in_batch_sparse_scores¶
Computes dot product between anchor, positive and negative activations.
Parameters¶
- activations
Examples¶
>>> from neural_cherche import utils
>>> activations = torch.tensor([
... [0, 0, 0, 0, 0, 0, 0, 0, 0, 1],
... [0, 0, 0, 0, 0, 0, 0, 0, 0, 2],
... [0, 0, 0, 0, 0, 0, 0, 0, 0, 3],
... ], device="mps")
>>> utils.in_batch_sparse_scores(
... activations=activations,
... )
tensor([5, 8, 9], device='mps:0')