File size: 7,412 Bytes
ade4f6a e4406a3 ade4f6a 8a34d65 e4406a3 8a3bc32 e4406a3 8a3bc32 778443c 8a3bc32 e4406a3 8a3bc32 e4406a3 8a3bc32 42742c6 e4406a3 8a3bc32 3d1d87d 8a3bc32 79f9224 42742c6 79f9224 42742c6 778443c 79f9224 8a34d65 42742c6 8a34d65 778443c 8a34d65 42742c6 e4406a3 42742c6 |
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 |
from pywhispercpp.model import Model
from pathlib import Path
import time
import csv
from silero_vad.utils_vad import languages
from scripts.asr_utils import get_origin_text_dict, get_text_distance
def save_csv(file_path, rows):
with open(file_path, "w", encoding="utf-8") as f:
writer = csv.writer(f)
writer.writerows(rows)
print(f"write csv to {file_path}")
def load_model():
models_dir = Path("/Users/jeqin/work/code/Translator/python_server/moyoyo_asr_models")
whisper_model = 'large-v3-turbo-q5_0'
t0 = time.time()
model = Model(
model=whisper_model,
models_dir=models_dir,
print_realtime=False,
print_progress=False,
print_timestamps=False,
translate=False,
# beam_search=1,
temperature=0.,
no_context=True
)
print("load model time: ", time.time()-t0)
return model
def run_recordings():
model = load_model()
audios = Path("../test_data/recordings/")
rows = [["file_name", "time", "inference_result"]]
original = get_origin_text_dict()
for audio in sorted(audios.glob("*.wav"), key=lambda x: int(x.stem)):
print(audio)
t1 = time.time()
output = model.transcribe(str(audio), language="zh", initial_prompt="以下是普通话句子，这是一段会议内容。")# initial_prompt="这是一段中文的会议内容。"
t = time.time() - t1
print("inference time:", t)
text = " ".join([a.text for a in output])
print(text)
d, nd, diff = get_text_distance(original[audio.stem], text)
rows.append([audio.name, round(t, 3), text, d, round(nd,3), diff])
save_csv("csv/pywhisper.csv", rows)
def run_test_audios():
model = load_model()
lang = "zh"
audios = Path("../test_data/audio_clips/")
rows = [["file_name", "time", "inference_result"]]
for audio in sorted(audios.glob(f"*{lang}*/*.wav")):
print(audio)
t1 = time.time()
output = model.transcribe(str(audio), language=lang, initial_prompt="以下是普通话句子，这是一段会议内容。")# initial_prompt="这是一段中文的会议内容。"
t = time.time() - t1
print("inference time:", t)
text = " ".join([a.text for a in output])
print(text)
rows.append([f"{audio.parent.name}/{audio.name}", round(t, 3), text])
save_csv("csv/whisper.csv", rows)
def run_test_dataset():
from test_data.audios import read_dataset
model = load_model()
test_data = Path("../test_data/dataset/dataset.txt")
audio_parent = Path("../test_data/")
rows = [["file_name", "time", "inference_result"]]
result_list = []
count = 0
try:
for audio_path, sentence, duration in read_dataset(test_data):
count += 1
print(f"processing {count}: {audio_path}")
t1 = time.time()
output = model.transcribe(str(audio_parent/audio_path), language="zh")# , initial_prompt="以下是普通话句子，这是一段会议内容。"
t = time.time() - t1
print("inference time:", t)
text = " ".join([a.text for a in output])
print(text)
result_list.append({
"index": count,
"audio_path": audio_path,
"reference": sentence,
"duration": duration,
"inference_time": round(t, 3),
"inference_result": text
})
except Exception as e:
print(e)
except KeyboardInterrupt as e:
print(e)
import json
with open("csv/whisper_dataset_results.json", "w", encoding="utf-8") as f:
json.dump(result_list, f, ensure_ascii=False, indent=2)
def run_test_emilia():
from test_data.audios import read_emilia
model = load_model()
parent = Path("../test_data/ZH-B000000")
result_list = []
count = 0
try:
for audio_path, sentence, duration in read_emilia(parent, count_limit=5000):
count += 1
print(f"processing {count}: {audio_path.name}")
t1 = time.time()
output = model.transcribe(str(audio_path), language="zh")# , initial_prompt="以下是普通话句子，这是一段会议内容。"
t = time.time() - t1
print("inference time:", t)
text = " ".join([a.text for a in output])
print(text)
result_list.append({
"index": count,
"audio_path": audio_path.name,
"reference": sentence,
"duration": duration,
"inference_time": round(t, 3),
"inference_result": text
})
except Exception as e:
print(e)
except KeyboardInterrupt as e:
print(e)
import json
with open("csv/whisper_emilia_results.json", "w", encoding="utf-8") as f:
json.dump(result_list, f, ensure_ascii=False, indent=2)
def run_test_st():
from test_data.audios import read_st
model = load_model()
# parent = Path("../test_data/ST-CMDS-20170001_1-OS")
result_list = []
count = 0
try:
for audio_path, sentence in read_st(count_limit=5000):
count += 1
print(f"processing {count}: {audio_path}")
t1 = time.time()
output = model.transcribe(
str(audio_path), language="zh"
)
t = time.time() - t1
print("inference time:", t)
text = " ".join([a.text for a in output])
print(text)
result_list.append({
"index": count,
"audio_path": audio_path.name,
"reference": sentence,
# "duration": duration,
"inference_time": round(t, 3),
"inference_result": text
})
except Exception as e:
print(e)
except KeyboardInterrupt as e:
print(e)
import json
with open("csv/whisper_st_results.json", "w", encoding="utf-8") as f:
json.dump(result_list, f, ensure_ascii=False, indent=2)
def run_test_wenet():
from test_data.audios import read_wenet
model = load_model()
result_list = []
count = 0
try:
for audio_path, sentence in read_wenet(count_limit=5000):
count += 1
print(f"processing {count}: {audio_path}")
t1 = time.time()
output = model.transcribe(
str(audio_path), language="zh"
)
t = time.time() - t1
print("inference time:", t)
text = " ".join([a.text for a in output])
print(text)
result_list.append({
"index": count,
"audio_path": audio_path.name,
"reference": sentence,
# "duration": duration,
"inference_time": round(t, 3),
"inference_result": text
})
except Exception as e:
print(e)
except KeyboardInterrupt as e:
print(e)
import json
with open("csv/whisper_wenet_results.json", "w", encoding="utf-8") as f:
json.dump(result_list, f, ensure_ascii=False, indent=2)
if __name__ == '__main__':
# run_test_emilia()
# run_recordings()
run_test_wenet() |