tf.contrib.seq2seq.sequence_loss

Weighted cross-entropy loss for a sequence of logits.

Depending on the values of average_across_timesteps / sum_over_timesteps and average_across_batch / sum_over_batch, the return Tensor will have rank 0, 1, or 2 as these arguments reduce the cross-entropy at each target, which has shape [batch_size, sequence_length], over their respective dimensions. For example, if average_across_timesteps is True and average_across_batch is False, then the return Tensor will have shape [batch_size].

Note that average_across_timesteps and sum_over_timesteps cannot be True at same time. Same for average_across_batch and sum_over_batch.

The recommended loss reduction in tf 2.0 has been changed to sum_over, instead of weighted average. User are recommend to use sum_over_timesteps and sum_over_batch for reduction.

Args
logits A Tensor of shape [batch_size, sequence_length, num_decoder_symbols] and dtype float. The logits correspond to the prediction across all classes at each timestep.
targets A Tensor of shape [batch_size, sequence_length] and dtype int. The target represents the true class at each timestep.
weights A Tensor of shape [batch_size, sequence_length] and dtype float. weights constitutes the weighting of each prediction in the sequence. When using weights as masking, set all valid timesteps to 1 and all padded timesteps to 0, e.g. a mask returned by tf.sequence_mask.
average_across_timesteps If set, sum the cost across the sequence dimension and divide the cost by the total label weight across timesteps.
average_across_batch If set, sum the cost across the batch dimension and divide the returned cost by the batch size.
sum_over_timesteps If set, sum the cost across the sequence dimension and divide the size of the sequence. Note that any element with 0 weights will be excluded from size calculation.
sum_over_batch if set, sum the cost across the batch dimension and divide the total cost by the batch size. Not that any element with 0 weights will be excluded from size calculation.
softmax_loss_function Function (labels, logits) -> loss-batch to be used instead of the standard softmax (the default if this is None). Note that to avoid confusion, it is required for the function to accept named arguments.
name Optional name for this operation, defaults to "sequence_loss".
Returns
A float Tensor of rank 0, 1, or 2 depending on the average_across_timesteps and average_across_batch arguments. By default, it has rank 0 (scalar) and is the weighted average cross-entropy (log-perplexity) per symbol.
Raises
ValueError logits does not have 3 dimensions or targets does not have 2 dimensions or weights does not have 2 dimensions.

© 2020 The TensorFlow Authors. All rights reserved.
Licensed under the Creative Commons Attribution License 3.0.
Code samples licensed under the Apache 2.0 License.
https://www.tensorflow.org/versions/r1.15/api_docs/python/tf/contrib/seq2seq/sequence_loss