manim_ml.neural_network.layers package#

Submodules#

manim_ml.neural_network.layers.convolutional module#

manim_ml.neural_network.layers.convolutional_to_convolutional module#

manim_ml.neural_network.layers.embedding module#

class manim_ml.neural_network.layers.embedding.EmbeddingLayer(point_radius=0.02, mean=array([0, 0]), covariance=array([[1., 0.], [0., 1.]]), dist_theme='gaussian', paired_query_mode=False, **kwargs)#

Bases: VGroupNeuralNetworkLayer

NeuralNetwork embedding object that can show probability distributions

add_gaussian_distribution(gaussian_distribution)#

Adds given GaussianDistribution to the list

animation_overrides = {<class 'manim.animation.creation.Create'>: <function EmbeddingLayer._create_override>}#
construct_gaussian_point_cloud(mean, covariance, point_color='#FFFFFF', num_points=400)#

Plots points sampled from a Gaussian with the given mean and covariance

construct_layer(input_layer: NeuralNetworkLayer, output_layer: NeuralNetworkLayer, **kwargs)#

Constructs the layer at network construction time

Parameters:
get_distribution_location()#

Returns mean of latent distribution in axes frame

make_forward_pass_animation(layer_args={}, **kwargs)#

Forward pass animation

remove_gaussian_distribution(gaussian_distribution)#

Removes the given gaussian distribution from the embedding

sample_point_location_from_distribution()#

Samples from the current latent distribution

class manim_ml.neural_network.layers.embedding.NeuralNetworkEmbeddingTestScene(renderer=None, camera_class=<class 'manim.camera.camera.Camera'>, always_update_mobjects=False, random_seed=None, skip_animations=False)#

Bases: Scene

construct()#

Add content to the Scene.

From within Scene.construct(), display mobjects on screen by calling Scene.add() and remove them from screen by calling Scene.remove(). All mobjects currently on screen are kept in Scene.mobjects. Play animations by calling Scene.play().

Notes

Initialization code should go in Scene.setup(). Termination code should go in Scene.tear_down().

Examples

A typical manim script includes a class derived from Scene with an overridden Scene.contruct() method:

class MyScene(Scene):
    def construct(self):
        self.play(Write(Text("Hello World!")))

See also

Scene.setup(), Scene.render(), Scene.tear_down()

manim_ml.neural_network.layers.embedding_to_feed_forward module#

class manim_ml.neural_network.layers.embedding_to_feed_forward.EmbeddingToFeedForward(input_layer, output_layer, animation_dot_color='#FC6255', dot_radius=0.03, **kwargs)#

Bases: ConnectiveLayer

Feed Forward to Embedding Layer

animation_overrides = {<class 'manim.animation.creation.Create'>: <function EmbeddingToFeedForward._create_override>}#
construct_layer(input_layer: NeuralNetworkLayer, output_layer: NeuralNetworkLayer, **kwargs)#

Constructs the layer at network construction time

Parameters:
input_class#

alias of EmbeddingLayer

make_forward_pass_animation(layer_args={}, run_time=1.5, **kwargs)#

Makes dots diverge from the given location and move the decoder

output_class#

alias of FeedForwardLayer

manim_ml.neural_network.layers.feed_forward module#

class manim_ml.neural_network.layers.feed_forward.FeedForwardLayer(num_nodes, layer_buffer=0.05, node_radius=0.08, node_color='#58C4DD', node_outline_color='#FFFFFF', rectangle_color='#FFFFFF', node_spacing=0.3, rectangle_fill_color='#000000', node_stroke_width=2.0, rectangle_stroke_width=2.0, animation_dot_color='#FF862F', activation_function=None, **kwargs)#

Bases: VGroupNeuralNetworkLayer

Handles rendering a layer for a neural network

animation_overrides = {<class 'manim.animation.creation.Create'>: <function FeedForwardLayer._create_override>}#
construct_activation_function()#

Construct the activation function

construct_layer(input_layer: NeuralNetworkLayer, output_layer: NeuralNetworkLayer, **kwargs)#

Creates the neural network layer

get_center()#

Get center coordinates

get_height()#
get_left()#

Get left coordinates of a box bounding the Mobject

get_right()#

Get right coordinates of a box bounding the Mobject

make_dropout_forward_pass_animation(layer_args, **kwargs)#

Makes a forward pass animation with dropout

make_forward_pass_animation(layer_args={}, **kwargs)#
move_to(mobject_or_point)#

Moves the center of the layer to the given mobject or point

manim_ml.neural_network.layers.feed_forward_to_embedding module#

class manim_ml.neural_network.layers.feed_forward_to_embedding.FeedForwardToEmbedding(input_layer, output_layer, animation_dot_color='#FC6255', dot_radius=0.03, **kwargs)#

Bases: ConnectiveLayer

Feed Forward to Embedding Layer

animation_overrides = {<class 'manim.animation.creation.Create'>: <function FeedForwardToEmbedding._create_override>}#
construct_layer(input_layer: NeuralNetworkLayer, output_layer: NeuralNetworkLayer, **kwargs)#

Constructs the layer at network construction time

Parameters:
input_class#

alias of FeedForwardLayer

make_forward_pass_animation(layer_args={}, run_time=1.5, **kwargs)#

Makes dots converge on a specific location

output_class#

alias of EmbeddingLayer

manim_ml.neural_network.layers.feed_forward_to_feed_forward module#

class manim_ml.neural_network.layers.feed_forward_to_feed_forward.FeedForwardToFeedForward(input_layer, output_layer, passing_flash=True, dot_radius=0.05, animation_dot_color='#FF862F', edge_color='#FFFFFF', edge_width=1.5, camera=None, **kwargs)#

Bases: ConnectiveLayer

Layer for connecting FeedForward layer to FeedForwardLayer

animation_overrides = {<class 'manim.animation.creation.Create'>: <function FeedForwardToFeedForward._create_override>, <class 'manim.animation.fading.FadeOut'>: <function FeedForwardToFeedForward._fadeout_animation>}#
construct_edges()#
construct_layer(input_layer: NeuralNetworkLayer, output_layer: NeuralNetworkLayer, **kwargs)#

Constructs the layer at network construction time

Parameters:
input_class#

alias of FeedForwardLayer

make_forward_pass_animation(layer_args={}, run_time=1, feed_forward_dropout=0.0, **kwargs)#

Animation for passing information from one FeedForwardLayer to the next

modify_edge_colors(colors=None, magnitudes=None, color_scheme='inferno')#

Changes the colors of edges

modify_edge_stroke_widths(widths)#

Changes the widths of the edges

output_class#

alias of FeedForwardLayer

manim_ml.neural_network.layers.feed_forward_to_image module#

class manim_ml.neural_network.layers.feed_forward_to_image.FeedForwardToImage(input_layer, output_layer, animation_dot_color='#FC6255', dot_radius=0.05, **kwargs)#

Bases: ConnectiveLayer

Image Layer to FeedForward layer

animation_overrides = {<class 'manim.animation.creation.Create'>: <function FeedForwardToImage._create_override>}#
construct_layer(input_layer: NeuralNetworkLayer, output_layer: NeuralNetworkLayer, **kwargs)#

Constructs the layer at network construction time

Parameters:
input_class#

alias of FeedForwardLayer

make_forward_pass_animation(layer_args={}, **kwargs)#

Makes dots diverge from the given location and move to the feed forward nodes decoder

output_class#

alias of ImageLayer

manim_ml.neural_network.layers.feed_forward_to_vector module#

class manim_ml.neural_network.layers.feed_forward_to_vector.FeedForwardToVector(input_layer, output_layer, animation_dot_color='#FC6255', dot_radius=0.05, **kwargs)#

Bases: ConnectiveLayer

Image Layer to FeedForward layer

animation_overrides = {<class 'manim.animation.creation.Create'>: <function FeedForwardToVector._create_override>}#
construct_layer(input_layer: NeuralNetworkLayer, output_layer: NeuralNetworkLayer, **kwargs)#

Constructs the layer at network construction time

Parameters:
input_class#

alias of FeedForwardLayer

make_forward_pass_animation(layer_args={}, **kwargs)#

Makes dots diverge from the given location and move to the feed forward nodes decoder

output_class#

alias of VectorLayer

manim_ml.neural_network.layers.image module#

class manim_ml.neural_network.layers.image.ImageLayer(numpy_image, height=1.5, show_image_on_create=True, **kwargs)#

Bases: NeuralNetworkLayer

Single Image Layer for Neural Network

animation_overrides = {<class 'manim.animation.creation.Create'>: <function ImageLayer._create_override>}#
construct_layer(input_layer, output_layer, **kwargs)#

Construct layer method

Parameters:

  • input_layer – Input layer

  • output_layer – Output layer

classmethod from_path(image_path, grayscale=True, **kwargs)#

Creates a query using the paths

get_right()#

Override get right

property height#

The height of the mobject.

Return type:

float

Examples

Example: HeightExample 

from manim import *

class HeightExample(Scene):
    def construct(self):
        decimal = DecimalNumber().to_edge(UP)
        rect = Rectangle(color=BLUE)
        rect_copy = rect.copy().set_stroke(GRAY, opacity=0.5)

        decimal.add_updater(lambda d: d.set_value(rect.height))

        self.add(rect_copy, rect, decimal)
        self.play(rect.animate.set(height=5))
        self.wait()

See also

length_over_dim()

make_forward_pass_animation(layer_args={}, **kwargs)#
scale(scale_factor, **kwargs)#

Scales the image mobject

property width#

The width of the mobject.

Return type:

float

Examples

Example: WidthExample 

from manim import *

class WidthExample(Scene):
    def construct(self):
        decimal = DecimalNumber().to_edge(UP)
        rect = Rectangle(color=BLUE)
        rect_copy = rect.copy().set_stroke(GRAY, opacity=0.5)

        decimal.add_updater(lambda d: d.set_value(rect.width))

        self.add(rect_copy, rect, decimal)
        self.play(rect.animate.set(width=7))
        self.wait()

See also

length_over_dim()

manim_ml.neural_network.layers.image_to_feed_forward module#

class manim_ml.neural_network.layers.image_to_feed_forward.ImageToFeedForward(input_layer, output_layer, animation_dot_color='#FC6255', dot_radius=0.05, **kwargs)#

Bases: ConnectiveLayer

Image Layer to FeedForward layer

animation_overrides = {<class 'manim.animation.creation.Create'>: <function ImageToFeedForward._create_override>}#
construct_layer(input_layer: NeuralNetworkLayer, output_layer: NeuralNetworkLayer, **kwargs)#

Constructs the layer at network construction time

Parameters:
input_class#

alias of ImageLayer

make_forward_pass_animation(layer_args={}, **kwargs)#

Makes dots diverge from the given location and move to the feed forward nodes decoder

output_class#

alias of FeedForwardLayer

manim_ml.neural_network.layers.paired_query module#

class manim_ml.neural_network.layers.paired_query.PairedQueryLayer(positive, negative, stroke_width=5, font_size=18, spacing=0.5, **kwargs)#

Bases: NeuralNetworkLayer

Paired Query Layer

animation_overrides = {<class 'manim.animation.creation.Create'>: <function PairedQueryLayer._create_override>}#
construct_layer(input_layer: NeuralNetworkLayer, output_layer: NeuralNetworkLayer, **kwargs)#

Constructs the layer at network construction time

Parameters:
classmethod from_paths(positive_path, negative_path, grayscale=True, **kwargs)#

Creates a query using the paths

make_assets()#

Constructs the assets needed for a query layer

make_forward_pass_animation(layer_args={}, **kwargs)#

Forward pass for query

manim_ml.neural_network.layers.paired_query_to_feed_forward module#

class manim_ml.neural_network.layers.paired_query_to_feed_forward.PairedQueryToFeedForward(input_layer, output_layer, animation_dot_color='#FC6255', dot_radius=0.02, **kwargs)#

Bases: ConnectiveLayer

PairedQuery layer to FeedForward layer

animation_overrides = {<class 'manim.animation.creation.Create'>: <function PairedQueryToFeedForward._create_override>}#
construct_layer(input_layer: NeuralNetworkLayer, output_layer: NeuralNetworkLayer, **kwargs)#

Constructs the layer at network construction time

Parameters:
input_class#

alias of PairedQueryLayer

make_forward_pass_animation(layer_args={}, **kwargs)#

Makes dots diverge from the given location and move to the feed forward nodes decoder

output_class#

alias of FeedForwardLayer

manim_ml.neural_network.layers.parent_layers module#

class manim_ml.neural_network.layers.parent_layers.BlankConnective(input_layer, output_layer, **kwargs)#

Bases: ConnectiveLayer

Connective layer to be used when the given pair of layers is undefined

animation_overrides = {<class 'manim.animation.creation.Create'>: <function BlankConnective._create_override>}#
make_forward_pass_animation(run_time=1.5, layer_args={}, **kwargs)#
class manim_ml.neural_network.layers.parent_layers.ConnectiveLayer(input_layer, output_layer, **kwargs)#

Bases: VGroupNeuralNetworkLayer

Forward pass animation for a given pair of layers

animation_overrides = {<class 'manim.animation.creation.Create'>: <function ConnectiveLayer._create_override>}#
abstract make_forward_pass_animation(run_time=2.0, layer_args={}, **kwargs)#
class manim_ml.neural_network.layers.parent_layers.NeuralNetworkLayer(text=None, *args, **kwargs)#

Bases: ABC, Group

Abstract Neural Network Layer class

animation_overrides = {<class 'manim.animation.creation.Create'>: <function NeuralNetworkLayer._create_override>}#
abstract construct_layer(input_layer: NeuralNetworkLayer, output_layer: NeuralNetworkLayer, **kwargs)#

Constructs the layer at network construction time

Parameters:
abstract make_forward_pass_animation(layer_args={}, **kwargs)#
class manim_ml.neural_network.layers.parent_layers.ThreeDLayer#

Bases: ABC

Abstract class for 3D layers

class manim_ml.neural_network.layers.parent_layers.VGroupNeuralNetworkLayer(*args, **kwargs)#

Bases: NeuralNetworkLayer

animation_overrides = {<class 'manim.animation.creation.Create'>: <function VGroupNeuralNetworkLayer._create_override>}#
abstract make_forward_pass_animation(**kwargs)#

manim_ml.neural_network.layers.triplet module#

class manim_ml.neural_network.layers.triplet.TripletLayer(anchor, positive, negative, stroke_width=5, font_size=22, buff=0.2, **kwargs)#

Bases: NeuralNetworkLayer

Shows triplet images

animation_overrides = {<class 'manim.animation.creation.Create'>: <function TripletLayer._create_override>}#
construct_layer(input_layer: NeuralNetworkLayer, output_layer: NeuralNetworkLayer, **kwargs)#

Constructs the layer at network construction time

Parameters:
classmethod from_paths(anchor_path, positive_path, negative_path, grayscale=True, font_size=22, buff=0.2)#

Creates a triplet using the anchor paths

make_assets()#

Constructs the assets needed for a triplet layer

make_forward_pass_animation(layer_args={}, **kwargs)#

Forward pass for triplet

manim_ml.neural_network.layers.triplet_to_feed_forward module#

class manim_ml.neural_network.layers.triplet_to_feed_forward.TripletToFeedForward(input_layer, output_layer, animation_dot_color='#FC6255', dot_radius=0.02, **kwargs)#

Bases: ConnectiveLayer

TripletLayer to FeedForward layer

animation_overrides = {<class 'manim.animation.creation.Create'>: <function TripletToFeedForward._create_override>}#
construct_layer(input_layer: NeuralNetworkLayer, output_layer: NeuralNetworkLayer, **kwargs)#

Constructs the layer at network construction time

Parameters:
input_class#

alias of TripletLayer

make_forward_pass_animation(layer_args={}, **kwargs)#

Makes dots diverge from the given location and move to the feed forward nodes decoder

output_class#

alias of FeedForwardLayer

manim_ml.neural_network.layers.util module#

manim_ml.neural_network.layers.util.get_connective_layer(input_layer, output_layer)#

Deduces the relevant connective layer

manim_ml.neural_network.layers.vector module#

class manim_ml.neural_network.layers.vector.VectorLayer(num_values, value_func=<function VectorLayer.<lambda>>, **kwargs)#

Bases: VGroupNeuralNetworkLayer

Shows a vector

animation_overrides = {<class 'manim.animation.creation.Create'>: <function VectorLayer._create_override>}#
construct_layer(input_layer: NeuralNetworkLayer, output_layer: NeuralNetworkLayer, **kwargs)#

Constructs the layer at network construction time

Parameters:
make_forward_pass_animation(layer_args={}, **kwargs)#
make_vector()#

Makes the vector

Module contents#