From 055406a67e48b7df854c0979985806119ac96d46 Mon Sep 17 00:00:00 2001 From: Fangjun Kuang Date: Wed, 23 Mar 2022 17:35:39 +0800 Subject: [PATCH] Minor fixes. --- .../ASR/transducer_stateless_modified-2/pretrained.py | 6 +++--- egs/aishell/ASR/transducer_stateless_modified/pretrained.py | 6 +++--- egs/librispeech/ASR/transducer_stateless/pretrained.py | 3 --- .../ASR/transducer_stateless_multi_datasets/pretrained.py | 3 --- 4 files changed, 6 insertions(+), 12 deletions(-) diff --git a/egs/aishell/ASR/transducer_stateless_modified-2/pretrained.py b/egs/aishell/ASR/transducer_stateless_modified-2/pretrained.py index 05489beba..9e6ed96b1 100755 --- a/egs/aishell/ASR/transducer_stateless_modified-2/pretrained.py +++ b/egs/aishell/ASR/transducer_stateless_modified-2/pretrained.py @@ -55,7 +55,6 @@ from typing import List import kaldifeat import torch -import torch.nn as nn import torchaudio from beam_search import ( beam_search, @@ -67,7 +66,6 @@ from torch.nn.utils.rnn import pad_sequence from train import get_params, get_transducer_model from icefall.lexicon import Lexicon -from icefall.utils import AttributeDict def get_parser(): @@ -262,7 +260,9 @@ def main(): ) hyp_list.append(hyp) - hyps = [sp.decode(hyp).split() for hyp in hyp_list] + hyps = [] + for hyp in hyp_list: + hyps.append([lexicon.token_table[i] for i in hyp]) s = "\n" for filename, hyp in zip(params.sound_files, hyps): diff --git a/egs/aishell/ASR/transducer_stateless_modified/pretrained.py b/egs/aishell/ASR/transducer_stateless_modified/pretrained.py index 4f7da3bde..f7c5b24ba 100755 --- a/egs/aishell/ASR/transducer_stateless_modified/pretrained.py +++ b/egs/aishell/ASR/transducer_stateless_modified/pretrained.py @@ -55,7 +55,6 @@ from typing import List import kaldifeat import torch -import torch.nn as nn import torchaudio from beam_search import ( beam_search, @@ -67,7 +66,6 @@ from torch.nn.utils.rnn import pad_sequence from train import get_params, get_transducer_model from icefall.lexicon import Lexicon -from icefall.utils import AttributeDict def get_parser(): @@ -262,7 +260,9 @@ def main(): ) hyp_list.append(hyp) - hyps = [sp.decode(hyp).split() for hyp in hyp_list] + hyps = [] + for hyp in hyp_list: + hyps.append([lexicon.token_table[i] for i in hyp]) s = "\n" for filename, hyp in zip(params.sound_files, hyps): diff --git a/egs/librispeech/ASR/transducer_stateless/pretrained.py b/egs/librispeech/ASR/transducer_stateless/pretrained.py index 23ccdf6ad..4fb5d92c5 100755 --- a/egs/librispeech/ASR/transducer_stateless/pretrained.py +++ b/egs/librispeech/ASR/transducer_stateless/pretrained.py @@ -59,7 +59,6 @@ from typing import List import kaldifeat import sentencepiece as spm import torch -import torch.nn as nn import torchaudio from beam_search import ( beam_search, @@ -70,8 +69,6 @@ from beam_search import ( from torch.nn.utils.rnn import pad_sequence from train import get_params, get_transducer_model -from icefall.utils import AttributeDict - def get_parser(): parser = argparse.ArgumentParser( diff --git a/egs/librispeech/ASR/transducer_stateless_multi_datasets/pretrained.py b/egs/librispeech/ASR/transducer_stateless_multi_datasets/pretrained.py index bd84d8d36..df9c3186f 100755 --- a/egs/librispeech/ASR/transducer_stateless_multi_datasets/pretrained.py +++ b/egs/librispeech/ASR/transducer_stateless_multi_datasets/pretrained.py @@ -59,7 +59,6 @@ from typing import List import kaldifeat import sentencepiece as spm import torch -import torch.nn as nn import torchaudio from beam_search import ( beam_search, @@ -70,8 +69,6 @@ from beam_search import ( from torch.nn.utils.rnn import pad_sequence from train import get_params, get_transducer_model -from icefall.utils import AttributeDict - def get_parser(): parser = argparse.ArgumentParser(