From 3958472fb2a9bbac5cb2a30e3fb11925c7b5d3d8 Mon Sep 17 00:00:00 2001
From: Yabin Li <wucong.lyb@alibaba-inc.com>
Date: 星期三, 20 三月 2024 19:21:33 +0800
Subject: [PATCH] Update SDK_advanced_guide_offline_en_zh.md
---
funasr/models/monotonic_aligner/model.py | 33 ++++++++++++++++++++-------------
1 files changed, 20 insertions(+), 13 deletions(-)
diff --git a/funasr/models/monotonic_aligner/model.py b/funasr/models/monotonic_aligner/model.py
index 1b43c2f..718923b 100644
--- a/funasr/models/monotonic_aligner/model.py
+++ b/funasr/models/monotonic_aligner/model.py
@@ -1,22 +1,27 @@
+#!/usr/bin/env python3
+# -*- encoding: utf-8 -*-
+# Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
+# MIT License (https://opensource.org/licenses/MIT)
+
import time
import copy
import torch
from torch.cuda.amp import autocast
from typing import Union, Dict, List, Tuple, Optional
+from funasr.register import tables
+from funasr.models.ctc.ctc import CTC
+from funasr.utils import postprocess_utils
+from funasr.utils.datadir_writer import DatadirWriter
from funasr.models.paraformer.cif_predictor import mae_loss
from funasr.train_utils.device_funcs import force_gatherable
from funasr.models.transformer.utils.add_sos_eos import add_sos_eos
from funasr.models.transformer.utils.nets_utils import make_pad_mask
from funasr.utils.timestamp_tools import ts_prediction_lfr6_standard
-from funasr.utils import postprocess_utils
-from funasr.utils.datadir_writer import DatadirWriter
-from funasr.register import tables
-from funasr.models.ctc.ctc import CTC
from funasr.utils.load_utils import load_audio_text_image_video, extract_fbank
-@tables.register("model_classes", "monotonicaligner")
+@tables.register("model_classes", "MonotonicAligner")
class MonotonicAligner(torch.nn.Module):
"""
Author: Speech Lab of DAMO Academy, Alibaba Group
@@ -41,15 +46,15 @@
super().__init__()
if specaug is not None:
- specaug_class = tables.specaug_classes.get(specaug.lower())
+ specaug_class = tables.specaug_classes.get(specaug)
specaug = specaug_class(**specaug_conf)
if normalize is not None:
- normalize_class = tables.normalize_classes.get(normalize.lower())
+ normalize_class = tables.normalize_classes.get(normalize)
normalize = normalize_class(**normalize_conf)
- encoder_class = tables.encoder_classes.get(encoder.lower())
+ encoder_class = tables.encoder_classes.get(encoder)
encoder = encoder_class(input_size=input_size, **encoder_conf)
encoder_output_size = encoder.output_size()
- predictor_class = tables.predictor_classes.get(predictor.lower())
+ predictor_class = tables.predictor_classes.get(predictor)
predictor = predictor_class(**predictor_conf)
self.specaug = specaug
self.normalize = normalize
@@ -143,7 +148,7 @@
return encoder_out, encoder_out_lens
- def generate(self,
+ def inference(self,
data_in,
data_lengths=None,
key: list=None,
@@ -178,9 +183,11 @@
results = []
ibest_writer = None
- if ibest_writer is None and kwargs.get("output_dir") is not None:
- writer = DatadirWriter(kwargs.get("output_dir"))
- ibest_writer = writer["tp_res"]
+ if kwargs.get("output_dir") is not None:
+ if not hasattr(self, "writer"):
+ self.writer = DatadirWriter(kwargs.get("output_dir"))
+ ibest_writer = self.writer["tp_res"]
+
for i, (us_alpha, us_peak, token_int) in enumerate(zip(us_alphas, us_peaks, text_token_int_list)):
token = tokenizer.ids2tokens(token_int)
timestamp_str, timestamp = ts_prediction_lfr6_standard(us_alpha[:encoder_out_lens[i] * 3],
--
Gitblit v1.9.1