From 0e622e694e6cb4459955f1e5942a7c53349ce640 Mon Sep 17 00:00:00 2001
From: 游雁 <zhifu.gzf@alibaba-inc.com>
Date: 星期二, 19 十二月 2023 21:58:14 +0800
Subject: [PATCH] funasr2
---
funasr/frontends/default.py | 17 ++++++++---------
1 files changed, 8 insertions(+), 9 deletions(-)
diff --git a/funasr/models/frontend/default.py b/funasr/frontends/default.py
similarity index 95%
rename from funasr/models/frontend/default.py
rename to funasr/frontends/default.py
index b4e518a..8ac1ca8 100644
--- a/funasr/models/frontend/default.py
+++ b/funasr/frontends/default.py
@@ -6,20 +6,19 @@
import humanfriendly
import numpy as np
import torch
+import torch.nn as nn
try:
from torch_complex.tensor import ComplexTensor
except:
print("Please install torch_complex firstly")
-from funasr.models.frontend.utils.log_mel import LogMel
-from funasr.models.frontend.utils.stft import Stft
-from funasr.models.frontend.abs_frontend import AbsFrontend
-from funasr.models.frontend.utils.frontend import Frontend
-from funasr.utils.get_default_kwargs import get_default_kwargs
+from funasr.frontends.utils.log_mel import LogMel
+from funasr.frontends.utils.stft import Stft
+from funasr.frontends.utils.frontend import Frontend
from funasr.models.transformer.utils.nets_utils import make_pad_mask
-class DefaultFrontend(AbsFrontend):
+class DefaultFrontend(nn.Module):
"""Conventional frontend structure for ASR.
Stft -> WPE -> MVDR-Beamformer -> Power-spec -> Mel-Fbank -> CMVN
"""
@@ -38,7 +37,7 @@
fmin: int = None,
fmax: int = None,
htk: bool = False,
- frontend_conf: Optional[dict] = get_default_kwargs(Frontend),
+ frontend_conf: Optional[dict] = None,
apply_stft: bool = True,
use_channel: int = None,
):
@@ -139,7 +138,7 @@
return input_stft, feats_lens
-class MultiChannelFrontend(AbsFrontend):
+class MultiChannelFrontend(nn.Module):
"""Conventional frontend structure for ASR.
Stft -> WPE -> MVDR-Beamformer -> Power-spec -> Mel-Fbank -> CMVN
"""
@@ -160,7 +159,7 @@
fmin: int = None,
fmax: int = None,
htk: bool = False,
- frontend_conf: Optional[dict] = get_default_kwargs(Frontend),
+ frontend_conf: Optional[dict] = None,
apply_stft: bool = True,
use_channel: int = None,
lfr_m: int = 1,
--
Gitblit v1.9.1