View Source Bumblebee.Vision.ConvNext (Bumblebee v0.6.0)

ConvNeXT model family.

Architectures

:base - plain ConvNeXT without any head on top
:for_image_classification - ConvNeXT with a classification head. The head consists of a single dense layer on top of the pooled features

"pixel_values" - {batch_size, height, width, num_channels}
Featurized image pixel values (224x224).

:output_hidden_states - when true, the model output includes all hidden states

:num_channels - the number of channels in the input. Defaults to 3
:patch_size - the size of the patch spatial dimensions. Defaults to 4
:hidden_sizes - the dimensionality of hidden layers at each stage. Defaults to [96, 192, 384, 768]
:depths - the depth (number of residual blocks) at each stage. Defaults to [3, 3, 9, 3]
:activation - the activation function. Defaults to :gelu
:scale_initial_value - the initial value for scaling layers. Defaults to 1.0e-6
:drop_path_rate - the drop path rate used to for stochastic depth. Defaults to 0.0
:layer_norm_epsilon - the epsilon used by the layer normalization layers. Defaults to 1.0e-12
:initializer_scale - the standard deviation of the normal initializer used for initializing kernel parameters. Defaults to 0.02
:num_labels - the number of labels to use in the last layer for the classification task. Defaults to 2
:id_to_label - a map from class index to label. Defaults to %{}