tfrs.layers.loss.RemoveAccidentalHits

Zeroes the logits of accidental negatives.

tfrs.layers.loss.RemoveAccidentalHits(
    trainable=True, name=None, dtype=None, dynamic=False, **kwargs
)

Methods

call(
    labels: tf.Tensor, logits: tf.Tensor, candidate_ids: tf.Tensor
) -> tf.Tensor

Zeros selected logits.

For each row in the batch, zeros the logits of negative candidates that have the same id as the positive candidate in that row.

Args
`labels`	[batch_size, num_candidates] one-hot labels tensor.
`logits`	[batch_size, num_candidates] logits tensor.
`candidate_ids`	[num_candidates] candidate identifiers tensor

Returns
`logits`	Modified logits.

Except as otherwise noted, the content of this page is licensed under the Creative Commons Attribution 4.0 License, and code samples are licensed under the Apache 2.0 License. For details, see the Google Developers Site Policies. Java is a registered trademark of Oracle and/or its affiliates.

Last updated 2024-04-26 UTC.