def main():
import argparse
global model, spect_parser, decoder, args, device
parser = argparse.ArgumentParser(
description='DeepSpeech transcription server')
parser.add_argument('--host',
type=str,
default='0.0.0.0',
help='Host to be used by the server')
parser.add_argument('--port',
type=int,
default=8888,
help='Port to be used by the server')
parser = add_inference_args(parser)
parser = add_decoder_args(parser)
args = parser.parse_args()
logging.getLogger().setLevel(logging.DEBUG)
logging.info('Setting up server...')
torch.set_grad_enabled(False)
device = torch.device("cuda" if args.cuda else "cpu")
model = load_model(device, args.model_path, args.half)
if args.decoder == "beam":
from deepspeech_pytorch.decoder import BeamCTCDecoder
decoder = BeamCTCDecoder(model.labels,
lm_path=args.lm_path,
alpha=args.alpha,
beta=args.beta,
cutoff_top_n=args.cutoff_top_n,
cutoff_prob=args.cutoff_prob,
beam_width=args.beam_width,
num_processes=args.lm_workers)
else:
decoder = GreedyDecoder(model.labels,
blank_index=model.labels.index('_'))
spect_parser = SpectrogramParser(model.audio_conf, normalize=True)
logging.info('Server initialised')
app.run(host=args.host, port=args.port, debug=True, use_reloader=False)
import argparse
import json
import torch
from deepspeech_pytorch.decoder import GreedyDecoder
from deepspeech_pytorch.loader.data_loader import SpectrogramDataset, AudioDataLoader
from deepspeech_pytorch.opts import add_decoder_args, add_inference_args
from deepspeech_pytorch.testing import evaluate
from deepspeech_pytorch.utils import load_model, load_decoder
parser = argparse.ArgumentParser(description='DeepSpeech transcription')
parser = add_inference_args(parser)
parser.add_argument('--test-manifest',
metavar='DIR',
help='path to validation manifest csv',
default='data/test_manifest.csv')
parser.add_argument('--batch-size',
default=5,
type=int,
help='Batch size for testing')
parser.add_argument('--num-workers',
default=8,
type=int,
help='Number of workers used in dataloading')
parser.add_argument('--verbose',
action="store_true",
help="print out decoded output and error of each sample")
parser.add_argument('--save-output',
default=None,
help="Saves output of model from test to this file_path")
spect_parser=spect_parser,
model=model,
decoder=decoder,
device=device,
use_half=args.half)
transcription = decode_results(
decoded_output, decoded_offsets)["output"][0]["transcription"]
print(transcription)
if __name__ == "__main__":
arg_parser = argparse.ArgumentParser(
description='DeepSpeech transcription')
arg_parser = add_inference_args(arg_parser)
arg_parser.add_argument('--offsets',
dest='offsets',
action='store_true',
help='Returns time offset information')
arg_parser = add_decoder_args(arg_parser)
args = arg_parser.parse_args()
device = torch.device("cuda" if args.cuda else "cpu")
model = load_model(device, args.model_path, args.half)
decoder = load_decoder(decoder_type=args.decoder,
labels=model.labels,
lm_path=args.lm_path,
alpha=args.alpha,
beta=args.beta,
cutoff_top_n=args.cutoff_top_n,