def get_normalized_probs(self, ctc_logits, logits, log_probs):
"""Get normalized probabilities (or log probs) from a net's output."""
if log_probs:
ctc_res = utils.log_softmax(ctc_logits.float(), dim=-1)
res = utils.log_softmax(logits.float(), dim=-1)
else:
ctc_res = utils.softmax(ctc_logits.float(), dim=-1)
res = utils.softmax(logits.float(), dim=-1)
ctc_res.batch_first = True
res.batch_first = True
return ctc_res, res
def get_normalized_probs(self, net_output, log_probs):
"""Get normalized probabilities (or log probs) from a net's output."""
logits = net_output["encoder_out"]
if log_probs:
return utils.log_softmax(logits.float(), dim=-1)
else:
return utils.softmax(logits.float(), dim=-1)
def get_normalized_probs(
self,
net_output: Tuple[Tensor, Optional[Dict[str, List[Optional[Tensor]]]]],
log_probs: bool,
sample: Optional[Dict[str, Tensor]] = None,
):
"""Get normalized probabilities (or log probs) from a net's output."""
if hasattr(self, "adaptive_softmax") and self.adaptive_softmax is not None:
if sample is not None:
assert "target" in sample
target = sample["target"]
else:
target = None
out = self.adaptive_softmax.get_log_prob(net_output[0], target=target)
return out.exp_() if not log_probs else out
logits = net_output[0]
if log_probs:
return utils.log_softmax(logits, dim=-1, onnx_trace=self.onnx_trace)
else:
return utils.softmax(logits, dim=-1, onnx_trace=self.onnx_trace)