Merge pull request #4752 from espnet/master

Update UASR branch with latest ESPnet functions

egs2/README.md

@@ -99,6 +99,7 @@ See: https://espnet.github.io/espnet/espnet2_tutorial.html#recipes-using-espnet2
 | ofuton_p_utagoe_db      | Ofuton_p_utagoe Singing voice synthesis corpus                                                                                   | SVS                     | JPN                  | https://sites.google.com/view/oftn-utagoedb/%E3%83%9B%E3%83%BC%E3%83%A0                                      |              |
 | open_li110              | Corpus combination with 110 languages                                                                                            | Multilingual ASR        | 100+ languages       |                                                                                                              |              |
 | open_li52               | Corpus combination with 52 languages(Commonvocie + voxforge)                                                                     | Multilingual ASR        | 52 languages         |                                                                                                              |              |
+| opencpop                | Opencpop: Mandarin singing voice synthesis corpus                                                                                | SVS                     | CMN                  | https://wenet.org.cn/opencpop/                                                                               |              |
 | polyphone_swiss_french  | Swiss French Polyphone corpus                                                                                                    | ASR                     | FRA                  | http://catalog.elra.info/en-us/repository/browse/ELRA-S0030_02                                               |              |
 | primewords_chinese      | Primewords Chinese Corpus Set 1                                                                                                  | ASR                     | CMN                  | https://www.openslr.org/47/                                                                                  |              |
 | puebla_nahuatl          | Highland Puebla Nahuatl corpus (endangered language in central Mexico)                                                           | ASR/ST                  | HPN                  | https://www.openslr.org/92/                                                                                  |              |

egs2/TEMPLATE/asr1/db.sh

@@ -158,6 +158,8 @@ HARPERVALLEY=downloads
 TALROMUR=downloads
 DCASE=
 TEDX_SPANISH=downloads
+OFUTON=
+OPENCPOP=
 M_AILABS=downloads
 
 # For only CMU TIR environment

egs2/ofuton_p_utagoe_db/svs1/local/prep_segments_from_xml.py

@@ -112,7 +112,7 @@ def make_segment(file_id, labels, tempo, threshold=13.5):
         segment.add(label.start, label.end, label.label_id, label.note)
         # Split by breath mark
         for arti in label.note.articulations:
-            if arti.name == "breath mark":
+            if arti.name in ["breath mark", "up bow"]:
                 segments.extend(segment.split(threshold=threshold))
                 segment = SegInfo()
 

egs2/ofuton_p_utagoe_db/svs1/run.sh

@@ -15,13 +15,7 @@ win_length=1200
 
 score_feats_extract=frame_score_feats   # frame_score_feats | syllable_score_feats
 
-opts=
-if [ "${fs}" -eq 48000 ]; then
-    # To suppress recreation, specify wav format
-    opts="--audio_format wav "
-else
-    opts="--audio_format flac "
-fi
+opts="--audio_format wav "
 
 train_set=tr_no_dev
 valid_set=dev
@@ -56,5 +50,4 @@ cleaner=none
     --test_sets "${test_sets}" \
     --score_feats_extract "${score_feats_extract}" \
     --srctexts "data/${train_set}/text" \
-    --ngpu 1 \
     ${opts} "$@"

egs2/opencpop/svs1/cmd.sh

@@ -0,0 +1,110 @@
+# ====== About run.pl, queue.pl, slurm.pl, and ssh.pl ======
+# Usage: <cmd>.pl [options] JOB=1:<nj> <log> <command...>
+# e.g.
+#   run.pl --mem 4G JOB=1:10 echo.JOB.log echo JOB
+#
+# Options:
+#   --time <time>: Limit the maximum time to execute.
+#   --mem <mem>: Limit the maximum memory usage.
+#   -–max-jobs-run <njob>: Limit the number parallel jobs. This is ignored for non-array jobs.
+#   --num-threads <ngpu>: Specify the number of CPU core.
+#   --gpu <ngpu>: Specify the number of GPU devices.
+#   --config: Change the configuration file from default.
+#
+# "JOB=1:10" is used for "array jobs" and it can control the number of parallel jobs.
+# The left string of "=", i.e. "JOB", is replaced by <N>(Nth job) in the command and the log file name,
+# e.g. "echo JOB" is changed to "echo 3" for the 3rd job and "echo 8" for 8th job respectively.
+# Note that the number must start with a positive number, so you can't use "JOB=0:10" for example.
+#
+# run.pl, queue.pl, slurm.pl, and ssh.pl have unified interface, not depending on its backend.
+# These options are mapping to specific options for each backend and
+# it is configured by "conf/queue.conf" and "conf/slurm.conf" by default.
+# If jobs failed, your configuration might be wrong for your environment.
+#
+#
+# The official documentation for run.pl, queue.pl, slurm.pl, and ssh.pl:
+#   "Parallelization in Kaldi": http://kaldi-asr.org/doc/queue.html
+# =========================================================~
+
+
+# Select the backend used by run.sh from "local", "stdout", "sge", "slurm", or "ssh"
+cmd_backend='local'
+
+# Local machine, without any Job scheduling system
+if [ "${cmd_backend}" = local ]; then
+
+    # The other usage
+    export train_cmd="run.pl"
+    # Used for "*_train.py": "--gpu" is appended optionally by run.sh
+    export cuda_cmd="run.pl"
+    # Used for "*_recog.py"
+    export decode_cmd="run.pl"
+
+# Local machine logging to stdout and log file, without any Job scheduling system
+elif [ "${cmd_backend}" = stdout ]; then
+
+    # The other usage
+    export train_cmd="stdout.pl"
+    # Used for "*_train.py": "--gpu" is appended optionally by run.sh
+    export cuda_cmd="stdout.pl"
+    # Used for "*_recog.py"
+    export decode_cmd="stdout.pl"
+
+
+# "qsub" (Sun Grid Engine, or derivation of it)
+elif [ "${cmd_backend}" = sge ]; then
+    # The default setting is written in conf/queue.conf.
+    # You must change "-q g.q" for the "queue" for your environment.
+    # To know the "queue" names, type "qhost -q"
+    # Note that to use "--gpu *", you have to setup "complex_value" for the system scheduler.
+
+    export train_cmd="queue.pl"
+    export cuda_cmd="queue.pl"
+    export decode_cmd="queue.pl"
+
+
+# "qsub" (Torque/PBS.)
+elif [ "${cmd_backend}" = pbs ]; then
+    # The default setting is written in conf/pbs.conf.
+
+    export train_cmd="pbs.pl"
+    export cuda_cmd="pbs.pl"
+    export decode_cmd="pbs.pl"
+
+
+# "sbatch" (Slurm)
+elif [ "${cmd_backend}" = slurm ]; then
+    # The default setting is written in conf/slurm.conf.
+    # You must change "-p cpu" and "-p gpu" for the "partition" for your environment.
+    # To know the "partion" names, type "sinfo".
+    # You can use "--gpu * " by default for slurm and it is interpreted as "--gres gpu:*"
+    # The devices are allocated exclusively using "${CUDA_VISIBLE_DEVICES}".
+
+    export train_cmd="slurm.pl"
+    export cuda_cmd="slurm.pl"
+    export decode_cmd="slurm.pl"
+
+elif [ "${cmd_backend}" = ssh ]; then
+    # You have to create ".queue/machines" to specify the host to execute jobs.
+    # e.g. .queue/machines
+    #   host1
+    #   host2
+    #   host3
+    # Assuming you can login them without any password, i.e. You have to set ssh keys.
+
+    export train_cmd="ssh.pl"
+    export cuda_cmd="ssh.pl"
+    export decode_cmd="ssh.pl"
+
+# This is an example of specifying several unique options in the JHU CLSP cluster setup.
+# Users can modify/add their own command options according to their cluster environments.
+elif [ "${cmd_backend}" = jhu ]; then
+
+    export train_cmd="queue.pl --mem 2G"
+    export cuda_cmd="queue-freegpu.pl --mem 2G --gpu 1 --config conf/queue.conf"
+    export decode_cmd="queue.pl --mem 4G"
+
+else
+    echo "$0: Error: Unknown cmd_backend=${cmd_backend}" 1>&2
+    return 1
+fi

egs2/opencpop/svs1/conf/decode.yaml

@@ -0,0 +1 @@
+tuning/decode_rnn.yaml

egs2/opencpop/svs1/conf/train.yaml

@@ -0,0 +1 @@
+tuning/train_naive_rnn.yaml

egs2/opencpop/svs1/conf/tuning/decode_rnn.yaml

@@ -0,0 +1,10 @@
+# This configuration is the decoding setting for FastSpeech or FastSpeech2.
+
+##########################################################
+#                    DECODING SETTING                    #
+##########################################################
+# speed_control_alpha: 1   # alpha to control the speed of generated speech
+                           # 1 < alpha makes slower and 1 > alpha makes faster
+use_teacher_forcing: false # whether to use teacher forcing
+                           # if true, we use groundtruth of durations
+                           # (+ pitch & energy for FastSpeech2)

egs2/opencpop/svs1/conf/tuning/train_naive_rnn.yaml

@@ -0,0 +1,69 @@
+
+
+##########################################################
+#                  SVS MODEL SETTING                     #
+##########################################################
+svs: naive_rnn                   # model architecture
+svs_conf:                        # keyword arguments for the selected model
+    midi_dim: 129                # midi dimension (note number + silence)
+    embed_dim: 512               # char or phn embedding dimension
+    eprenet_conv_layers: 0       # prenet (from bytesing) conv layers
+    eprenet_conv_chans: 256      # prenet (from bytesing) conv channels numbers
+    eprenet_conv_filts: 3        # prenet (from bytesing) conv filters size
+    elayers: 3                   # number of lstm layers in encoder
+    eunits: 512                  # number of lstm units
+    ebidirectional: True         # if bidirectional in encoder
+    midi_embed_integration_type: add # how to integrate midi information
+    dlayers: 5                   # number of lstm layers in decoder
+    dunits: 1024                 # number of lstm units in decoder
+    dbidirectional: True         # if bidirectional in decoder
+    postnet_layers: 5            # number of layers in postnet
+    postnet_chans: 512           # number of channels in postnet
+    postnet_filts: 5             # filter size of postnet layer
+    use_batch_norm: true         # whether to use batch normalization in postnet
+    reduction_factor: 1          # reduction factor
+    eprenet_dropout_rate: 0.2    # prenet dropout rate
+    edropout_rate: 0.1           # encoder dropout rate
+    ddropout_rate: 0.1           # decoder dropout rate
+    postnet_dropout_rate: 0.5    # postnet dropout_rate
+    init_type: pytorch           # parameter initialization
+    use_masking: true            # whether to apply masking for padded part in loss calculation
+    loss_type: L1
+
+
+##########################################################
+#                  OPTIMIZER SETTING                     #
+##########################################################
+optim: adam           # optimizer type
+optim_conf:           # keyword arguments for selected optimizer
+    lr: 1.0e-03       # learning rate
+    eps: 1.0e-06      # epsilon
+    weight_decay: 0.0 # weight decay coefficient
+
+##########################################################
+#                OTHER TRAINING SETTING                  #
+##########################################################
+# num_iters_per_epoch: 200    # number of iterations per epoch
+max_epoch: 500              # number of epochs
+grad_clip: 1.0              # gradient clipping norm
+grad_noise: false           # whether to use gradient noise injection
+accum_grad: 1               # gradient accumulation
+
+batch_type: sorted
+batch_size: 16
+
+sort_in_batch: descending   # how to sort data in making batch
+sort_batch: descending      # how to sort created batches
+num_workers: 8              # number of workers of data loader
+train_dtype: float32        # dtype in training
+log_interval: null          # log interval in iterations
+keep_nbest_models: 2        # number of models to keep
+num_att_plot: 3             # number of attention figures to be saved in every check
+seed: 0                     # random seed number
+best_model_criterion:
+-   - valid
+    - loss
+    - min
+-   - train
+    - loss
+    - min

egs2/opencpop/svs1/conf/tuning/train_naive_rnn_dp.yaml

@@ -0,0 +1,69 @@
+
+
+##########################################################
+#                  SVS MODEL SETTING                     #
+##########################################################
+svs: naive_rnn_dp                   # model architecture
+svs_conf:                        # keyword arguments for the selected model
+    midi_dim: 129                # midi dimension (note number + silence)
+    embed_dim: 512               # char or phn embedding dimension
+    tempo_dim: 500
+    eprenet_conv_layers: 0       # prenet (from bytesing) conv layers
+    eprenet_conv_chans: 256      # prenet (from bytesing) conv channels numbers
+    eprenet_conv_filts: 3        # prenet (from bytesing) conv filters size
+    elayers: 3                   # number of lstm layers in encoder
+    eunits: 256                  # number of lstm units
+    ebidirectional: True         # if bidirectional in encoder
+    midi_embed_integration_type: add # how to integrate midi information
+    dlayers: 2                   # number of lstm layers in decoder
+    dunits: 256                  # number of lstm units in decoder
+    dbidirectional: True         # if bidirectional in decoder
+    postnet_layers: 5            # number of layers in postnet
+    postnet_chans: 512           # number of channels in postnet
+    postnet_filts: 5             # filter size of postnet layer
+    use_batch_norm: true         # whether to use batch normalization in postnet
+    reduction_factor: 1          # reduction factor
+    eprenet_dropout_rate: 0.2    # prenet dropout rate
+    edropout_rate: 0.1           # encoder dropout rate
+    ddropout_rate: 0.1           # decoder dropout rate
+    postnet_dropout_rate: 0.5    # postnet dropout_rate
+    init_type: pytorch           # parameter initialization
+    use_masking: true            # whether to apply masking for padded part in loss calculation
+
+
+##########################################################
+#                  OPTIMIZER SETTING                     #
+##########################################################
+optim: adam           # optimizer type
+optim_conf:           # keyword arguments for selected optimizer
+    lr: 1.0e-03       # learning rate
+    eps: 1.0e-06      # epsilon
+    weight_decay: 0.0 # weight decay coefficient
+
+##########################################################
+#                OTHER TRAINING SETTING                  #
+##########################################################
+# num_iters_per_epoch: 200    # number of iterations per epoch
+max_epoch: 500              # number of epochs
+grad_clip: 1.0              # gradient clipping norm
+grad_noise: false           # whether to use gradient noise injection
+accum_grad: 1               # gradient accumulation
+
+batch_type: sorted
+batch_size: 16
+
+sort_in_batch: descending   # how to sort data in making batch
+sort_batch: descending      # how to sort created batches
+num_workers: 8              # number of workers of data loader
+train_dtype: float32        # dtype in training
+log_interval: null          # log interval in iterations
+keep_nbest_models: 2        # number of models to keep
+num_att_plot: 3             # number of attention figures to be saved in every check
+seed: 0                     # random seed number
+best_model_criterion:
+-   - valid
+    - loss
+    - min
+-   - train
+    - loss
+    - min

egs2/opencpop/svs1/db.sh

@@ -0,0 +1 @@
+../../TEMPLATE/svs1/db.sh

egs2/opencpop/svs1/local/data.sh

@@ -0,0 +1,72 @@
+#!/usr/bin/env bash
+
+set -e
+set -u
+set -o pipefail
+
+. ./path.sh || exit 1;
+. ./cmd.sh || exit 1;
+. ./db.sh || exit 1;
+
+log() {
+    local fname=${BASH_SOURCE[1]##*/}
+    echo -e "$(date '+%Y-%m-%dT%H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
+}
+
+SECONDS=0
+stage=1
+stop_stage=100
+fs=24000
+
+log "$0 $*"
+
+. utils/parse_options.sh || exit 1;
+
+if [ -z "${OPENCPOP}" ]; then
+    log "Fill the value of 'OPENCPOP' of db.sh"
+    exit 1
+fi
+
+mkdir -p ${OPENCPOP}
+
+train_set="tr_no_dev"
+train_dev="dev"
+
+if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
+    log "stage 0: Data Download"
+    # The Opencpop data should be downloaded from https://wenet.org.cn/opencpop/
+fi
+
+if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
+    log "stage 1: Data preparaion "
+
+    mkdir -p xml_dump
+    mkdir -p wav_dump
+    # we convert the music score to xml format
+    python local/data_prep.py ${OPENCPOP} --midi_note_scp local/midi-note.scp \
+        --xml_dumpdir xml_dump \
+        --wav_dumpdir wav_dump \
+        --sr ${fs}
+    for src_data in train eval; do
+        utils/utt2spk_to_spk2utt.pl < data/${src_data}/utt2spk > data/${src_data}/spk2utt
+        utils/fix_data_dir.sh --utt_extra_files "label musicxml.scp" data/${src_data}
+    done
+fi
+
+if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
+    log "stage 2: Held out validation set"
+
+    utils/copy_data_dir.sh data/train data/${train_set}
+    utils/copy_data_dir.sh data/train data/${train_dev}
+    for dset in ${train_set} ${train_dev}; do
+        for extra_file in label musicxml.scp; do
+            cp data/train/${extra_file} data/${dset}
+        done
+    done
+    tail -n 50 data/train/wav.scp > data/dev/wav.scp
+    utils/filter_scp.pl --exclude data/dev/wav.scp data/train/wav.scp > data/tr_no_dev/wav.scp
+
+    utils/fix_data_dir.sh --utt_extra_files "label musicxml.scp" data/tr_no_dev
+    utils/fix_data_dir.sh --utt_extra_files "label musicxml.scp" data/dev
+
+fi