From eb92e79fb94e7b3df8f27c8ce3e607a70dff2a2e Mon Sep 17 00:00:00 2001
From: 语帆 <yf352572@alibaba-inc.com>
Date: 星期三, 28 二月 2024 15:21:32 +0800
Subject: [PATCH] test
---
funasr/models/lcbnet/model.py | 20 ++++++++++++--------
1 files changed, 12 insertions(+), 8 deletions(-)
diff --git a/funasr/models/lcbnet/model.py b/funasr/models/lcbnet/model.py
index d1ebc5c..45b1ee5 100644
--- a/funasr/models/lcbnet/model.py
+++ b/funasr/models/lcbnet/model.py
@@ -21,6 +21,7 @@
from funasr.utils import postprocess_utils
from funasr.utils.datadir_writer import DatadirWriter
from funasr.register import tables
+
import pdb
@tables.register("model_classes", "LCBNet")
class LCBNet(nn.Module):
@@ -91,6 +92,7 @@
fusion_encoder = fusion_encoder_class(**fusion_encoder_conf)
bias_predictor_class = tables.encoder_classes.get(bias_predictor)
bias_predictor = bias_predictor_class(**bias_predictor_conf)
+
if decoder is not None:
decoder_class = tables.decoder_classes.get(decoder)
@@ -272,15 +274,15 @@
ind: int
"""
with autocast(False):
-
+ pdb.set_trace()
# Data augmentation
if self.specaug is not None and self.training:
speech, speech_lengths = self.specaug(speech, speech_lengths)
-
+ pdb.set_trace()
# Normalization for feature: e.g. Global-CMVN, Utterance-CMVN
if self.normalize is not None:
speech, speech_lengths = self.normalize(speech, speech_lengths)
-
+ pdb.set_trace()
# Forward encoder
# feats: (Batch, Length, Dim)
# -> encoder_out: (Batch, Length2, Dim2)
@@ -297,7 +299,7 @@
if intermediate_outs is not None:
return (encoder_out, intermediate_outs), encoder_out_lens
-
+ pdb.set_trace()
return encoder_out, encoder_out_lens
def _calc_att_loss(
@@ -424,21 +426,23 @@
else:
# extract fbank feats
time1 = time.perf_counter()
- pdb.set_trace()
- audio_sample_list = load_audio_text_image_video(data_in, fs=frontend.fs, audio_fs=kwargs.get("fs", 16000),
+ sample_list = load_audio_text_image_video(data_in, fs=frontend.fs, audio_fs=kwargs.get("fs", 16000),
data_type=kwargs.get("data_type", "sound"),
tokenizer=tokenizer)
time2 = time.perf_counter()
meta_data["load_data"] = f"{time2 - time1:0.3f}"
- pdb.set_trace()
+ audio_sample_list = sample_list[0]
+ ocr_sample_list = sample_list[1]
speech, speech_lengths = extract_fbank(audio_sample_list, data_type=kwargs.get("data_type", "sound"),
frontend=frontend)
time3 = time.perf_counter()
meta_data["extract_feat"] = f"{time3 - time2:0.3f}"
- meta_data["batch_data_time"] = speech_lengths.sum().item() * frontend.frame_shift * frontend.lfr_n / 1000
+ frame_shift = 10
+ meta_data["batch_data_time"] = speech_lengths.sum().item() * frame_shift / 1000
speech = speech.to(device=kwargs["device"])
speech_lengths = speech_lengths.to(device=kwargs["device"])
+ pdb.set_trace()
# Encoder
encoder_out, encoder_out_lens = self.encode(speech, speech_lengths)
if isinstance(encoder_out, tuple):
--
Gitblit v1.9.1