File size: 8,501 Bytes
cd8454d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 | # Copyright (C) 2025. Huawei Technologies Co., Ltd. All Rights Reserved. (authors: Xiao Chen)
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
# http://www.apache.org/licenses/LICENSE-2.0
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from transformers import WhisperProcessor, WhisperForConditionalGeneration
import soundfile as sf
import scipy
import argparse
from whisper_normalizer.english import EnglishTextNormalizer
import os
import string
import lingvo.tasks.asr.tools.simple_wer_v2 as WER
from tqdm import tqdm
import logging
import torch
keyphrases = None
english_normalizer = EnglishTextNormalizer()
device = torch.device("cuda")
en_asr_model_path = "./whisper-large-v3"
wer_obj = WER.SimpleWER(
key_phrases=keyphrases,
html_handler=WER.HighlightAlignedHtmlHandler(WER.HighlightAlignedHtml),
preprocess_handler=WER.RemoveCommentTxtPreprocess,
)
def dummy_split_text(text):
return text
def remove_punct(text):
puncts = set(string.punctuation)
output = ""
for char in text:
if char not in puncts:
output += char
output = output.replace(" ", " ")
return output
def get_gt_ref_texts_and_wav_files(
args, gt_test_lst, gt_folder, punct_remover, text_spliter
):
wav_file_list = []
reference = []
with open(gt_test_lst, "r") as fp:
for line in fp:
fields = line.strip().split("|")
wav_file = f"{gt_folder}/{fields[0]}.wav"
if not os.path.isfile(wav_file):
continue
wav_file_list.append(wav_file)
text = fields[-1].lower()
if args.norm_text:
truth_text = english_normalizer(text) # " ".join(fields[-1])
elif args.remove_punct:
truth_text = punct_remover(text)
else:
truth_text = text
truth_text = text_spliter(truth_text)
reference.append([truth_text, fields[-1]])
assert len(reference) == len(wav_file_list)
return reference, wav_file_list
def get_ref_texts_and_gen_files(
args, test_lst, test_folder, punct_remover, text_spliter
):
reference = []
gen_file_list = []
with open(test_lst, "r") as fp:
for line in fp:
fields = line.strip().split("|")
filename = fields[2].split("/")[-1]
filename = filename.split(".")[0]
gen_file = f"{filename}_gen.wav"
gen_file_list.append(f"{test_folder}/{gen_file}")
text = fields[-1].lower()
if args.norm_text:
truth_text = english_normalizer(text) # " ".join(fields[-1])
elif args.remove_punct:
truth_text = punct_remover(text)
else:
truth_text = text
truth_text = text_spliter(truth_text)
reference.append([truth_text, fields[-1]])
assert len(reference) == len(gen_file_list)
return reference, gen_file_list
def get_hypo_texts(args, results_list, punct_remover, text_spliter):
hypothesis = []
for res in results_list:
text = res["text"].lower()
if args.norm_text:
hypo_text = english_normalizer(text)
elif args.remove_punct:
hypo_text = punct_remover(text)
else:
hypo_text = text
hypo_text = text_spliter(hypo_text)
hypothesis.append([hypo_text, res["text"]])
return hypothesis
def calc_wer(reference, hypothesis, test_lst):
logging.info(f"calc WER:")
for idx in tqdm(range(len(hypothesis))):
hypo = hypothesis[idx][0].strip()
ref = reference[idx][0].strip()
wer_obj.AddHypRef(hypo, ref)
str_summary, str_details, str_keyphrases_info = wer_obj.GetSummaries()
logging.info(f"WER summary:")
logging.info(str_summary)
logging.info(str_details)
logging.info(str_keyphrases_info)
try:
fn_output = test_lst + "_diagnosis.html"
aligned_html = "<br>".join(wer_obj.aligned_htmls)
with open(fn_output, "wt") as fp:
fp.write("<body><html>")
fp.write("<div>%s</div>" % aligned_html)
fp.write("</body></html>")
fp.close()
text_output = test_lst + "_rawtext.lst"
with open(text_output, "w") as fp:
for ref, hypo in zip(reference, hypothesis):
fp.write(f"{ref[1]}|{hypo[1]}\n")
fp.close()
logging.info(f"Save {fn_output} and {text_output} for diagnosis")
except IOError:
logging.info("failed to write diagnosis html")
def load_en_model():
processor = WhisperProcessor.from_pretrained(en_asr_model_path)
model = WhisperForConditionalGeneration.from_pretrained(en_asr_model_path).to(
device
)
return processor, model
def process_wavs(wav_file_list, batch_size=300):
results = []
processor, model = load_en_model()
for wav_file_path in tqdm(wav_file_list):
wav, sr = sf.read(wav_file_path)
if sr != 16000:
wav = scipy.signal.resample(wav, int(len(wav) * 16000 / sr))
input_features = processor(
wav, sampling_rate=16000, return_tensors="pt"
).input_features
input_features = input_features.to(device)
forced_decoder_ids = processor.get_decoder_prompt_ids(
language="english", task="transcribe"
)
predicted_ids = model.generate(
input_features, forced_decoder_ids=forced_decoder_ids
)
transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)[
0
]
results.append({"text": transcription.strip()})
return results
def main(args):
handler = logging.FileHandler(filename=args.log_file, mode="w")
logging.root.setLevel(logging.INFO)
logging.root.addHandler(handler)
test_path = (
args.test_path
) # './40ms.AISHELL2.test_with_single_ref.base.chunk25.gen'
lst_path = args.test_lst # "40ms.AISHELL2.test_with_single_ref.base.lst"
logging.info(
f"Evaluate {args.test_path} with Text Normalization: {args.norm_text} and Remove Punct: {args.remove_punct}"
)
if args.eval_gt:
logging.info(f"run ASR for GT: {lst_path}")
reference, wav_file_list = get_gt_ref_texts_and_wav_files(
args, lst_path, test_path, remove_punct, dummy_split_text
)
results = process_wavs(wav_file_list, batch_size=12)
else:
logging.info(f"run ASR for detok: {lst_path}")
reference, gen_file_list = get_ref_texts_and_gen_files(
args, lst_path, test_path, remove_punct, dummy_split_text
)
results = process_wavs(gen_file_list, batch_size=12)
hypothesis = get_hypo_texts(args, results, remove_punct, dummy_split_text)
assert len(hypothesis) == len(reference)
logging.info(f"Finish runing ASR for {lst_path}")
logging.info(f"hypothesis: {len(hypothesis)} vs reference: {len(reference)}")
calc_wer(reference, hypothesis, test_path)
logging.info(f"Finish evaluate {lst_path}, results are in {args.log_file}")
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument(
"--test-path",
required=True,
type=str,
help=f"folder of wav files",
)
parser.add_argument(
"--test-lst",
required=True,
type=str,
help=f"path to test file lst",
)
parser.add_argument(
"--log-file",
required=False,
type=str,
default=None,
help=f"path to test file lst",
)
parser.add_argument(
"--norm-text",
default=False,
action="store_true",
help=f"normalized GT and hypo texts",
)
parser.add_argument(
"--remove-punct",
default=False,
action="store_true",
help=f"remove punct from GT and hypo texts",
)
parser.add_argument(
"--eval-gt",
default=False,
action="store_true",
help=f"remove punct from GT and hypo texts",
)
args = parser.parse_args()
main(args)
|