Initial commit

asr.sh

+../../TEMPLATE/asr1/asr.sh
\ No newline at end of file

cmd.sh

+# ====== 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 documentaion 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='sge'
+
+# 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 [ "${cmd_backend}" = sge ]; then
+    # The default setting is written in conf/queue.conf.
+    # You must change "-q g.q" for the "queue" for your environment.
+    # To know the "queue" names, type "qhost -q"
+    # Note that to use "--gpu *", you have to setup "complex_value" for the system scheduler.
+
+    export train_cmd="queue.pl --mem 20G --config conf/queue.conf"
+    export cuda_cmd="queue.pl --mem 20G --config conf/queue.conf"
+    export decode_cmd="queue.pl"
+
+
+# "qsub" (Torque/PBS.)
+elif [ "${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 [ "${cmd_backend}" = slurm ]; then
+    # The default setting is written in conf/slurm.conf.
+    # You must change "-p cpu" and "-p gpu" for the "partition" for your environment.
+    # To know the "partion" names, type "sinfo".
+    # You can use "--gpu * " by defualt for slurm and it is interpreted as "--gres gpu:*"
+    # The devices are allocated exclusively using "${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

conf/decode_asr.yaml

+lm_weight: 0.6
+ctc_weight: 0.4
+beam_size: 3
+penalty: 0.0
+maxlenratio: 0.0
+minlenratio: 0.0
+streaming: True

conf/decode_asr_streaming.yaml

+tuning/decode_asr_streaming.yaml
\ No newline at end of file

conf/fbank.conf

+--sample-frequency=16000 
+--num-mel-bins=80

conf/pbs.conf

+# Default configuration
+command qsub -V -v PATH -S /bin/bash
+option name=* -N $0
+option mem=* -l mem=$0
+option mem=0          # Do not add anything to qsub_opts
+option num_threads=* -l ncpus=$0
+option num_threads=1  # Do not add anything to qsub_opts
+option num_nodes=* -l nodes=$0:ppn=1
+default gpu=0
+option gpu=0
+option gpu=* -l ngpus=$0

conf/pitch.conf

+--sample-frequency=16000

conf/queue.conf

+# Default configuration
+command qsub -v PATH -cwd -S /bin/bash -j y -l arch=*64*
+option name=* -N $0
+option mem=* -l mem_free=$0,ram_free=$0
+option mem=0          # Do not add anything to qsub_opts
+option num_threads=* -pe smp $0
+option num_threads=1  # Do not add anything to qsub_opts
+option max_jobs_run=* -tc $0
+option num_nodes=* -pe mpi $0  # You must set this PE as allocation_rule=1
+default gpu=0
+option gpu=0
+option gpu=* -l gpu=$0 -q Default

conf/slurm.conf

+# Default configuration
+command sbatch --export=PATH
+option name=* --job-name $0
+option time=* --time $0
+option mem=* --mem-per-cpu $0
+option mem=0
+option num_threads=* --cpus-per-task $0
+option num_threads=1 --cpus-per-task 1
+option num_nodes=* --nodes $0
+default gpu=0
+option gpu=0 -p cpu
+option gpu=* -p gpu --gres=gpu:$0 -c $0  # Recommend allocating more CPU than, or equal to the number of GPU
+# note: the --max-jobs-run option is supported as a special case
+# by slurm.pl and you don't have to handle it in the config file.

conf/train_asr_streaming_transformer.yaml

+tuning/train_asr_streaming_transformer.yaml
\ No newline at end of file

conf/train_asr_transformer.yaml

+tuning/train_asr_transformer.yaml
\ No newline at end of file

conf/train_lm.yaml

+tuning/train_lm_adam.yaml
\ No newline at end of file

conf/train_lm_transformer.yaml

+tuning/train_lm_transformer.yaml
\ No newline at end of file

conf/tuning/decode_asr_streaming.yaml

+# lm_weight: 0.3
+beam_size: 10
+penalty: 0.0
+maxlenratio: 0.0
+minlenratio: 0.0
+ctc_weight: 0.3
+sim_chunk_length: 512
+disable_repetition_detection: true
+decoder_text_length_limit: 0
+encoded_feat_length_limit: 0

conf/tuning/train_asr_conformer.yaml

+batch_type: numel
+batch_bins: 140000000
+accum_grad: 1
+max_epoch: 60
+patience: none
+init: xavier_uniform
+val_scheduler_criterion:
+    - valid
+    - acc
+best_model_criterion:
+-   - valid
+    - acc
+    - max
+keep_nbest_models: 10
+
+encoder: conformer
+encoder_conf:
+    output_size: 512
+    attention_heads: 8
+    linear_units: 2048
+    num_blocks: 12
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.1
+    attention_dropout_rate: 0.1
+    input_layer: conv2d
+    normalize_before: true
+    macaron_style: true
+    pos_enc_layer_type: "rel_pos"
+    selfattention_layer_type: "rel_selfattn"
+    activation_type: "swish"
+    use_cnn_module:  true
+    cnn_module_kernel: 31
+
+decoder: transformer
+decoder_conf:
+    attention_heads: 8
+    linear_units: 2048
+    num_blocks: 6
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.1
+    self_attention_dropout_rate: 0.1
+    src_attention_dropout_rate: 0.1
+
+model_conf:
+    ctc_weight: 0.3
+    lsm_weight: 0.1
+    length_normalized_loss: false
+
+optim: adam
+optim_conf:
+    lr: 0.0015
+scheduler: warmuplr
+scheduler_conf:
+    warmup_steps: 25000
+
+frontend_conf:
+  n_fft: 512
+  hop_length: 256
+
+specaug: specaug
+specaug_conf:
+    apply_time_warp: true
+    time_warp_window: 5
+    time_warp_mode: bicubic
+    apply_freq_mask: true
+    freq_mask_width_range:
+    - 0
+    - 30
+    num_freq_mask: 2
+    apply_time_mask: true
+    time_mask_width_range:
+    - 0
+    - 40
+    num_time_mask: 2
\ No newline at end of file

conf/tuning/train_asr_streaming_conformer.yaml

+# network architecture
+# encoder related
+encoder: contextual_block_conformer    # contextual_block_conformer is the core of streaming conformer
+encoder_conf:
+    output_size: 256    # dimension of attention
+    attention_heads: 4
+    linear_units: 2048  # the number of units of position-wise feed forward
+    num_blocks: 12      # the number of encoder blocks
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.1
+    attention_dropout_rate: 0.0
+    input_layer: conv2d # encoder architecture type
+    normalize_before: true
+    activation_type: swish
+    macaron_style: true
+    use_cnn_module: true
+    cnn_module_kernel: 15
+    block_size: 40      # streaming configuration
+    hop_size: 16        # streaming configuration
+    look_ahead: 16      # streaming configuration
+    init_average: true  # streaming configuration
+    ctx_pos_enc: true   # streaming configuration
+
+# decoder related
+decoder: transformer
+decoder_conf:
+    attention_heads: 4
+    linear_units: 2048
+    num_blocks: 6
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.1
+    self_attention_dropout_rate: 0.0
+    src_attention_dropout_rate: 0.0
+
+# hybrid CTC/attention
+model_conf:
+    ctc_weight: 0.3
+    lsm_weight: 0.1     # label smoothing option
+    length_normalized_loss: false
+
+# minibatch related
+# batch_type: folded
+# batch_size: 128
+batch_type: numel
+batch_bins: 50000000
+#valid_batch_size: 1
+# optimization related
+accum_grad: 1
+grad_clip: 5
+patience: 3
+max_epoch: 50
+val_scheduler_criterion:
+    - valid
+    - acc
+best_model_criterion:
+-   - valid
+    - acc
+    - max
+early_stopping_criterion:
+    - valid
+    - cer_ctc
+    - min
+keep_nbest_models: 10
+
+# NoamLR is deprecated. Use WarmupLR.
+# The following is equivalent setting for NoamLR:
+#
+#    optim: adam
+#    optim_conf:
+#        lr: 10.
+#    scheduler: noamlr
+#    scheduler_conf:
+#        model_size: 256
+#        warmup_steps: 25000
+#
+optim: adam
+optim_conf:
+    lr: 0.0005
+scheduler: warmuplr     # pytorch v1.1.0+ required
+scheduler_conf:
+    warmup_steps: 30000
+num_att_plot: 0
+specaug: specaug
+specaug_conf:
+    apply_time_warp: true
+    time_warp_window: 5
+    time_warp_mode: bicubic
+    apply_freq_mask: true
+    freq_mask_width_range:
+    - 0
+    - 30
+    num_freq_mask: 2
+    apply_time_mask: true
+    time_mask_width_range:
+    - 0
+    - 40
+    num_time_mask: 2
\ No newline at end of file

conf/tuning/train_asr_streaming_transformer.yaml

+# This configuration requires 4 GPUs with 32GB memory
+batch_type: numel
+batch_bins: 140000000
+accum_grad: 6
+max_epoch: 33
+patience: none
+# The initialization method for model parameters
+init: xavier_uniform
+best_model_criterion:
+-   - valid
+    - acc
+    - max
+keep_nbest_models: 10
+
+encoder: contextual_block_transformer
+encoder_conf:
+    output_size: 512
+    attention_heads: 8
+    linear_units: 2048
+    num_blocks: 18
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.1
+    attention_dropout_rate: 0.1
+    input_layer: conv2d6
+    normalize_before: true
+    block_size: 40
+    hop_size: 16
+    look_ahead: 16
+    init_average: true
+    ctx_pos_enc: true
+
+decoder: transformer
+decoder_conf:
+    attention_heads: 8
+    linear_units: 2048
+    num_blocks: 6
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.1
+    self_attention_dropout_rate: 0.1
+    src_attention_dropout_rate: 0.1
+
+model_conf:
+    ctc_weight: 0.3
+    lsm_weight: 0.1
+    length_normalized_loss: false
+
+optim: adam
+optim_conf:
+    lr: 0.002
+scheduler: warmuplr
+scheduler_conf:
+    warmup_steps: 25000
+num_att_plot: 0
+specaug: specaug
+specaug_conf:
+    apply_time_warp: true
+    time_warp_window: 5
+    time_warp_mode: bicubic
+    apply_freq_mask: true
+    freq_mask_width_range:
+    - 0
+    - 30
+    num_freq_mask: 2
+    apply_time_mask: true
+    time_mask_width_range:
+    - 0
+    - 40
+    num_time_mask: 2

conf/tuning/train_asr_transformer.yaml

+# encoder related
+encoder: transformer
+encoder_conf:
+    output_size: 512
+    attention_heads: 8
+    linear_units: 2048
+    num_blocks: 12
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.1
+    attention_dropout_rate: 0.0
+    input_layer: conv2d
+    normalize_before: true
+
+# decoder related
+decoder: transformer
+decoder_conf:
+    attention_heads: 8
+    linear_units: 2048
+    num_blocks: 6
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.1
+    self_attention_dropout_rate: 0.0
+    src_attention_dropout_rate: 0.0
+
+# hybrid CTC/attention
+model_conf:
+    ctc_weight: 0.3
+    lsm_weight: 0.1     # label smoothing option
+    length_normalized_loss: false
+
+# minibatch related
+# batch_type: folded
+# batch_size: 32
+
+# optimization related
+batch_type: numel
+batch_bins: 16000000
+accum_grad: 4
+# grad_clip : 5
+max_epoch: 200
+# patience: 3
+patience: none
+init: xavier_uniform
+val_scheduler_criterion:
+    - valid
+    - acc
+best_model_criterion:
+-   - valid
+    - acc
+    - max
+keep_nbest_models: 10
+
+# NoamLR is deprecated. Use WarmupLR.
+optim: adam
+optim_conf:
+    lr: 0.002
+scheduler: warmuplr
+scheduler_conf:
+    warmup_steps: 25000
+
+specaug: specaug
+specaug_conf:
+    apply_time_warp: true
+    time_warp_window: 5
+    time_warp_mode: bicubic
+    apply_freq_mask: true
+    freq_mask_width_range:
+    - 0
+    - 30
+    num_freq_mask: 2
+    apply_time_mask: true
+    time_mask_width_range:
+    - 0
+    - 40
+    num_time_mask: 2

conf/tuning/train_asr_transformer3.yaml

+batch_type: numel
+batch_bins: 15000000
+accum_grad: 6
+max_epoch: 100
+patience: none
+# The initialization method for model parameters
+init: xavier_uniform
+best_model_criterion:
+-   - valid
+    - acc
+    - max
+keep_nbest_models: 10
+
+encoder: transformer
+encoder_conf:
+    output_size: 512
+    attention_heads: 8
+    linear_units: 2048
+    num_blocks: 18
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.1
+    attention_dropout_rate: 0.1
+    input_layer: conv2d6
+    normalize_before: true
+
+decoder: transformer
+decoder_conf:
+    attention_heads: 8
+    linear_units: 2048
+    num_blocks: 6
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.1
+    self_attention_dropout_rate: 0.1
+    src_attention_dropout_rate: 0.1
+
+model_conf:
+    ctc_weight: 0.3
+    lsm_weight: 0.1
+    length_normalized_loss: false
+
+optim: adam
+optim_conf:
+    lr: 0.002
+scheduler: warmuplr
+scheduler_conf:
+    warmup_steps: 25000
+
+specaug: specaug
+specaug_conf:
+    apply_time_warp: true
+    time_warp_window: 5
+    time_warp_mode: bicubic
+    apply_freq_mask: true
+    freq_mask_width_range:
+    - 0
+    - 30
+    num_freq_mask: 2
+    apply_time_mask: true
+    time_mask_width_range:
+    - 0
+    - 40
+    num_time_mask: 2

conf/tuning/train_lm_adam.yaml

+lm_conf:
+    nlayers: 4
+    unit: 2048
+optim: adam
+optim_conf:
+    lr: 0.001
+batch_type: folded
+batch_size: 400   # batch size in LM training
+max_epoch: 20     # if the data size is large, we can reduce this
+patience: 3
+
+best_model_criterion:
+-   - valid
+    - loss
+    - min
+keep_nbest_models: 1

conf/tuning/train_lm_transformer.yaml

+# Trained with Nvidia TESLA V100, with 16GM RAM, x4
+lm: transformer
+lm_conf:
+    pos_enc: null
+    embed_unit: 128
+    att_unit: 512
+    head: 8
+    unit: 2048
+    layer: 16
+    dropout_rate: 0.1
+
+# optimization related
+grad_clip: 5.0
+batch_type: numel
+batch_bins: 2000000
+accum_grad: 1
+max_epoch: 15  # 15epoch is enougth
+
+optim: adam
+optim_conf:
+   lr: 0.001
+scheduler: warmuplr
+scheduler_conf:
+   warmup_steps: 25000
+
+best_model_criterion:
+-   - valid
+    - loss
+    - min
+keep_nbest_models: 10  # 10 is good.
\ No newline at end of file

db.sh

+../../TEMPLATE/asr1/db.sh
\ No newline at end of file

decode.sh

+../asr1/decode.sh
\ No newline at end of file

local/data.sh

+#!/bin/bash
+# Set bash to 'debug' mode, it will exit on :
+# -e 'error', -u 'undefined variable', -o ... 'error in pipeline', -x 'print commands',
+set -e
+set -u
+set -o pipefail
+
+log() {
+    local fname=${BASH_SOURCE[1]##*/}
+    echo -e "$(date '+%Y-%m-%dT%H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
+}
+SECONDS=0
+
+
+stage=0
+nj=32
+stop_stage=100000
+train_set=train_all
+valid_set=dev_all
+test_sets="tat-vol1-test tat-vol2-test tat-tmp-test"
+use_noise=false
+global_path=
+
+log "$0 $*"
+. ./path.sh || exit 1
+. ./cmd.sh || exit 1
+. utils/parse_options.sh || exit 1
+
+
+if [ $# -ne 0 ]; then
+    log "Error: No positional arguments are required."
+    exit 2
+fi
+
+if [ ${stage} -le 1 ]; then
+  # combine the sub sets into the train_set
+
+  echo "Stage 0: Combine Multiple Train Data Source"
+  utils/combine_data.sh --extra-files utt2num_frames data/${train_set} \
+      data/train-data/{tat-vol1-train,tat-vol2-train,tat-tmp}
+
+  # combine the sub sets into the dev_set
+  echo "Stage 0: Combine Multiple Dev Data Source"
+  utils/combine_data.sh --extra-files utt2num_frames data/${valid_set} \
+      data/train-data/{tat-vol1-dev,tat-vol2-dev}
+
+  if $use_noise; then
+    echo "Use FaNT to add noise"
+    train_set=${train_set}_noise
+    rirsdir=
+    noise_opt=()
+    dir_opt=()
+    dest_opt=()
+
+    # use FaNT to increase data diversity
+    noise_opt+=("/nfs/TS-1635AX/Corpora/DNS-Challenge")
+    for n in ${noise_opt[@]}; do
+      srcdir=data/train_all
+      _n=$(echo $n | awk -F'/' '{print $NF}')
+      local/multi_condition/perturb_data_dir_fant_convert.sh --nj $nj \
+                                                            --noisedir $n \
+                                                            $srcdir
+      dest_opt+=(${srcdir}_fant_${_n})
+    done
+    utils/data/combine_data.sh data/train_all_fant ${dest_opt[@]}
+    rm -r ${dest_opt[@]}
+
+    dir_opt+=(data/train_all_fant)
+    dir_opt+=(data/train_all)
+
+    # add RIRs, simulated RIRs, isotropic noises and point-source noises
+    if [[ -n "$rirsdir" ]]; then
+      srcdir=data/train_all_fant
+      samplerate=16000
+      # Make a version with reverberated speech
+      rvb_opts=()
+      rvb_opts+=(--rir-set-parameters "0.5, ${rirsdir}/simulated_rirs/smallroom/rir_list")
+      rvb_opts+=(--rir-set-parameters "0.5, ${rirsdir}/simulated_rirs/mediumroom/rir_list")
+      # Make a reverberated version of the SWBD+SRE list. Note that we don't add any
+      # additive noise here.
+      python3 steps/data/reverberate_data_dir.py "${rvb_opts[@]}" \
+                                                --prefix "reverb" \
+                                                --speech-rvb-probability 1 \
+                                                --pointsource-noise-addition-probability 0 \
+                                                --isotropic-noise-addition-probability 0 \
+                                                --num-replications 1 \
+                                                --source-sampling-rate $samplerate \
+                                                ${srcdir} ${srcdir}_reverb
+      dir_opt+=(${srcdir}_reverb)
+    fi
+    utils/data/combine_data.sh data/${train_set} ${dir_opt[@]}
+    utils/fix_data_dir.sh data/${train_set}
+  fi
+fi
+
+#if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
+#  # use external data
+#  echo "$0: preparing extra corpus for subword LM training..."
+#  mkdir -p data/local/other_text
+#  local/lm/prepare_extra_text.sh --normjobs $nj \
+#                              --global-path $global_path \
+#                              data/local/lm data/local/lm/corpus || exit 1
+#  if [ ! -e data/local/other_text/text ]; then
+#    # provide utterance id to each texts
+#    find data/local/lm/norm -mindepth 1 -maxdepth 3 -type f | xargs cat > data/local/other_text/text
+#  fi
+#fi
+
+if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then
+  for f in ${test_sets}; do
+    cp -r data/test-data/${f} data
+    utils/fix_data_dir.sh data/${f}
+  done
+fi
+
+log "Successfully finished. [elapsed=${SECONDS}s]"

local/path.sh

+# We only need the NCTU parser stuff below for the optional text normalization(for LM-training) step
+PARSER_ROOT=tools/new_parser_UNICODE
+export PATH=$PATH:$PARSER_ROOT
+
+# FaNT is needed for noise/phone
+FANT_ROOT=tools/fant
+export PATH=$PATH:$FANT_ROOT
+
+# g729a/b is needed for noise/phone
+G729_ROOT=tools/g729a
+export PATH=$PATH:$G729_ROOT

path.sh

+../../TEMPLATE/asr1/path.sh
\ No newline at end of file

pyscripts

+../../TEMPLATE/asr1/pyscripts
\ No newline at end of file

run.sh

+#!/bin/bash
+# Set bash to 'debug' mode, it will exit on :
+# -e 'error', -u 'undefined variable', -o ... 'error in pipeline', -x 'print commands',
+set -e
+set -u
+set -o pipefail
+
+# E2E model related
+train_set=train_all
+valid_set=dev_all
+test_sets="tat-vol1-test tat-vol2-test tat-tmp-test"
+use_noise=false
+global_path=`pwd`
+nj=150
+stage=0
+stop_stage=10000
+
+. ./path.sh
+. ./cmd.sh
+. ./utils/parse_options.sh
+
+asr_config=conf/tuning/train_asr_streaming_conformer.yaml
+lm_config=conf/tuning/train_lm_transformer.yaml
+inference_config=conf/decode_asr_streaming.yaml
+
+if "${use_noise}"; then train_set=${train_set}_noise; fi
+
+./asr.sh \ß
+    --stage $stage \
+    --stop_stage $stop_stage \
+    --use_streaming true \
+    --use_lm false \
+    --nj $nj \
+    --lang tw \
+    --ngpu 10 \
+    --num_nodes 1 \
+    --nbpe 5000 \
+    --token_type word \
+    --feats_type raw \
+    --audio_format wav \
+    --max_wav_duration 30 \
+    --speed_perturb_factors "0.9 1.0 1.1" \
+    --asr_config "${asr_config}" \
+    --lm_config "${lm_config}" \
+    --inference_config "${inference_config}" \
+    --local_data_opts "--global-path $global_path --nj $nj --stage 1" \
+    --train_set "${train_set}" \
+    --valid_set "${valid_set}" \
+    --test_sets "${test_sets}" \
+    --lm_train_text "data/${train_set}/text" \
+    --bpe_train_text "data/${train_set}/text" "$@"
+    

run_decode.sh

+#!/bin/bash
+# Set bash to 'debug' mode, it will exit on :
+# -e 'error', -u 'undefined variable', -o ... 'error in pipeline', -x 'print commands',
+set -e
+set -u
+set -o pipefail
+
+# E2E model related
+train_set=train_all
+valid_set=dev_all
+test_sets="tat-vol1-test tat-vol2-test tat-tmp-test"
+ßuse_noise=false
+global_path=`pwd`
+nj=100
+
+. ./path.sh
+. ./cmd.sh
+. ./utils/parse_options.sh
+
+asr_config=conf/tuning/train_asr_streaming_conformer.yaml
+lm_config=conf/tuning/train_lm_transformer.yaml
+inference_config=conf/decode_asr_streaming.yaml
+
+if "${use_noise}"; then train_set=${train_set}_noise; fi
+
+./decode.sh \
+    --stage 0 \
+    --stop_stage 10000 \
+    --use_streaming true \
+    --gpu_inference false \
+    --inference_nj $nj \
+    --use_lm false \
+    --nj $nj \
+    --lang tw \
+    --ngpu 10 \
+    --num_nodes 1 \
+    --nbpe 5000 \
+    --token_type word \
+    --feats_type raw \
+    --audio_format wav \
+    --speed_perturb_factors "0.9 1.0 1.1" \
+    --asr_config "${asr_config}" \
+    --lm_config "${lm_config}" \
+    --inference_config "${inference_config}" \
+    --local_data_opts "--global-path $global_path --stage 1" \
+    --train_set "${train_set}" \
+    --valid_set "${valid_set}" \
+    --test_sets "${test_sets}" \
+    --asr_speech_fold_length 512 \
+    --asr_text_fold_length 150 \
+    --lm_fold_length 150 \
+    --lm_train_text "data/${train_set}/text"
+    

scripts

+../../TEMPLATE/asr1/scripts
\ No newline at end of file

steps

+../../../tools/kaldi/egs/wsj/s5/steps
\ No newline at end of file

utils

+../../../tools/kaldi/egs/wsj/s5/utils
\ No newline at end of file