From 8ae9fa8365eba7d33c8d8f5fa51d12903ca6a409 Mon Sep 17 00:00:00 2001
From: haoneng.lhn <haoneng.lhn@alibaba-inc.com>
Date: 星期四, 21 九月 2023 16:26:51 +0800
Subject: [PATCH] update
---
funasr/models/e2e_asr_paraformer.py | 14 ++++----------
1 files changed, 4 insertions(+), 10 deletions(-)
diff --git a/funasr/models/e2e_asr_paraformer.py b/funasr/models/e2e_asr_paraformer.py
index 09af2cd..e157454 100644
--- a/funasr/models/e2e_asr_paraformer.py
+++ b/funasr/models/e2e_asr_paraformer.py
@@ -10,7 +10,6 @@
import torch
import random
import numpy as np
-from typeguard import check_argument_types
from funasr.layers.abs_normalize import AbsNormalize
from funasr.losses.label_smoothing_loss import (
@@ -80,7 +79,6 @@
postencoder: Optional[AbsPostEncoder] = None,
use_1st_decoder_loss: bool = False,
):
- assert check_argument_types()
assert 0.0 <= ctc_weight <= 1.0, ctc_weight
assert 0.0 <= interctc_weight < 1.0, interctc_weight
@@ -242,7 +240,7 @@
loss = self.ctc_weight * loss_ctc + (1 - self.ctc_weight) * loss_att + loss_pre * self.predictor_weight
if self.use_1st_decoder_loss and pre_loss_att is not None:
- loss = loss + pre_loss_att
+ loss = loss + (1 - self.ctc_weight) * pre_loss_att
# Collect Attn branch stats
stats["loss_att"] = loss_att.detach() if loss_att is not None else None
@@ -645,7 +643,6 @@
postencoder: Optional[AbsPostEncoder] = None,
use_1st_decoder_loss: bool = False,
):
- assert check_argument_types()
assert 0.0 <= ctc_weight <= 1.0, ctc_weight
assert 0.0 <= interctc_weight < 1.0, interctc_weight
@@ -1160,8 +1157,8 @@
mask_chunk_predictor=mask_chunk_predictor,
target_label_length=None,
)
- predictor_alignments, predictor_alignments_len = self.predictor.gen_frame_alignments(pre_alphas[:, :-1],
- encoder_out_lens)
+ predictor_alignments, predictor_alignments_len = self.predictor.gen_frame_alignments(pre_alphas,
+ encoder_out_lens+1 if self.predictor.tail_threshold > 0.0 else encoder_out_lens)
scama_mask = None
if self.encoder.overlap_chunk_cls is not None and self.decoder_attention_chunk_type == 'chunk':
@@ -1255,7 +1252,6 @@
preencoder: Optional[AbsPreEncoder] = None,
postencoder: Optional[AbsPostEncoder] = None,
):
- assert check_argument_types()
assert 0.0 <= ctc_weight <= 1.0, ctc_weight
assert 0.0 <= interctc_weight < 1.0, interctc_weight
@@ -1528,7 +1524,6 @@
preencoder: Optional[AbsPreEncoder] = None,
postencoder: Optional[AbsPostEncoder] = None,
):
- assert check_argument_types()
assert 0.0 <= ctc_weight <= 1.0, ctc_weight
assert 0.0 <= interctc_weight < 1.0, interctc_weight
@@ -1806,7 +1801,6 @@
preencoder: Optional[AbsPreEncoder] = None,
postencoder: Optional[AbsPostEncoder] = None,
):
- assert check_argument_types()
assert 0.0 <= ctc_weight <= 1.0, ctc_weight
assert 0.0 <= interctc_weight < 1.0, interctc_weight
@@ -2113,7 +2107,7 @@
return loss_att, acc_att, cer_att, wer_att, loss_pre
- def cal_decoder_with_predictor(self, encoder_out, encoder_out_lens, sematic_embeds, ys_pad_lens, hw_list=None):
+ def cal_decoder_with_predictor(self, encoder_out, encoder_out_lens, sematic_embeds, ys_pad_lens, hw_list=None, clas_scale=1.0):
if hw_list is None:
# default hotword list
hw_list = [torch.Tensor([self.sos]).long().to(encoder_out.device)] # empty hotword list
--
Gitblit v1.9.1