Spaces:
Sleeping
Sleeping
modified getting pinyin method
Browse files
server.py
CHANGED
@@ -12,6 +12,7 @@ import jiwer
|
|
12 |
import librosa
|
13 |
from svs_utils import load_song_database, estimate_sentence_length
|
14 |
from svs_eval import singmos_warmup, singmos_evaluation
|
|
|
15 |
|
16 |
|
17 |
asr_pipeline = pipeline(
|
@@ -143,12 +144,14 @@ def on_click_metrics():
|
|
143 |
# OWSM ctc + PER
|
144 |
y, sr = librosa.load("tmp/response.wav", sr=16000)
|
145 |
asr_result = asr_pipeline(y, generate_kwargs={"language": "mandarin"} )['text']
|
146 |
-
hyp_pinin = lazy_pinyin(asr_result)
|
|
|
147 |
|
148 |
with open(f"tmp/llm.txt", "r") as f:
|
149 |
ref = f.read().replace(' ', '')
|
150 |
|
151 |
-
ref_pinin = lazy_pinyin(ref)
|
|
|
152 |
per = jiwer.wer(" ".join(ref_pinin), " ".join(hyp_pinin))
|
153 |
|
154 |
audio = librosa.load(f"tmp/response.wav", sr=sample_rate)[0]
|
|
|
12 |
import librosa
|
13 |
from svs_utils import load_song_database, estimate_sentence_length
|
14 |
from svs_eval import singmos_warmup, singmos_evaluation
|
15 |
+
from util import get_pinyin
|
16 |
|
17 |
|
18 |
asr_pipeline = pipeline(
|
|
|
144 |
# OWSM ctc + PER
|
145 |
y, sr = librosa.load("tmp/response.wav", sr=16000)
|
146 |
asr_result = asr_pipeline(y, generate_kwargs={"language": "mandarin"} )['text']
|
147 |
+
# hyp_pinin = lazy_pinyin(asr_result)
|
148 |
+
hyp_pinin = get_pinyin(asr_result)
|
149 |
|
150 |
with open(f"tmp/llm.txt", "r") as f:
|
151 |
ref = f.read().replace(' ', '')
|
152 |
|
153 |
+
# ref_pinin = lazy_pinyin(ref)
|
154 |
+
ref_pinin = get_pinyin(ref)
|
155 |
per = jiwer.wer(" ".join(ref_pinin), " ".join(hyp_pinin))
|
156 |
|
157 |
audio = librosa.load(f"tmp/response.wav", sr=sample_rate)[0]
|