Composes several transforms together. Args: transforms(list of ``Transform`` objects): list of transforms to compose. Example: Ret

r""" Applies a linear transformation together with a non-linear activation to the incoming tensor: :math:`y = \mathrm{Activation}(x A^T + b)`

r""" Biaffine layer for first-order scoring. This function has a tensor of weights :math:`W` and bias terms if needed. The score :math:`s

Conditional random field. This module implements a conditional random field [LMP01]_. The forward computation of this class computes the log

Categorical Top-K accuracy. Assumes integer labels, with each item to be classified having a single correct class. Tie break enables equa

Given an input shaped like `(batch_size, time_steps, [rest])` and a `Module` that takes inputs like `(batch_size, [rest])`, `TimeDistributed`

Multi-headed attention from 'Attention Is All You Need' paper

A basic sequence tagger that decodes from inputs of word embeddings

Adam enables L2 weight decay and clip_by_global_norm on gradients. Just adding the square of the weights to the loss function is *not* the

Transformer decoder consisting of *config.decoder_layers* layers. Each layer is a [`BartDecoderLayer`] Args: config: BartConfig

This module learns positional embeddings up to a fixed maximum size.

Conditional Random Field layer (tf.keras) `CRF` can be used as the last layer in a network (as a classifier). Input shape (features) must be e

r""" The implementation of CRF Constituency Parser, also called FANCY (abbr. of Fast and Accurate Neural Crf constituencY) Parser. Refere

Computes a parameterised scalar mixture of N tensors, ``mixture = gamma * sum(s_k * tensor_k)`` where ``s = softmax(w)``, with ``w`` and ``gamma``