From 6f7e27eb7c2d0a7649ec8f14d167c8da8e29f906 Mon Sep 17 00:00:00 2001
From: jmwang66 <wangjiaming.wjm@alibaba-inc.com>
Date: 星期二, 16 五月 2023 15:07:20 +0800
Subject: [PATCH] Merge pull request #518 from alibaba-damo-academy/dev_wjm2
---
funasr/models/e2e_tp.py | 15 ++++++---------
1 files changed, 6 insertions(+), 9 deletions(-)
diff --git a/funasr/models/e2e_tp.py b/funasr/models/e2e_tp.py
index d1367ab..33948f9 100644
--- a/funasr/models/e2e_tp.py
+++ b/funasr/models/e2e_tp.py
@@ -17,9 +17,8 @@
from funasr.modules.add_sos_eos import add_sos_eos
from funasr.modules.nets_utils import make_pad_mask, pad_list
from funasr.torch_utils.device_funcs import force_gatherable
-from funasr.train.abs_espnet_model import AbsESPnetModel
+from funasr.models.base_model import FunASRModel
from funasr.models.predictor.cif import CifPredictorV3
-
if LooseVersion(torch.__version__) >= LooseVersion("1.6.0"):
from torch.cuda.amp import autocast
@@ -30,7 +29,7 @@
yield
-class TimestampPredictor(AbsESPnetModel):
+class TimestampPredictor(FunASRModel):
"""
Author: Speech Lab of DAMO Academy, Alibaba Group
"""
@@ -56,7 +55,7 @@
self.predictor_bias = predictor_bias
self.criterion_pre = mae_loss()
self.token_list = token_list
-
+
def forward(
self,
speech: torch.Tensor,
@@ -65,7 +64,6 @@
text_lengths: torch.Tensor,
) -> Tuple[torch.Tensor, Dict[str, torch.Tensor], torch.Tensor]:
"""Frontend + Encoder + Decoder + Calc loss
-
Args:
speech: (Batch, Length, ...)
speech_lengths: (Batch, )
@@ -113,7 +111,6 @@
self, speech: torch.Tensor, speech_lengths: torch.Tensor
) -> Tuple[torch.Tensor, torch.Tensor]:
"""Frontend + Encoder. Note that this method is used by asr_inference.py
-
Args:
speech: (Batch, Length, ...)
speech_lengths: (Batch, )
@@ -128,7 +125,7 @@
encoder_out, encoder_out_lens, _ = self.encoder(feats, feats_lengths)
return encoder_out, encoder_out_lens
-
+
def _extract_feats(
self, speech: torch.Tensor, speech_lengths: torch.Tensor
) -> Tuple[torch.Tensor, torch.Tensor]:
@@ -151,8 +148,8 @@
encoder_out_mask = (~make_pad_mask(encoder_out_lens, maxlen=encoder_out.size(1))[:, None, :]).to(
encoder_out.device)
ds_alphas, ds_cif_peak, us_alphas, us_peaks = self.predictor.get_upsample_timestamp(encoder_out,
- encoder_out_mask,
- token_num)
+ encoder_out_mask,
+ token_num)
return ds_alphas, ds_cif_peak, us_alphas, us_peaks
def collect_feats(
--
Gitblit v1.9.1