms.Composeto combine multiple image sequence transformations.

Note rsp.ml.multi_transforms.MultiTransform is a base class and should be inherited.

5.19.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.19.2 __init__

TOC

Description

Initializes a new instance.

5.20 Stack : MultiTransform

TOC

Description

MultiTransform is an extension to keep the same transformation over a sequence of images instead of initializing a new transformation for every single image. It is inspired by torchvision.transforms and could be used for video augmentation. Use rsp.ml.multi_transforms.Composeto combine multiple image sequence transformations.

Note rsp.ml.multi_transforms.MultiTransform is a base class and should be inherited.

5.20.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.20.2 __init__

TOC

Description

Initializes a new instance.

5.21 ToCVImage : MultiTransform

TOC

Description

Converts a torch.Tensorto Open CV image by changing dimensions (d0, d1, d2) -> (d1, d2, d0) and converting torch.Tensor to numpy.

5.21.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.21.2 __init__

TOC

Description

Initializes a new instance.

5.22 ToNumpy : MultiTransform

TOC

Description

Converts a torch.Tensorto numpy

5.22.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.22.2 __init__

TOC

Description

Initializes a new instance.

5.23 ToPILImage : MultiTransform

TOC

Description

Converts sequence of images to sequence of PIL.Image.

5.23.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.23.2 __init__

TOC

Description

Initializes a new instance.

5.24 ToTensor : MultiTransform

TOC

Description

Converts a sequence of images to torch.Tensor.

5.24.1 __call__

TOC

Description

Call self as a function.

Parameters

Name Type Description
input torch.Tensor
List[PIL.Image]
List[numpy.array]		 Sequence of images

5.24.2 __init__

TOC

Description

Initializes a new instance.

6 run

TOC

The module rsp.ml.run provides some tools for storing, loading and visualizing data during training of models using PyTorch.

6.1 Run : builtins.object

TOC

Description

Run class to store and manage training

Example

from rsp.ml.run import Run
import rsp.ml.metrics as m

metrics = [
    m.top_1_accuracy
]
config = {
    m.top_1_accuracy.__name__: {
        'ymin': 0,
        'ymax': 1
    }
}
run = Run(id='run0001', metrics=metrics, config=config, ignore_outliers_in_chart_scaling=True)

for epoch in range(100):
    """here goes some training code, giving us inputs, predictions and targets"""
    acc = m.top_1_accuracy(predictions, targets)
    run.append(m.top_1_accuracy.__name__, 'train', acc)

6.1.1 __init__

TOC

Description

Run class to store and manage training

Parameters

Name Type Description
id str, default = None Id of the run. If None, a new id is generated
moving_average_epochs int, default = 1 Number of epochs to average over
metrics list, default = None List of metrics to compute. Each metric should be a function that takes Y and T as input.
device str, default = None torch device to run on
ignore_outliers_in_chart_scaling bool, default = False Ignore outliers when scaling charts
config dict, default = {} Configuration dictionary. Keys are metric names and values are dictionaries with keys 'ymin' and 'ymax'

6.1.2 append

TOC

Description

Append value to key in phase.

Parameters

Name Type Description
key str Key to append to
phase str Phase to append to
value float Value to append

6.1.3 get_avg

TOC

Description

Get last average value of key in phase

Parameters

Name Type Description
key str Key to get
phase str Phase to get from

Returns

Last average value of key in phase. If key is not in data, returns np.nan : value : float

6.1.4 get_val

TOC

Description

Get last value of key in phase

Parameters

Name Type Description
key str Key to get
phase str Phase to get from

Returns

Last value of key in phase. If key is not in data, returns np.nan : value : float

6.1.5 len

TOC

Description

Get length of longest phase

6.1.6 load_best_state_dict

TOC

Description

Load best state_dict from runs/{id}/{fname}

Parameters

Name Type Description
model torch.nn.Module Model to load state_dict into
fname str, default = 'state_dict.pt' Filename to load from
verbose bool, default = False Print loaded file

6.1.7 load_state_dict

TOC

Description

Load state_dict from runs/{id}/{fname}

Parameters

Name Type Description
model torch.nn.Module Model to load state_dict into
fname str, default = None Filename to load from

6.1.8 pickle_dump

TOC

Description

Pickle model to runs/{id}/{fname}

Parameters

Name Type Description
model torch.nn.Module Model to pickle
fname str, default = 'model.pkl' Filename to save to

6.1.9 pickle_load

TOC

Description

Load model from runs/{id}/{fname}

Parameters

Name Type Description
fname str, default = 'model.pkl' Filename to load from

6.1.10 plot

TOC

Description

Plot all keys to runs/{id}/plot/{key}.jpg

6.1.11 recalculate_moving_average

TOC

Description

Recalculate moving average

6.1.12 save

TOC

Description

Save data to runs/{id}/data.json

6.1.13 save_best_state_dict

TOC

Description

Save state_dict if new_acc is better than previous best

Parameters

Name Type Description
state_dict dict State dict to save
new_acc float New accuracy
epoch int, default = None Epoch to save
fname str, default = 'state_dict.pt' Filename to save to

6.1.14 save_state_dict

TOC

Description

Save state_dict to runs/{id}/{fname}

Parameters

Name Type Description
state_dict dict State dict to save
fname str, default = 'state_dict.pt' Filename to save to

6.1.15 train_epoch

TOC

Description

Train one epoch.

Parameters

Name Type Description
dataloader DataLoader DataLoader to train on
model torch.nn.Module Model to train
optimizer torch.optim.Optimizer Optimizer to use
criterion torch.nn.Module Criterion to use
num_batches int, default = None Number of batches to train on. If None, train on all batches
return_YT bool, default = False Append Y and T to results

Returns

Dictionary with results : results : dict

6.1.16 validate_epoch

TOC

Description

Validate one epoch.

Parameters

Name Type Description
dataloader DataLoader DataLoader to validate on
model torch.nn.Module Model to validate
optimizer torch.optim.Optimizer Optimizer to use
criterion torch.nn.Module Criterion to use
num_batches int, default = None Number of batches to validate on. If None, validate on all batches
return_YT bool, default = False Append Y and T to results

Returns

Dictionary with results : results : dict

Project details

Verified details

These details have been verified by PyPI
Maintainers
Avatar for schulzr from gravatar.com schulzr

Unverified details

These details have not been verified by PyPI
Project links
Meta
  • License: MIT
  • Author: Robert Schulz
  • Tags python , Machine Learning
Classifiers

Release history Release notifications | RSS feed

0.0.149

0.0.148

0.0.147

0.0.146

0.0.145

0.0.144

0.0.143

0.0.142

0.0.141

This version

0.0.140

0.0.139

0.0.138

0.0.137

0.0.136

0.0.135

0.0.134

0.0.133

0.0.132

0.0.131

0.0.130

0.0.129

0.0.128

0.0.127

0.0.126

0.0.125

0.0.124

0.0.123

0.0.122

0.0.121

0.0.120

0.0.119

0.0.118

0.0.116

0.0.115

0.0.114

0.0.113

0.0.112

0.0.111

0.0.110

0.0.109

0.0.108

0.0.107

0.0.106

0.0.105

0.0.104

0.0.103

0.0.102

0.0.101

0.0.100

0.0.99

0.0.98

0.0.97

0.0.96

0.0.95

0.0.94

0.0.93

0.0.92

0.0.91

0.0.90

0.0.89

0.0.88

0.0.87

0.0.86

0.0.85

0.0.84

0.0.83

0.0.82

0.0.81

0.0.80

0.0.79

0.0.78

0.0.77

0.0.76

0.0.75

0.0.74

0.0.73

0.0.72

0.0.71

0.0.70

0.0.69

0.0.68

0.0.67

0.0.66

0.0.65

0.0.64

0.0.63

0.0.62

0.0.61

0.0.60

0.0.59

0.0.58

0.0.57

0.0.56

0.0.55

0.0.54

0.0.53

0.0.52

0.0.51

0.0.50

0.0.49

0.0.48

0.0.47

0.0.46

0.0.45

0.0.44

0.0.43

0.0.42

0.0.41

0.0.40

0.0.39

0.0.38

0.0.37

0.0.36

0.0.35

0.0.34

0.0.33

0.0.32

0.0.31

0.0.30

0.0.29

0.0.28

0.0.27

0.0.26

0.0.25

0.0.24

0.0.23

0.0.22

0.0.21

0.0.20

0.0.19

0.0.18

0.0.17

0.0.16

0.0.15

0.0.14

0.0.13

0.0.12

0.0.11

0.0.10

0.0.9

0.0.8

0.0.7

0.0.6

0.0.5

0.0.4

0.0.3

0.0.2

0.0.1

0.0.0

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

rsp_ml-0.0.140.tar.gz (114.1 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

rsp_ml-0.0.140-py3-none-any.whl (60.3 kB view details)

Uploaded Python 3

File details

Details for the file rsp_ml-0.0.140.tar.gz.

File metadata

  • Download URL: rsp_ml-0.0.140.tar.gz
  • Upload date:
  • Size: 114.1 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/5.1.1 CPython/3.11.0

File hashes

Hashes for rsp_ml-0.0.140.tar.gz
Algorithm Hash digest
SHA256 5d1b9fd18d84335346d6f55393b53c53eee7ca9d6f6eaf9fe92bd819cf7c9f81
MD5 52b3cae6529daa68b58f594280c980dc
BLAKE2b-256 ff5dcd72e3633408ac7661f99404fa2ca3ee9c15f3d9ed9e1f30532aceded77f

See more details on using hashes here.

File details

Details for the file rsp_ml-0.0.140-py3-none-any.whl.

File metadata

  • Download URL: rsp_ml-0.0.140-py3-none-any.whl
  • Upload date:
  • Size: 60.3 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/5.1.1 CPython/3.11.0

File hashes

Hashes for rsp_ml-0.0.140-py3-none-any.whl
Algorithm Hash digest
SHA256 0f091e48829104ad6a8f6a6f13283e4a503b9c40d10505b19952be430b7b0e88
MD5 e38a79a65b2752f93374a18f15df3230
BLAKE2b-256 bcf46770491ca2193de95350864623349a6b6909d056269f49aa003d2cfebfeb

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page