Spaces:
Running
Running
File size: 5,138 Bytes
297e244 |
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 |
import os
import tempfile
import re
import librosa
import torch
import json
import numpy as np
from transformers import Wav2Vec2ForCTC, AutoProcessor
from huggingface_hub import hf_hub_download
from torchaudio.models.decoder import ctc_decoder
uroman_dir = "uroman"
assert os.path.exists(uroman_dir)
UROMAN_PL = os.path.join(uroman_dir, "bin", "uroman.pl")
ASR_SAMPLING_RATE = 16_000
MODEL_ID = "facebook/mms-1b-all"
processor = AutoProcessor.from_pretrained(MODEL_ID)
model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
lm_decoding_config = {}
lm_decoding_configfile = hf_hub_download(
repo_id="facebook/mms-cclms",
filename="decoding_config.json",
subfolder="mms-1b-all",
)
with open(lm_decoding_configfile) as f:
lm_decoding_config = json.loads(f.read())
decoding_config = lm_decoding_config["eng"]
lm_file = hf_hub_download(
repo_id="facebook/mms-cclms",
filename=decoding_config["lmfile"].rsplit("/", 1)[1],
subfolder=decoding_config["lmfile"].rsplit("/", 1)[0],
)
token_file = hf_hub_download(
repo_id="facebook/mms-cclms",
filename=decoding_config["tokensfile"].rsplit("/", 1)[1],
subfolder=decoding_config["tokensfile"].rsplit("/", 1)[0],
)
def error_check_file(filepath):
if not isinstance(filepath, str):
return "Expected file to be of type 'str'. Instead got {}".format(
type(filepath)
)
if not os.path.exists(filepath):
return "Input file '{}' doesn't exists".format(type(filepath))
def norm_uroman(text):
text = text.lower()
text = text.replace("’", "'")
text = re.sub("([^a-z' ])", " ", text)
text = re.sub(' +', ' ', text)
return text.strip()
def uromanize(words):
iso = "xxx"
with tempfile.NamedTemporaryFile() as tf, tempfile.NamedTemporaryFile() as tf2:
with open(tf.name, "w") as f:
f.write("\n".join(words))
cmd = f"perl " + UROMAN_PL
cmd += f" -l {iso} "
cmd += f" < {tf.name} > {tf2.name}"
os.system(cmd)
lexicon = {}
with open(tf2.name) as f:
for idx, line in enumerate(f):
line = re.sub(r"\s+", " ", norm_uroman(line)).strip()
lexicon[words[idx]] = " ".join(line) + " |"
return lexicon
def load_lexicon(filepath):
words = []
with open(filepath) as f:
for line in f:
line = line.strip()
# ignore invalid words.
if not line or " " in line or len(line) > 50:
continue
words.append(line)
return uromanize(words)
def process(audio_data, words_file, lm_path=None):
if isinstance(audio_data, tuple):
# microphone
sr, audio_samples = audio_data
audio_samples = (audio_samples / 32768.0).astype(np.float)
assert sr == ASR_SAMPLING_RATE, "Invalid sampling rate"
else:
# file upload
assert isinstance(audio_data, str)
audio_samples = librosa.load(audio_data, sr=ASR_SAMPLING_RATE, mono=True)[0]
# print(audio_samples[:10])
# print("I'm here 102")
# print("len audio_samples", len(audio_samples))
lang_code = "eng"
processor.tokenizer.set_target_lang(lang_code)
# print("I'm here 107")
model.load_adapter(lang_code)
# print("I'm here 109")
inputs = processor(
audio_samples, sampling_rate=ASR_SAMPLING_RATE, return_tensors="pt"
)
# print("I'm here 106")
# set device
if torch.cuda.is_available():
device = torch.device("cuda")
elif (
hasattr(torch.backends, "mps")
and torch.backends.mps.is_available()
and torch.backends.mps.is_built()
):
device = torch.device("mps")
else:
device = torch.device("cpu")
model.to(device)
inputs = inputs.to(device)
# print("I'm here 122")
with torch.no_grad():
outputs = model(**inputs).logits
# Setup lexicon and decoder
# print("before uroman")
lexicon = load_lexicon(words_file)
# print("after uroman")
# print("len lexicon", len(lexicon))
with tempfile.NamedTemporaryFile() as lexicon_file:
with open(lexicon_file.name, "w") as f:
idx = 10
for word, spelling in lexicon.items():
f.write(word + " " + spelling + "\n")
if idx%100 == 0:
print(word, spelling, flush=True)
idx+=1
beam_search_decoder = ctc_decoder(
lexicon=lexicon_file.name,
tokens=token_file,
lm=None,
nbest=1,
beam_size=500,
beam_size_token=50,
lm_weight=float(decoding_config["lmweight"]),
word_score=float(decoding_config["wordscore"]),
sil_score=float(decoding_config["silweight"]),
blank_token="<s>",
)
beam_search_result = beam_search_decoder(outputs.to("cpu"))
transcription = " ".join(beam_search_result[0][0].words).strip()
return transcription
ZS_EXAMPLES = [
["upload/english.mp3", "upload/words_top10k.txt"]
]
# print(process("upload/english.mp3", "upload/words_top10k.txt")) |