def setUp(self):
self.recogniser = DummyRecogniser()
self.asr = ASR(self.recogniser, dummy_lattice_to_nbest)
import requests
from flask import Flask, request
from asr import ASR
app = Flask(__name__)
# ==============================================================================
# Parameters
# ==============================================================================
FB_API_URL = 'https://graph.facebook.com/v5.0/me/messages'
VERIFY_TOKEN = '7F151hSt4EBuUSxfc+OxEdNRR20ZCfkVe6i3ywb2ZUY='
PAGE_ACCESS_TOKEN = 'EAAvvvwDxUFABANyZBdYXUjeg6TRCLzgHJsctZCAq9xVRVrtZCgBbsbUPTZB9CnCJSmKBW1nqlyWBO1ZBFmAOME85iFr7pPeHGkm8oHHCBn9EfLXnxZCXVz1JKJnWywdW00aqgtAUa1s5AeeIrOmR3OGsZBt0xXMPvjlKfUQuhCxygZDZD'
asr_model = ASR()
# ==============================================================================
# Messager
# ==============================================================================
def triggers(text):
text = ' ' + text.lower() + ' '
if ' no ' in text:
response = ' -- Can you please take your medication'
elif ' not yet ' in text:
response = ' -- Can you please take your medication'
def setUp(self):
self.recogniser = DummyRecogniser()
self.asr = ASR(self.recogniser, dummy_to_nbest, dummy_to_best_path)
# ********* EXAMPLE SCENARIO *************** S0 = 45 #------->EUR sigma = 0.6 #--->EUR.day*-1/2 which corresponds to an annual volatility approximately equal to 21%. T = 63 #-------->trading days N = [22, 62] #-->The set of possible dates for delivery before expiry V = 4000000 #--->stocks day-1 Q = 20000000 #-->stocks eta = 0.1 #----->EUR stock-1 day-1 fi = 0.75 gamma = 2.5e-7 #>EUR-1 NQ = 10 # the computational grid for q INF = 1e9 # value for the infinity scenario = ASR(S0, sigma, T, N, V, Q, eta, fi, gamma) scenario.initialize(NQ, INF) # uncomment 2 of the 3 following lines to calculate and save TETAs # - use 'save_TETAs()' to save results to a text file # - use 'save_gzip_TETAs()' to save results to a gzip file scenario.get_TETAs() #scenario.save_TETAs() scenario.save_gzip_TETAs() # uncoment the 2 of the 4 following lines to read TETAs from a file: # - use 'read_TETAs' to read from a text file # - use 'read_gzip_TETAs' to read from a gzip file #filename = 'teta_qgrid_50_gamma_2.5e-07.txt' # define a filename to read TETAs #scenario.read_TETAs(filename) #filename = 'teta_qgrid_50_gamma_2.5e-07.gzip' # define a filename to read TETAs
def fromScratch(cls):
self = cls()
self.asr = ASR()
self.sentMatcher = SentMatcher(corpus="../quran-simple-clean.txt")
def main(config):
### create Experiment Directory ###
# combine all hyperparameters into a single file
hparams = load_hparams(config.exp_config)
hparams["model_config"] = load_hparams(config.model_config)
# create exp dir
sb.create_experiment_directory(experiment_directory=config.output_folder,
hyperparams_to_save=config.exp_config,
overrides=None)
### Datasets and Tokenizer ###
train_data, valid_data, test_data, tokenizer = dataio_prepare(hparams)
# Trainer initialization
run_opts = {
"device": "cuda:0"
} # certain args from yaml file will autoamtically get picked as run_opts
# see https://github.com/speechbrain/speechbrain/blob/develop/recipes/LibriSpeech/ASR/transformer/train.py#L372
# see https://github.com/speechbrain/speechbrain/blob/d6adc40e742107c34ae38dc63484171938b4d237/speechbrain/core.py#L124
#print(type(hparams["model_config"]["modules"]))
#print(type(hparams))
#exit()
asr_brain = ASR(
modules=hparams["model_config"]["modules"],
opt_class=hparams["model_config"]["Adam"],
hparams=hparams["model_config"],
run_opts=run_opts,
checkpointer=hparams["model_config"]["checkpointer"],
)
# adding objects to trainer:
asr_brain.tokenizer = tokenizer # hparams["tokenizer"]
# Training
asr_brain.fit(
asr_brain.hparams.epoch_counter,
train_data,
valid_data,
train_loader_kwargs=hparams["model_config"]["train_dataloader_opts"],
valid_loader_kwargs=hparams["model_config"]["valid_dataloader_opts"],
)
raise NotImplementedError
### get Train Data ###
# list of {'audio__file': str, 'transcript_all_file': str, 'transcript_uid': str, 'filter_criteria': str}
# meaning that <audio__file>'s transcript is the one in the <transcript_all_file> with id <transcript_uid>
train_corpus = get_utterance_manifest_from_data_config(
config.train_data_config)
for x in train_corpus:
assert os.path.exists(
x["transcript_all_file"]
), "data transcript file {} does not exist! Exiting!".format(
x["transcript_all_file"])
### create json file for SpeechBrain-->SentencePiece ###
selected_transcripts_json, annotation_read = create_transcripts_json(
train_corpus)
### train custom SentencePiece Tokenizer ###
with tempfile.NamedTemporaryFile(mode="w+", suffix=".json") as f:
f.write(json.dumps(selected_transcripts_json))
f.seek(0)
SentencePiece(model_dir=config.output_folder,
vocab_size=config.vocab_size,
annotation_train=f.name,
annotation_read=annotation_read,
annotation_format="json",
model_type=config.model_type,
character_coverage=config.character_coverage,
annotation_list_to_check=config.annotation_list_to_check)
default=DEFAULT_MODEL,
help="kaldi model, default: %s" % DEFAULT_MODEL)
(options, args) = parser.parse_args()
if options.verbose:
logging.basicConfig(level=logging.DEBUG)
else:
logging.basicConfig(level=logging.INFO)
kaldi_model_dir = options.model_dir
kaldi_model = options.model
asr = ASR(engine=ASR_ENGINE_NNET3,
model_dir=kaldi_model_dir,
kaldi_beam=DEFAULT_BEAM,
kaldi_acoustic_scale=DEFAULT_ACOUSTIC_SCALE,
kaldi_frame_subsampling_factor=DEFAULT_FRAME_SUBSAMPLING_FACTOR)
else:
logging.basicConfig(level=logging.INFO)
kaldi_model_dir = DEFAULT_MODEL_DIR
kaldi_model = DEFAULT_MODEL
asr = ASR(engine=ASR_ENGINE_NNET3,
model_dir=DEFAULT_MODEL_DIR,
kaldi_beam=DEFAULT_BEAM,
kaldi_acoustic_scale=DEFAULT_ACOUSTIC_SCALE,
kaldi_frame_subsampling_factor=DEFAULT_FRAME_SUBSAMPLING_FACTOR)
out = asr_model.transcribe(wav)
toc = time.time()
app.logger.info("ASR Model Transcription: "+out)
app.logger.info("ASR Duration: {} seconds".format(toc-tic))
# form response
flask_response= app.response_class(response=flask.json.dumps({"text": out}),
status=200,
mimetype='application/json' )
return flask_response
if __name__ == '__main__':
conf = parse_yaml("conf.yaml")
# load ASR model
app.logger.info("===== Loading ASR Model =====")
asr_conf = conf["asr"]
asr_model = ASR(asr_conf)
# load TTS model
app.logger.info("===== Loading TTS Model =====")
tts_conf = conf["tts"]
tts_model = TTS(tts_conf)
# run server
app.logger.info("===== Running the Server =====")
app.run(host="0.0.0.0", port=5000)
