speech-test commited on
Commit
5bee55f
1 Parent(s): ce0b25a

Fix eval script

Browse files
Files changed (1) hide show
  1. README.md +10 -9
README.md CHANGED
@@ -82,37 +82,38 @@ from tqdm.auto import tqdm
82
  from datasets import load_metric
83
  from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
84
 
85
- # Download the raw data instead of using HF datasets to save space
86
- data_url = "https://voice-prod-bundler-ee1969a6ce8178826482b88e843c335139bd3fb4.s3.amazonaws.com/cv-corpus-6.1-2020-12-11/sah.tar.gz"
87
  filestream = urllib.request.urlopen(data_url)
88
  data_file = tarfile.open(fileobj=filestream, mode="r|gz")
89
  data_file.extractall()
90
 
91
  wer = load_metric("wer")
92
 
93
- processor = Wav2Vec2Processor.from_pretrained("anton-l/wav2vec2-large-xlsr-53-sakha")
94
- model = Wav2Vec2ForCTC.from_pretrained("anton-l/wav2vec2-large-xlsr-53-sakha")
95
  model.to("cuda")
96
 
97
- cv_test = pd.read_csv("cv-corpus-6.1-2020-12-11/sah/test.tsv", sep='\\t')
98
- clips_path = "cv-corpus-6.1-2020-12-11/sah/clips/"
99
 
100
  def clean_sentence(sent):
101
  sent = sent.lower()
 
 
102
  # replace non-alpha characters with space
103
- sent = "".join(ch if ch.isalpha() else " " for ch in sent)
104
  # remove repeated spaces
105
  sent = " ".join(sent.split())
106
  return sent
107
 
108
- resampler = torchaudio.transforms.Resample(48_000, 16_000)
109
-
110
  targets = []
111
  preds = []
112
 
113
  for i, row in tqdm(cv_test.iterrows(), total=cv_test.shape[0]):
114
  row["sentence"] = clean_sentence(row["sentence"])
115
  speech_array, sampling_rate = torchaudio.load(clips_path + row["path"])
 
116
  row["speech"] = resampler(speech_array).squeeze().numpy()
117
 
118
  inputs = processor(row["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
82
  from datasets import load_metric
83
  from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
84
 
85
+ # Download the raw data instead of using HF datasets to save disk space
86
+ data_url = "https://voice-prod-bundler-ee1969a6ce8178826482b88e843c335139bd3fb4.s3.amazonaws.com/cv-corpus-6.1-2020-12-11/lt.tar.gz"
87
  filestream = urllib.request.urlopen(data_url)
88
  data_file = tarfile.open(fileobj=filestream, mode="r|gz")
89
  data_file.extractall()
90
 
91
  wer = load_metric("wer")
92
 
93
+ processor = Wav2Vec2Processor.from_pretrained("anton-l/wav2vec2-large-xlsr-53-lithuanian")
94
+ model = Wav2Vec2ForCTC.from_pretrained("anton-l/wav2vec2-large-xlsr-53-lithuanian")
95
  model.to("cuda")
96
 
97
+ cv_test = pd.read_csv("cv-corpus-6.1-2020-12-11/lt/test.tsv", sep='\t')
98
+ clips_path = "cv-corpus-6.1-2020-12-11/lt/clips/"
99
 
100
  def clean_sentence(sent):
101
  sent = sent.lower()
102
+ # normalize apostrophes
103
+ sent = sent.replace("’", "'")
104
  # replace non-alpha characters with space
105
+ sent = "".join(ch if ch.isalpha() or ch == "'" else " " for ch in sent)
106
  # remove repeated spaces
107
  sent = " ".join(sent.split())
108
  return sent
109
 
 
 
110
  targets = []
111
  preds = []
112
 
113
  for i, row in tqdm(cv_test.iterrows(), total=cv_test.shape[0]):
114
  row["sentence"] = clean_sentence(row["sentence"])
115
  speech_array, sampling_rate = torchaudio.load(clips_path + row["path"])
116
+ resampler = torchaudio.transforms.Resample(sampling_rate, 16_000)
117
  row["speech"] = resampler(speech_array).squeeze().numpy()
118
 
119
  inputs = processor(row["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)