mindspore.ops.CTCLossV2

class mindspore.ops.CTCLossV2(blank=0, reduction='none', zero_infinity=False)[source]

Calculates the CTC (Connectionist Temporal Classification) loss and the gradient.

The CTC algorithm is proposed in Connectionist Temporal Classification: Labeling Unsegmented Sequence Data with Recurrent Neural Networks.

Warning

This is an experimental API that is subject to change or deletion.

Parameters

  • blank (int, optional) – The blank label. Default: 0.

  • reduction (str, optional) – Apply specific reduction method to the output. Currently only support ‘none’, not case sensitive. Default: “none”.

  • zero_infinity (bool, optional) – If loss is infinite, this parameter determines whether to set that loss and its correlated gradient to zero. Default: False.

Inputs:

  • log_probs (Tensor) - A tensor of shape \((T, C, N)\), where \(T\) is input length, \(N\) is batch size and \(C\) is number of classes (including blank).

  • targets (Tensor) - A tensor of shape \((N, S)\), where \(S\) is max target length, means the target sequences.

  • input_lengths (Union(Tuple, Tensor)) - A tuple or Tensor of shape \((N)\). It means the lengths of the input.

  • target_lengths (Union(Tuple, Tensor)) - A tuple or Tensor of shape \((N)\). It means the lengths of the target.

Outputs:

  • neg_log_likelihood (Tensor) - A loss value which is differentiable with respect to each input node.

  • log_alpha (Tensor) - The probability of possible trace of input to target.

Raises

  • TypeError – If zero_infinity is not a bool.

  • TypeError – If reduction is not string.

  • TypeError – If the dtype of log_probs is not float or double.

  • TypeError – If the dtype of targets, input_lengths or target_lengths is not int32 or int64.

  • ValueError – If the rank of log_probs is not 3.

  • ValueError – If the rank of targets is not 2.

  • ValueError – If the shape of input_lengths does not match batch_size \(N\).

  • ValueError – If the shape of target_lengths does not match batch_size \(N\).

  • TypeError – If the types of targets, input_lengths or target_lengths are different.

  • ValueError – If the value of blank is not in range [0, C).

  • RuntimeError – If any value of input_lengths is larger than (num_labels|C).

  • RuntimeError – If any target_lengths[i] is not in range [0, input_length[i]].

Supported Platforms:

Ascend GPU CPU

Examples

>>> log_probs = Tensor(np.array([[[0.3, 0.6, 0.6]],
...                              [[0.9, 0.4, 0.2]]]).astype(np.float32))
>>> targets = Tensor(np.array([[0, 1]]), mstype.int32)
>>> input_lengths = Tensor(np.array([2]), mstype.int32)
>>> target_lengths = Tensor(np.array([1]), mstype.int32)
>>> CTCLossV2 = ops.CTCLossV2(blank=0, reduction='none', zero_infinity=False)
>>> neg_log_hood, log_alpha = CTCLossV2(
...     log_probs, targets, input_lengths, target_lengths)
>>> print(neg_log_hood)
[-2.2986124]
>>> print(log_alpha)
[[[0.3       0.3            -inf      -inf      -inf]
  [1.2       1.8931472 1.2            -inf      -inf]]]