ShigekiKarita

34c34
< data_url=www.openslr.org/resources/60
---
> data_url=www.openslr.org/resources/141
69c69
<         cp ${db_root}/LibriTTS/SPEAKERS.txt data/local
---
>         cp ${db_root}/LibriTTS/train-clean-100/SPEAKERS.txt data/local

ShigekiKarita

ShigekiKarita

git clone https://github.com/espnet/espnet
cd espnet
git checkout v.0.9.7

# Prepare TTS input
echo "HELLO WORLD" > egs/ljspeech/tts1/text

# Run TTS in docker. If you have GPUs, use "espnet/espnet:gpu-cuda10.1-cudnn7-u18"
# https://hub.docker.com/r/espnet/espnet/tags
docker run --rm -v $(pwd)/utils:/espnet/utils -v $(pwd)/espnet:/espnet/espnet -v $(pwd)/egs:/espnet/egs \

ShigekiKarita

# Count the number of parameters in the saved model.
import torch

# Run this commands to get this model.
# $ cd $ESPNET_ROOT/egs/mini_an4/asr1; ./run.sh
state = torch.load("./exp/train_nodev_pytorch_train/results/model.acc.best")
print(sum(v.numel() for v in state.values()))

ShigekiKarita

210c210
<     print B "( " . $cmd . ") 2>>$logfile >> $logfile";
---
>     print B "( " . $cmd . ") |& tee -a $logfile";

ShigekiKarita

ShigekiKarita

#!/bin/bash

# Copyright 2019 Nagoya University (Masao Someki)
# Copyright 2019 Shigeki Karita
#  Apache 2.0  (http://www.apache.org/licenses/LICENSE-2.0)

. ./path.sh || exit 1;
. ./cmd.sh || exit 1;

# general configuration

ShigekiKarita

import logging
from typing import Dict
from typing import NamedTuple
from typing import Tuple

import torch

from espnet.nets.e2e_asr_common import end_detect
from espnet.nets.scorer_interface import PartialScorerInterface
from espnet.nets.scorer_interface import ScorerInterface

ShigekiKarita

import logging
from typing import Any
from typing import Dict
from typing import List
from typing import NamedTuple
from typing import Tuple

import torch

from espnet.nets.e2e_asr_common import end_detect

ShigekiKarita

retrieved from https://interspeech2019.org/program/schedule/

Tutorials

[T6] Advanced methods for neural end-to-end speech processing – unification, integration, and implementation Sunday, 15 September, 1400–1730, Hall 1 Takaaki Hori (Mitsubishi Electric Research Laboratories), Tomoki Hayashi (Department of Information Science, Nagoya University), Shigeki Karita (NTT Communication Science Laboratories), Shinji Watanabe (Center for Language and Speech Processing, Johns Hopkins University)

[T8] Microphone array signal processing and deep learning for speech enhancement – strong together, Sunday, 15 September, 1400–1730, Hall 11 Reinhold Haeb-Umbach (Department of Communications Engineering, Paderborn University), Tomohiro Nakatani (NTT Communication Science Laboratories)