espnet.nets.chainer_backend.deterministic_embed_id.embed_id

source

espnet.nets.chainer_backend.deterministic_embed_id.embed_id(x, W, ignore_label=None)

Efficient linear function for one-hot input.

This function implements so called word embeddings. It takes two arguments: a set of IDs (words) x in $B$ dimensional integer vector, and a set of all ID (word) embeddings W in $V \times d$ float32 matrix. It outputs $B \times d$ matrix whose i-th column is the x[i]-th column of W. This function is only differentiable on the input W.

• Parameters:
  • x (chainer.Variable | np.ndarray) – Batch vectors of IDs. Each element must be signed integer.
  • W (chainer.Variable | np.ndarray) – Distributed representation of each ID (a.k.a. word embeddings).
  • ignore_label (int) – If ignore_label is an int value, i-th column of return value is filled with 0.
• Returns: Embedded variable.
• Return type: chainer.Variable

EmbedID

Examples

>>> x = np.array([2, 1]).astype('i')
>>> x
array([2, 1], dtype=int32)
>>> W = np.array([[0, 0, 0],
...               [1, 1, 1],
...               [2, 2, 2]]).astype('f')
>>> W
array([[ 0.,  0.,  0.],
       [ 1.,  1.,  1.],
       [ 2.,  2.,  2.]], dtype=float32)
>>> F.embed_id(x, W).data
array([[ 2.,  2.,  2.],
       [ 1.,  1.,  1.]], dtype=float32)
>>> F.embed_id(x, W, ignore_label=1).data
array([[ 2.,  2.,  2.],
       [ 0.,  0.,  0.]], dtype=float32)