diff --git a/egs2/README.md b/egs2/README.md
index c871591d867..185a20b875d 100755
--- a/egs2/README.md
+++ b/egs2/README.md
@@ -16,20 +16,21 @@ See: https://espnet.github.io/espnet/espnet2_tutorial.html#recipes-using-espnet2
 | an4                     | CMU AN4 database                                                                                                                 | ASR/TTS                 | ENG                   | http://www.speech.cs.cmu.edu/databases/an4/                                                                  |              |
 | babel                   | IARPA Babel corups                                                                                                               | ASR                     | ~20 languages         | https://www.iarpa.gov/index.php/research-programs/babel                                                      |              |
 | bn_openslr53            | Large bengali ASR training dataset                                                                                               | ASR                     | BEN                   | https://openslr.org/53/                                                                                      |              |
-| bur_openslr80           | Burmese ASR training dataset                                                                                                     | ASR                     | BUR                   | https://openslr.org/80/  
-| catslu               	  | CATSLU-MAPS                                                                                                                      | SLU                     | CMN           	      | https://sites.google.com/view/catslu/home                                                                    |              |
+| bur_openslr80           | Burmese ASR training dataset                                                                                                     | ASR                     | BUR                   | https://openslr.org/80/                                                                                      |              |
+| catslu               	  | CATSLU-MAPS                                                                                                                      | SLU                     | CMN           	      | https://sites.google.com/view/catslu/home                                                                     |              |
 | chime4                  | The 4th CHiME Speech Separation and Recognition Challenge                                                                        | ASR/Multichannel ASR    | ENG                   | http://spandh.dcs.shef.ac.uk/chime_challenge/chime2016/                                                      |              |
-| clarity21               | The First Clarity Enhancement Challenge CEC1                                         | SE                      | ENG                   | https://claritychallenge.github.io/clarity_CEC1_doc/                                                   |              |
+| clarity21               | The First Clarity Enhancement Challenge CEC1                                                                                     | SE                      | ENG                   | https://claritychallenge.github.io/clarity_CEC1_doc/                                                         |              |
 | cmu_indic               | CMU INDIC                                                                                                                        | TTS                     | 7 languages           | http://festvox.org/cmu_indic/                                                                                |              |
 | commonvoice             | The Mozilla Common Voice                                                                                                         | ASR                     | 13 languages          | https://voice.mozilla.org/datasets                                                                           |              |
-| conferencingspeech21    | Far-field Multi-channel Speech Enhancement Challenge for Video Conferencing (ConferencingSpeech 2021) | SE                     | ENG, CMN            | https://tea-lab.qq.com/conferencingspeech-2021                                                               |              |
+| conferencingspeech21    | Far-field Multi-channel Speech Enhancement Challenge for Video Conferencing (ConferencingSpeech 2021)                            | SE                      | ENG, CMN              | https://tea-lab.qq.com/conferencingspeech-2021                                                               |              |
+| covost2                 | Multilingual speech-to-text translation corpus from Common Voice                                                                 | ST                      | lang pairs from 22    | https://github.com/facebookresearch/covost                                                                   |              |
 | csj                     | Corpus of Spontaneous Japanese                                                                                                   | ASR                     | JPN                   | https://pj.ninjal.ac.jp/corpus_center/csj/en/                                                                |              |
 | csmsc                   | Chinese Standard Mandarin Speech Copus                                                                                           | TTS                     | CMN                   | https://www.data-baker.com/open_source.html                                                                  |              |
 | css10                   | CSS10: A Collection of Single Speaker Speech Datasets for 10 Languages                                                           | TTS                     | 10 langauges          | https://github.com/Kyubyong/css10                                                                            |              |
 | dirha_wsj               | Distant-speech Interaction for Robust Home Applications                                                                          | Multichannel ASR        | ENG                   | https://dirha.fbk.eu/, https://github.com/SHINE-FBK/DIRHA_English_wsj                                        |              |
-| dns_ins20               | Deep Noise Suppression Challenge – INTERSPEECH 2020                                                                              | SE                      | 7 languages + singing | https://www.microsoft.com/en-us/research/academic-program/deep-noise-suppression-challenge-interspeech-2020/ |              |
-| dns_icassp21            | Deep Noise Suppression Challenge – ICASSP 2021                                                                                   | SE                      | 11 languages + singing| https://www.microsoft.com/en-us/research/academic-program/deep-noise-suppression-challenge-icassp-2021/      |              |
-| dns_ins21               | Deep Noise Suppression Challenge – INTERSPEECH 2021                                                                              | SE                      | 11 languages + singing| https://www.microsoft.com/en-us/research/academic-program/deep-noise-suppression-challenge-interspeech-2021/ |              |
+| dns_ins20               | Deep Noise Suppression Challenge – INTERSPEECH 2020                                                                          | SE                      | 7 languages +singing  | https://www.microsoft.com/en-us/research/academic-program/deep-noise-suppression-challenge-interspeech-2020/ |              |
+| dns_icassp21            | Deep Noise Suppression Challenge – ICASSP 2021                                                                               | SE                      | 11 languages + singing| https://www.microsoft.com/en-us/research/academic-program/deep-noise-suppression-challenge-icassp-2021/      |              |
+| dns_ins21               | Deep Noise Suppression Challenge – INTERSPEECH 2021                                                                          | SE                      | 11 languages + singing| https://www.microsoft.com/en-us/research/academic-program/deep-noise-suppression-challenge-interspeech-2021/ |              |
 | dsing                   | Automatic Lyric Transcription from Karaoke Vocal Tracks (From DAMP Sing300x30x2)                                                 | ASR (ALT)               | ENG singing           | https://github.com/groadabike/Kaldi-Dsing-task                                                               |              |
 | fisher_callhome_spanish | Fisher and CALLHOME Spanish--English Speech Translation                                                                          | ASR/ST                  | SPA->ENG              | https://catalog.ldc.upenn.edu/LDC2014T23                                                                     |              |
 | fsc                     | Fluent Speech Commands Dataset                                                                                                   | SLU                     | ENG                   | https://fluent.ai/fluent-speech-commands-a-dataset-for-spoken-language-understanding-research/               |              |
@@ -60,15 +61,15 @@ See: https://espnet.github.io/espnet/espnet2_tutorial.html#recipes-using-espnet2
 | ljspeech                | The LJ Speech Dataset                                                                                                            | TTS                     | ENG                  | https://keithito.com/LJ-Speech-Dataset/                                                                      |              |
 | lrs3                    | The Oxford-BBC Lip Reading Sentences 3 (LRS3) Dataset                                                                            | ASR                     | ENG                  | https://www.robots.ox.ac.uk/~vgg/data/lip_reading/lrs3.html                                                  |              |
 | lrs2                    | The Oxford-BBC Lip Reading Sentences 2 (LRS2) Dataset                                                                            | Lipreading/ASR          | ENG                  | https://www.robots.ox.ac.uk/~vgg/data/lip_reading/lrs2.html                                                  |              |
-| microsoft_speech             |      Microsoft Speech Corpus (Indian languages)                                                                                                   | ASR                     | 3 languages          | https://msropendata.com/datasets/7230b4b1-912d-400e-be58-f84e0512985e                                                                           |              |
+| microsoft_speech        | Microsoft Speech Corpus (Indian languages)                                                                                       | ASR                     | 3 languages          | https://msropendata.com/datasets/7230b4b1-912d-400e-be58-f84e0512985e                                        |              |
 | mini_an4                | Mini version of CMU AN4 database for the integration test                                                                        | ASR/TTS/SE              | ENG                  | http://www.speech.cs.cmu.edu/databases/an4/                                                                  |              |
 | mini_librispeech        | Mini version of Librispeech corpus                                                                                               | DIAR                    | ENG                  | https://openslr.org/31/                                                                                      |              |
-| ml_openslr63            | Crowdsourced high-quality Malayalam multi-speaker speech data                           | ASR                     | MAL                  | https://openslr.org/63/                                                                                      |              |
+| ml_openslr63            | Crowdsourced high-quality Malayalam multi-speaker speech data                                                                    | ASR                     | MAL                  | https://openslr.org/63/                                                                                      |              |
 | mls                     | MLS (A large multilingual corpus derived from LibriVox audiobooks)                                                               | ASR                     | 8 languages          | http://www.openslr.org/94/                                                                                   |              |
 | mr_openslr64            | OpenSLR Marathi Corpus                                                                                                           | ASR                     | MAR                  | http://www.openslr.org/64/                                                                                   |              |
 | ms_indic_is18           | Microsoft Speech Corpus (Indian languages)                                                                                       | ASR                     | 3 langs: TEL TAM GUJ | https://msropendata.com/datasets/7230b4b1-912d-400e-be58-f84e0512985e                                        |              |
 | nsc                     | National Speech Corpus                                                                                                           | ASR                     | ENG-SG               | https://www.imda.gov.sg/programme-listing/digital-services-lab/national-speech-corpus                        |              |
-| open_li52               | Corpus combination with 52 languages(Commonvocie + voxforge)                                                                     | Multilingual ASR        | 52 languages         |                                                                                                             |              |
+| open_li52               | Corpus combination with 52 languages(Commonvocie + voxforge)                                                                     | Multilingual ASR        | 52 languages         |                                                                                                              |              |
 | 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                     | HPN                  | https://www.openslr.org/92/                                                                                  |              |
@@ -87,12 +88,12 @@ See: https://espnet.github.io/espnet/espnet2_tutorial.html#recipes-using-espnet2
 | su_openslr36            | Sundanese                                                                                                                        | ASR                     | SUN                  | http://www.openslr.org/36                                                                                    |              |
 | swbd                    | Switchboard Corpus for 2-channel Conversational Telephone Speech (300h)                                                          | ASR                     | ENG                  | https://catalog.ldc.upenn.edu/LDC97S62                                                                       |              |
 | swbd_da                 | NXT Switchboard Annotations                                                                                                      | SLU                     | ENG                  | https://catalog.ldc.upenn.edu/LDC2009T26                                                                     |              |
-| swbd_sentiment          | Speech Sentiment Annotations                                                                                                     | SLU                     | ENG                  | https://catalog.ldc.upenn.edu/LDC2020T14                                                                    |              |
+| swbd_sentiment          | Speech Sentiment Annotations                                                                                                     | SLU                     | ENG                  | https://catalog.ldc.upenn.edu/LDC2020T14                                                                     |              |
 | tedlium2                | TED-LIUM corpus release 2                                                                                                        | ASR                     | ENG                  | https://www.openslr.org/19/, http://www.lrec-conf.org/proceedings/lrec2014/pdf/1104_Paper.pdf                |              |
 | thchs30                 | A Free Chinese Speech Corpus Released by CSLT@Tsinghua University                                                                | TTS                     | CMN                  | https://www.openslr.org/18/                                                                                  |              |
 | timit                   | TIMIT Acoustic-Phonetic Continuous Speech Corpus                                                                                 | ASR                     | ENG                  | https://catalog.ldc.upenn.edu/LDC93S1                                                                        |              |
 | totonac                 | Highland Totonac corpus (endangered language in central Mexico)                                                                  | ASR                     | TOS                  | http://www.openslr.org/107/                                                                                  |              |
-| tsukuyomi               | つくよみちゃんコーパス                                                                                                                      | TTS                     | JPN                  | https://tyc.rei-yumesaki.net/material/corpus                                                                 |              |
+| tsukuyomi               | つくよみちゃんコーパス                                                                                | TTS                     | JPN                  | https://tyc.rei-yumesaki.net/material/corpus                                                                 |              |
 | vctk                    | English Multi-speaker Corpus for CSTR Voice Cloning Toolkit                                                                      | ASR/TTS                 | ENG                  | http://www.udialogue.org/download/cstr-vctk-corpus.html                                                      |              |
 | vctk_noisyreverb        | Noisy reverberant speech database (48kHz)                                                                                        | SE                      | ENG                  | https://datashare.ed.ac.uk/handle/10283/2826                                                                 |              |
 | vivos                   | VIVOS (Vietnamese corpus for ASR)                                                                                                | ASR                     | VIE                  | https://ailab.hcmus.edu.vn/vivos/                                                                            |              |
diff --git a/egs2/covost2/asr1/asr.sh b/egs2/covost2/asr1/asr.sh
new file mode 120000
index 00000000000..60b05122cfd
--- /dev/null
+++ b/egs2/covost2/asr1/asr.sh
@@ -0,0 +1 @@
+../../TEMPLATE/asr1/asr.sh
\ No newline at end of file
diff --git a/egs2/covost2/asr1/cmd.sh b/egs2/covost2/asr1/cmd.sh
new file mode 100644
index 00000000000..2aae6919fef
--- /dev/null
+++ b/egs2/covost2/asr1/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
diff --git a/egs2/covost2/asr1/conf/decode_asr.yaml b/egs2/covost2/asr1/conf/decode_asr.yaml
new file mode 120000
index 00000000000..f3f59d5ac2b
--- /dev/null
+++ b/egs2/covost2/asr1/conf/decode_asr.yaml
@@ -0,0 +1 @@
+tuning/decode_transformer.yaml
\ No newline at end of file
diff --git a/egs2/covost2/asr1/conf/fbank.conf b/egs2/covost2/asr1/conf/fbank.conf
new file mode 100644
index 00000000000..82ac7bd0dbc
--- /dev/null
+++ b/egs2/covost2/asr1/conf/fbank.conf
@@ -0,0 +1,2 @@
+--sample-frequency=16000 
+--num-mel-bins=80
diff --git a/egs2/covost2/asr1/conf/pbs.conf b/egs2/covost2/asr1/conf/pbs.conf
new file mode 100644
index 00000000000..119509938ce
--- /dev/null
+++ b/egs2/covost2/asr1/conf/pbs.conf
@@ -0,0 +1,11 @@
+# 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
diff --git a/egs2/covost2/asr1/conf/pitch.conf b/egs2/covost2/asr1/conf/pitch.conf
new file mode 100644
index 00000000000..e959a19d5b8
--- /dev/null
+++ b/egs2/covost2/asr1/conf/pitch.conf
@@ -0,0 +1 @@
+--sample-frequency=16000
diff --git a/egs2/covost2/asr1/conf/queue.conf b/egs2/covost2/asr1/conf/queue.conf
new file mode 100644
index 00000000000..500582fab31
--- /dev/null
+++ b/egs2/covost2/asr1/conf/queue.conf
@@ -0,0 +1,12 @@
+# 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 g.q
diff --git a/egs2/covost2/asr1/conf/slurm.conf b/egs2/covost2/asr1/conf/slurm.conf
new file mode 100644
index 00000000000..3b229673638
--- /dev/null
+++ b/egs2/covost2/asr1/conf/slurm.conf
@@ -0,0 +1,14 @@
+# 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.
diff --git a/egs2/covost2/asr1/conf/train_asr.yaml b/egs2/covost2/asr1/conf/train_asr.yaml
new file mode 120000
index 00000000000..00ea2f24791
--- /dev/null
+++ b/egs2/covost2/asr1/conf/train_asr.yaml
@@ -0,0 +1 @@
+tuning/train_asr_transformer.yaml
\ No newline at end of file
diff --git a/egs2/covost2/asr1/conf/train_lm.yaml b/egs2/covost2/asr1/conf/train_lm.yaml
new file mode 100644
index 00000000000..5bc7b7fcbfc
--- /dev/null
+++ b/egs2/covost2/asr1/conf/train_lm.yaml
@@ -0,0 +1,14 @@
+lm_conf:
+    nlayers: 2
+    unit: 650
+optim: sgd        # or adam
+batch_type: folded
+batch_size: 64   # 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
diff --git a/egs2/covost2/asr1/conf/tuning/decode_transformer.yaml b/egs2/covost2/asr1/conf/tuning/decode_transformer.yaml
new file mode 100644
index 00000000000..d89db079882
--- /dev/null
+++ b/egs2/covost2/asr1/conf/tuning/decode_transformer.yaml
@@ -0,0 +1,7 @@
+batch_size: 1
+beam_size: 10
+penalty: 0.0
+maxlenratio: 0.0
+minlenratio: 0.0
+ctc_weight: 0.5
+lm_weight: 0.3
diff --git a/egs2/covost2/asr1/conf/tuning/train_asr_transformer.yaml b/egs2/covost2/asr1/conf/tuning/train_asr_transformer.yaml
new file mode 100644
index 00000000000..a889e110cf3
--- /dev/null
+++ b/egs2/covost2/asr1/conf/tuning/train_asr_transformer.yaml
@@ -0,0 +1,52 @@
+# network architecture
+# encoder related
+encoder: transformer
+encoder_conf:
+    input_layer: "conv2d"
+    num_blocks: 12
+    linear_units: 2048
+    dropout_rate: 0.1
+    output_size: 256  # dimension of attention
+    attention_heads: 4
+    attention_dropout_rate: 0.0
+
+# decoder related
+# decoder related
+decoder: transformer
+decoder_conf:
+    input_layer: "embed"
+    num_blocks: 6
+    linear_units: 2048
+    dropout_rate: 0.1
+
+# hybrid CTC/attention
+model_conf:
+    ctc_weight: 0.3
+    lsm_weight: 0.1
+    length_normalized_loss: false
+
+# minibatch related
+batch_type: folded
+batch_size: 32
+
+# optimization related
+optim: adam
+accum_grad: 2
+grad_clip: 5
+patience: 0
+max_epoch: 100
+optim_conf:
+    lr: 1.0
+scheduler: noamlr
+scheduler_conf:
+    warmup_steps: 25000
+
+# others:
+best_model_criterion:
+-  - valid
+   - acc
+   - max
+keep_nbest_models: 10
+
+init: xavier_uniform # Our empirical studies shows that this initialization
+                     # is very important to low-resource ASR training
diff --git a/egs2/covost2/asr1/db.sh b/egs2/covost2/asr1/db.sh
new file mode 120000
index 00000000000..50d86130898
--- /dev/null
+++ b/egs2/covost2/asr1/db.sh
@@ -0,0 +1 @@
+../../TEMPLATE/asr1/db.sh
\ No newline at end of file
diff --git a/egs2/covost2/asr1/local/data.sh b/egs2/covost2/asr1/local/data.sh
new file mode 100755
index 00000000000..bae3393ae0b
--- /dev/null
+++ b/egs2/covost2/asr1/local/data.sh
@@ -0,0 +1,84 @@
+#!/usr/bin/env bash
+
+# Copyright 2021 Carnegie Mellon University (Jiatong Shi)
+#  Apache 2.0  (http://www.apache.org/licenses/LICENSE-2.0)
+
+. ./path.sh || exit 1;
+. ./cmd.sh || exit 1;
+. ./db.sh || exit 1;
+
+# general configuration
+stage=0       # start from 0 if you need to start from data preparation
+stop_stage=100
+SECONDS=0
+src_lang=es
+tgt_lang=en
+
+ . utils/parse_options.sh || exit 1;
+
+
+log() {
+    local fname=${BASH_SOURCE[1]##*/}
+    echo -e "$(date '+%Y-%m-%dT%H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
+}
+
+if [ -z "${COVOST2}" ]; then
+    log "Fill the value of 'COVOST2' of db.sh"
+    exit 1
+fi
+mkdir -p ${COVOST2}
+
+if [ -z "${COMMONVOICE}" ]; then
+    log "Fill the value of 'COMMONVOICE' of db.sh"
+    exit 1
+fi
+mkdir -p ${COMMONVOICE}
+
+# 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
+
+
+
+if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
+    log "stage 0: Data Downloading"
+
+    # base url for downloads.
+    data_url=https://voice-prod-bundler-ee1969a6ce8178826482b88e843c335139bd3fb4.s3.amazonaws.com/cv-corpus-4-2019-12-10/${src_lang}.tar.gz
+
+    # Download CommonVoice
+    mkdir -p ${COMMONVOICE}/${src_lang}
+    local/download_and_untar_commonvoice.sh ${COMMONVOICE}/${src_lang} ${data_url} ${src_lang}.tar.gz
+
+    # Download translation
+    if [[ ${src_lang} != en ]]; then
+        wget --no-check-certificate https://dl.fbaipublicfiles.com/covost/covost_v2.${src_lang}_${tgt_lang}.tsv.tar.gz \
+            -P ${COVOST2}
+        tar -xzf ${COVOST2}/covost_v2.${src_lang}_${tgt_lang}.tsv.tar.gz -C ${COVOST2}
+    fi
+    wget --no-check-certificate https://dl.fbaipublicfiles.com/covost/covost2.zip \
+          -P ${COVOST2}
+    unzip ${COVOST2}/covost2.zip -d ${COVOST2}
+    # NOTE: some non-English target languages lack translation from English
+fi
+
+if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
+    log "stage 1: Data Preparation"
+    # use underscore-separated names in data directories.
+    local/data_prep_commonvoice.pl "${COMMONVOICE}/${src_lang}" validated data/validated.${src_lang}
+
+    # text preprocessing (tokenization, case, punctuation marks etc.)
+    local/data_prep_covost2.sh ${COVOST2} ${src_lang} ${tgt_lang} || exit 1;
+    # NOTE: train/dev/test splits are different from original CommonVoice
+fi
+
+if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
+    log "stage 2: ASR Adaptation"
+    for x in train dev test; do
+        cp data/${x}.${src_lang}-${tgt_lang}/text.lc.rm.${src_lang} data/${x}.${src_lang}-${tgt_lang}/text
+    done
+fi
+
+log "Successfully finished. [elapsed=${SECONDS}s]"
diff --git a/egs2/covost2/asr1/local/data_prep_commonvoice.pl b/egs2/covost2/asr1/local/data_prep_commonvoice.pl
new file mode 120000
index 00000000000..665a6fbfa65
--- /dev/null
+++ b/egs2/covost2/asr1/local/data_prep_commonvoice.pl
@@ -0,0 +1 @@
+../../../../egs/covost2/asr1/local/data_prep_commonvoice.pl
\ No newline at end of file
diff --git a/egs2/covost2/asr1/local/data_prep_covost2.sh b/egs2/covost2/asr1/local/data_prep_covost2.sh
new file mode 120000
index 00000000000..ab5069cc162
--- /dev/null
+++ b/egs2/covost2/asr1/local/data_prep_covost2.sh
@@ -0,0 +1 @@
+../../../../egs/covost2/asr1/local/data_prep_covost2.sh
\ No newline at end of file
diff --git a/egs2/covost2/asr1/local/download_and_untar_commonvoice.sh b/egs2/covost2/asr1/local/download_and_untar_commonvoice.sh
new file mode 120000
index 00000000000..2c59b3f2047
--- /dev/null
+++ b/egs2/covost2/asr1/local/download_and_untar_commonvoice.sh
@@ -0,0 +1 @@
+../../../../egs/covost2/asr1/local/download_and_untar_commonvoice.sh
\ No newline at end of file
diff --git a/egs2/covost2/asr1/local/path.sh b/egs2/covost2/asr1/local/path.sh
new file mode 100644
index 00000000000..e69de29bb2d
diff --git a/egs2/covost2/asr1/local/process_tsv.py b/egs2/covost2/asr1/local/process_tsv.py
new file mode 120000
index 00000000000..80316c81614
--- /dev/null
+++ b/egs2/covost2/asr1/local/process_tsv.py
@@ -0,0 +1 @@
+../../../../egs/covost2/asr1/local/process_tsv.py
\ No newline at end of file
diff --git a/egs2/covost2/asr1/path.sh b/egs2/covost2/asr1/path.sh
new file mode 120000
index 00000000000..c9ac0a75bc6
--- /dev/null
+++ b/egs2/covost2/asr1/path.sh
@@ -0,0 +1 @@
+../../TEMPLATE/asr1/path.sh
\ No newline at end of file
diff --git a/egs2/covost2/asr1/pyscripts b/egs2/covost2/asr1/pyscripts
new file mode 120000
index 00000000000..ac68ad75b60
--- /dev/null
+++ b/egs2/covost2/asr1/pyscripts
@@ -0,0 +1 @@
+../../TEMPLATE/asr1/pyscripts
\ No newline at end of file
diff --git a/egs2/covost2/asr1/run.sh b/egs2/covost2/asr1/run.sh
new file mode 100755
index 00000000000..e40acf30dd7
--- /dev/null
+++ b/egs2/covost2/asr1/run.sh
@@ -0,0 +1,99 @@
+#!/usr/bin/env 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
+
+# language related
+src_lang=es
+tgt_lang=en
+# English (en)
+# French (fr)
+# German (de)
+# Spanish (es)
+# Catalan (ca)
+# Italian (it)
+# Russian (ru)
+# Chinese (zh-CN)
+# Portuguese (pt)
+# Persian (fa)
+# Estonian (et)
+# Mongolian (mn)
+# Dutch (nl)
+# Turkish (tr)
+# Arabic (ar)
+# Swedish (sv-SE)
+# Latvian (lv)
+# Slovenian (sl)
+# Tamil (ta)
+# Japanese (ja)
+# Indonesian (id)
+# Welsh (cy)
+nbpe=1000
+
+train_set=train.${src_lang}-${tgt_lang}
+train_dev=dev.${src_lang}-${tgt_lang}
+test_set="test.${src_lang}-${tgt_lang} dev.${src_lang}-${tgt_lang} "
+
+asr_config=conf/train_asr.yaml
+lm_config=conf/train_lm.yaml
+inference_config=conf/decode_asr.yaml
+
+# verify language directions
+is_exist=false
+is_low_resource=false
+if [[ ${src_lang} == en ]]; then
+    tgt_langs=de_ca_zh-CN_fa_et_mn_tr_ar_sv-SE_lv_sl_ta_ja_id_cy
+    for lang in $(echo ${tgt_langs} | tr '_' ' '); do
+        if [[ ${lang} == "${tgt_lang}" ]]; then
+            is_exist=true
+            break
+        fi
+    done
+else
+    lr_src_langs=it_ru_zh-CN_pt_fa_et_mn_nl_tr_ar_sv-SE_lv_sl_ta_ja_id_cy
+    for lang in $(echo ${lr_src_langs} | tr '_' ' '); do
+        if [[ ${lang} == "${src_lang}" ]]; then
+            is_low_resource=true
+            break
+        fi
+    done
+    src_langs=fr_de_es_ca_it_ru_zh-CN_pt_fa_et_mn_nl_tr_ar_sv-SE_lv_sl_ta_ja_id_cy
+    for lang in $(echo ${src_langs} | tr '_' ' '); do
+        if [[ ${lang} == "${src_lang}" ]]; then
+            is_exist=true
+            break
+        fi
+    done
+fi
+if [[ ${is_exist} == false ]]; then
+    echo "No language direction: ${src_lang} to ${tgt_lang}" && exit 1;
+fi
+
+if [ ${is_low_resource} = true ]; then
+    speed_perturb_factors="0.9 1.0 1.1"
+else
+    speed_perturb_factors="0.8 0.9 1.0 1.1 1.2"
+fi
+
+if [ ${src_lang} == ja ] || [ ${src_lang} == zh-CN ]; then
+    nbpe=4000
+fi
+
+./asr.sh \
+    --ngpu 1 \
+    --local_data_opts "--stage 0 --src_lang ${src_lang} --tgt_lang ${tgt_lang}" \
+    --use_lm false \
+    --lm_config "${lm_config}" \
+    --token_type bpe \
+    --nbpe $nbpe \
+    --feats_type raw \
+    --speed_perturb_factors "${speed_perturb_factors}" \
+    --asr_config "${asr_config}" \
+    --inference_config "${inference_config}" \
+    --train_set "${train_set}" \
+    --valid_set "${train_dev}" \
+    --test_sets "${test_set}" \
+    --bpe_train_text "data/${train_set}/text" \
+    --lm_train_text "data/${train_set}/text" "$@"
diff --git a/egs2/covost2/asr1/scripts b/egs2/covost2/asr1/scripts
new file mode 120000
index 00000000000..b25829705dc
--- /dev/null
+++ b/egs2/covost2/asr1/scripts
@@ -0,0 +1 @@
+../../TEMPLATE/asr1/scripts
\ No newline at end of file
diff --git a/egs2/covost2/asr1/steps b/egs2/covost2/asr1/steps
new file mode 120000
index 00000000000..91f2d234e20
--- /dev/null
+++ b/egs2/covost2/asr1/steps
@@ -0,0 +1 @@
+../../../tools/kaldi/egs/wsj/s5/steps
\ No newline at end of file
diff --git a/egs2/covost2/asr1/utils b/egs2/covost2/asr1/utils
new file mode 120000
index 00000000000..f49247da827
--- /dev/null
+++ b/egs2/covost2/asr1/utils
@@ -0,0 +1 @@
+../../../tools/kaldi/egs/wsj/s5/utils
\ No newline at end of file
diff --git a/egs2/covost2/mt1/cmd.sh b/egs2/covost2/mt1/cmd.sh
new file mode 100644
index 00000000000..2aae6919fef
--- /dev/null
+++ b/egs2/covost2/mt1/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
diff --git a/egs2/covost2/mt1/conf/decode_mt.yaml b/egs2/covost2/mt1/conf/decode_mt.yaml
new file mode 100644
index 00000000000..6570a89920d
--- /dev/null
+++ b/egs2/covost2/mt1/conf/decode_mt.yaml
@@ -0,0 +1,5 @@
+beam_size: 5
+lm_weight: 0.0
+maxlenratio: 1.6
+minlenratio: 0.0
+penalty: 0.2
diff --git a/egs2/covost2/mt1/conf/pbs.conf b/egs2/covost2/mt1/conf/pbs.conf
new file mode 100644
index 00000000000..119509938ce
--- /dev/null
+++ b/egs2/covost2/mt1/conf/pbs.conf
@@ -0,0 +1,11 @@
+# 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
diff --git a/egs2/covost2/mt1/conf/queue.conf b/egs2/covost2/mt1/conf/queue.conf
new file mode 100644
index 00000000000..500582fab31
--- /dev/null
+++ b/egs2/covost2/mt1/conf/queue.conf
@@ -0,0 +1,12 @@
+# 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 g.q
diff --git a/egs2/covost2/mt1/conf/slurm.conf b/egs2/covost2/mt1/conf/slurm.conf
new file mode 100644
index 00000000000..3b229673638
--- /dev/null
+++ b/egs2/covost2/mt1/conf/slurm.conf
@@ -0,0 +1,14 @@
+# 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.
diff --git a/egs2/covost2/mt1/conf/train_mt.yaml b/egs2/covost2/mt1/conf/train_mt.yaml
new file mode 120000
index 00000000000..050cda0e4d0
--- /dev/null
+++ b/egs2/covost2/mt1/conf/train_mt.yaml
@@ -0,0 +1 @@
+tuning/train_mt_transformer_lr3e-3_warmup10k_share_enc_dec_input_dropout0.3.yaml
\ No newline at end of file
diff --git a/egs2/covost2/mt1/conf/tuning/train_mt_transformer_lr3e-3_warmup10k_share_enc_dec_input_dropout0.3.yaml b/egs2/covost2/mt1/conf/tuning/train_mt_transformer_lr3e-3_warmup10k_share_enc_dec_input_dropout0.3.yaml
new file mode 100644
index 00000000000..8b2d8844238
--- /dev/null
+++ b/egs2/covost2/mt1/conf/tuning/train_mt_transformer_lr3e-3_warmup10k_share_enc_dec_input_dropout0.3.yaml
@@ -0,0 +1,59 @@
+frontend: embed     # embedding + positional encoding
+frontend_conf:
+    embed_dim: 512
+    positional_dropout_rate: 0.3
+
+encoder: transformer
+encoder_conf:
+    output_size: 512
+    attention_heads: 4
+    linear_units: 1024
+    num_blocks: 6
+    dropout_rate: 0.3
+    positional_dropout_rate: 0.3
+    attention_dropout_rate: 0.3
+    input_layer: null
+    normalize_before: true
+
+decoder: transformer
+decoder_conf:
+    attention_heads: 4
+    linear_units: 1024
+    num_blocks: 6
+    dropout_rate: 0.3
+    positional_dropout_rate: 0.3
+    self_attention_dropout_rate: 0.3
+    src_attention_dropout_rate: 0.3
+
+model_conf:
+    lsm_weight: 0.1
+    length_normalized_loss: false
+    share_decoder_input_output_embed: false
+    share_encoder_decoder_input_embed: true
+
+num_att_plot: 1
+log_interval: 100
+num_workers: 2
+batch_type: numel
+batch_bins: 400000000
+accum_grad: 1
+max_epoch: 200
+patience: none
+init: none
+best_model_criterion:
+-   - valid
+    - acc
+    - max
+keep_nbest_models: 10
+
+optim: adam
+optim_conf:
+    lr: 0.003
+    betas:
+    - 0.9
+    - 0.98
+    eps: 0.000000001
+    weight_decay: 0.0001
+scheduler: warmuplr
+scheduler_conf:
+    warmup_steps: 10000
diff --git a/egs2/covost2/mt1/db.sh b/egs2/covost2/mt1/db.sh
new file mode 120000
index 00000000000..a11c0666fa1
--- /dev/null
+++ b/egs2/covost2/mt1/db.sh
@@ -0,0 +1 @@
+../../TEMPLATE/mt1/db.sh
\ No newline at end of file
diff --git a/egs2/covost2/mt1/local/data.sh b/egs2/covost2/mt1/local/data.sh
new file mode 120000
index 00000000000..18a913cdc75
--- /dev/null
+++ b/egs2/covost2/mt1/local/data.sh
@@ -0,0 +1 @@
+../../st1/local/data.sh
\ No newline at end of file
diff --git a/egs2/covost2/mt1/local/data_prep_commonvoice.pl b/egs2/covost2/mt1/local/data_prep_commonvoice.pl
new file mode 120000
index 00000000000..67d2c51579c
--- /dev/null
+++ b/egs2/covost2/mt1/local/data_prep_commonvoice.pl
@@ -0,0 +1 @@
+../../st1/local/data_prep_commonvoice.pl
\ No newline at end of file
diff --git a/egs2/covost2/mt1/local/data_prep_covost2.sh b/egs2/covost2/mt1/local/data_prep_covost2.sh
new file mode 120000
index 00000000000..21b916c502a
--- /dev/null
+++ b/egs2/covost2/mt1/local/data_prep_covost2.sh
@@ -0,0 +1 @@
+../../st1/local/data_prep_covost2.sh
\ No newline at end of file
diff --git a/egs2/covost2/mt1/local/download_and_untar_commonvoice.sh b/egs2/covost2/mt1/local/download_and_untar_commonvoice.sh
new file mode 120000
index 00000000000..6979fcaffea
--- /dev/null
+++ b/egs2/covost2/mt1/local/download_and_untar_commonvoice.sh
@@ -0,0 +1 @@
+../../st1/local/download_and_untar_commonvoice.sh
\ No newline at end of file
diff --git a/egs2/covost2/mt1/local/path.sh b/egs2/covost2/mt1/local/path.sh
new file mode 120000
index 00000000000..476483f066c
--- /dev/null
+++ b/egs2/covost2/mt1/local/path.sh
@@ -0,0 +1 @@
+../../st1/local/path.sh
\ No newline at end of file
diff --git a/egs2/covost2/mt1/local/process_tsv.py b/egs2/covost2/mt1/local/process_tsv.py
new file mode 120000
index 00000000000..8ec19bcb081
--- /dev/null
+++ b/egs2/covost2/mt1/local/process_tsv.py
@@ -0,0 +1 @@
+../../st1/local/process_tsv.py
\ No newline at end of file
diff --git a/egs2/covost2/mt1/mt.sh b/egs2/covost2/mt1/mt.sh
new file mode 120000
index 00000000000..9f4c1d5c0bb
--- /dev/null
+++ b/egs2/covost2/mt1/mt.sh
@@ -0,0 +1 @@
+../../TEMPLATE/mt1/mt.sh
\ No newline at end of file
diff --git a/egs2/covost2/mt1/path.sh b/egs2/covost2/mt1/path.sh
new file mode 120000
index 00000000000..a2d87d29a46
--- /dev/null
+++ b/egs2/covost2/mt1/path.sh
@@ -0,0 +1 @@
+../../TEMPLATE/mt1/path.sh
\ No newline at end of file
diff --git a/egs2/covost2/mt1/pyscripts b/egs2/covost2/mt1/pyscripts
new file mode 120000
index 00000000000..bca5bde44f3
--- /dev/null
+++ b/egs2/covost2/mt1/pyscripts
@@ -0,0 +1 @@
+../../TEMPLATE/mt1/pyscripts
\ No newline at end of file
diff --git a/egs2/covost2/mt1/run.sh b/egs2/covost2/mt1/run.sh
new file mode 100755
index 00000000000..11265d7ec87
--- /dev/null
+++ b/egs2/covost2/mt1/run.sh
@@ -0,0 +1,99 @@
+#!/usr/bin/env 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
+
+# language related
+src_lang=es
+tgt_lang=en
+# English (en)
+# French (fr)
+# German (de)
+# Spanish (es)
+# Catalan (ca)
+# Italian (it)
+# Russian (ru)
+# Chinese (zh-CN)
+# Portuguese (pt)
+# Persian (fa)
+# Estonian (et)
+# Mongolian (mn)
+# Dutch (nl)
+# Turkish (tr)
+# Arabic (ar)
+# Swedish (sv-SE)
+# Latvian (lv)
+# Slovenian (sl)
+# Tamil (ta)
+# Japanese (ja)
+# Indonesian (id)
+# Welsh (cy)
+
+src_nbpe=1000
+tgt_nbpe=1000
+src_case=lc.rm
+tgt_case=lc.rm
+
+train_set=train.${src_lang}-${tgt_lang}
+train_dev=dev.${src_lang}-${tgt_lang}
+test_set="test.${src_lang}-${tgt_lang} dev.${src_lang}-${tgt_lang} "
+
+mt_config=conf/train_mt.yaml
+inference_config=conf/decode_mt.yaml
+
+# verify language directions
+is_exist=false
+if [[ ${src_lang} == en ]]; then
+    tgt_langs=de_ca_zh-CN_fa_et_mn_tr_ar_sv-SE_lv_sl_ta_ja_id_cy
+    for lang in $(echo ${tgt_langs} | tr '_' ' '); do
+        if [[ ${lang} == "${tgt_lang}" ]]; then
+            is_exist=true
+            break
+        fi
+    done
+else
+    src_langs=fr_de_es_ca_it_ru_zh-CN_pt_fa_et_mn_nl_tr_ar_sv-SE_lv_sl_ta_ja_id_cy
+    for lang in $(echo ${src_langs} | tr '_' ' '); do
+        if [[ ${lang} == "${src_lang}" ]]; then
+            is_exist=true
+            break
+        fi
+    done
+fi
+if [[ ${is_exist} == false ]]; then
+    echo "No language direction: ${src_lang} to ${tgt_lang}" && exit 1;
+fi
+
+if [ ${src_lang} == ja ] || [ ${src_lang} == zh-CN ]; then
+    src_nbpe=4000
+fi
+
+if [ ${tgt_lang} == ja ] || [ ${tgt_lang} == zh-CN ]; then
+    tgt_nbpe=4000
+fi
+
+./mt.sh \
+    --ngpu 1 \
+    --local_data_opts "--stage 0 --src_lang ${src_lang} --tgt_lang ${tgt_lang}" \
+    --use_lm false \
+    --feats_type raw \
+    --audio_format "flac.ark" \
+    --token_joint false \
+    --src_lang ${src_lang} \
+    --tgt_lang ${tgt_lang} \
+    --src_token_type "bpe" \
+    --src_nbpe $src_nbpe \
+    --tgt_token_type "bpe" \
+    --tgt_nbpe $tgt_nbpe \
+    --src_case ${src_case} \
+    --tgt_case ${tgt_case} \
+    --mt_config "${mt_config}" \
+    --inference_config "${inference_config}" \
+    --train_set "${train_set}" \
+    --valid_set "${train_dev}" \
+    --test_sets "${test_set}" \
+    --src_bpe_train_text "data/${train_set}/text.${src_case}.${src_lang}" \
+    --tgt_bpe_train_text "data/${train_set}/text.${tgt_case}.${tgt_lang}" \
+    --lm_train_text "data/${train_set}/text.${tgt_case}.${tgt_lang}"  "$@"
diff --git a/egs2/covost2/mt1/scripts b/egs2/covost2/mt1/scripts
new file mode 120000
index 00000000000..1c11b3c3c7b
--- /dev/null
+++ b/egs2/covost2/mt1/scripts
@@ -0,0 +1 @@
+../../TEMPLATE/mt1/scripts
\ No newline at end of file
diff --git a/egs2/covost2/mt1/steps b/egs2/covost2/mt1/steps
new file mode 120000
index 00000000000..91f2d234e20
--- /dev/null
+++ b/egs2/covost2/mt1/steps
@@ -0,0 +1 @@
+../../../tools/kaldi/egs/wsj/s5/steps
\ No newline at end of file
diff --git a/egs2/covost2/mt1/utils b/egs2/covost2/mt1/utils
new file mode 120000
index 00000000000..f49247da827
--- /dev/null
+++ b/egs2/covost2/mt1/utils
@@ -0,0 +1 @@
+../../../tools/kaldi/egs/wsj/s5/utils
\ No newline at end of file
diff --git a/egs2/covost2/st1/cmd.sh b/egs2/covost2/st1/cmd.sh
new file mode 100644
index 00000000000..2aae6919fef
--- /dev/null
+++ b/egs2/covost2/st1/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
diff --git a/egs2/covost2/st1/conf/decode_st.yaml b/egs2/covost2/st1/conf/decode_st.yaml
new file mode 100644
index 00000000000..2967ee6fc0f
--- /dev/null
+++ b/egs2/covost2/st1/conf/decode_st.yaml
@@ -0,0 +1,5 @@
+batch_size: 1
+beam_size: 10
+nbest: 1
+lm_weight: 0.0
+
diff --git a/egs2/covost2/st1/conf/fbank.conf b/egs2/covost2/st1/conf/fbank.conf
new file mode 100644
index 00000000000..d75ddde4df8
--- /dev/null
+++ b/egs2/covost2/st1/conf/fbank.conf
@@ -0,0 +1,2 @@
+--sample-frequency=8000 
+--num-mel-bins=80
diff --git a/egs2/covost2/st1/conf/pbs.conf b/egs2/covost2/st1/conf/pbs.conf
new file mode 100644
index 00000000000..119509938ce
--- /dev/null
+++ b/egs2/covost2/st1/conf/pbs.conf
@@ -0,0 +1,11 @@
+# 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
diff --git a/egs2/covost2/st1/conf/pitch.conf b/egs2/covost2/st1/conf/pitch.conf
new file mode 100644
index 00000000000..926bcfca92a
--- /dev/null
+++ b/egs2/covost2/st1/conf/pitch.conf
@@ -0,0 +1 @@
+--sample-frequency=8000
diff --git a/egs2/covost2/st1/conf/queue.conf b/egs2/covost2/st1/conf/queue.conf
new file mode 100644
index 00000000000..500582fab31
--- /dev/null
+++ b/egs2/covost2/st1/conf/queue.conf
@@ -0,0 +1,12 @@
+# 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 g.q
diff --git a/egs2/covost2/st1/conf/slurm.conf b/egs2/covost2/st1/conf/slurm.conf
new file mode 100644
index 00000000000..3b229673638
--- /dev/null
+++ b/egs2/covost2/st1/conf/slurm.conf
@@ -0,0 +1,14 @@
+# 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.
diff --git a/egs2/covost2/st1/conf/train_st.yaml b/egs2/covost2/st1/conf/train_st.yaml
new file mode 120000
index 00000000000..9e97e84c995
--- /dev/null
+++ b/egs2/covost2/st1/conf/train_st.yaml
@@ -0,0 +1 @@
+tuning/train_transformer_st.yaml
\ No newline at end of file
diff --git a/egs2/covost2/st1/conf/tuning/train_transformer_st.yaml b/egs2/covost2/st1/conf/tuning/train_transformer_st.yaml
new file mode 100644
index 00000000000..168a6a7a174
--- /dev/null
+++ b/egs2/covost2/st1/conf/tuning/train_transformer_st.yaml
@@ -0,0 +1,87 @@
+# network architecture
+
+# frontend related
+frontend: default
+frontend_conf:
+    n_fft: 512
+    win_length: 400
+    hop_length: 160
+
+# encoder related
+encoder: transformer
+encoder_conf:
+    input_layer: conv2d
+    num_blocks: 12
+    linear_units: 2048
+    dropout_rate: 0.1
+    output_size: 256  # dimension of attention
+    attention_heads: 4
+    attention_dropout_rate: 0.0
+
+# decoder related
+decoder: transformer
+decoder_conf:
+    input_layer: embed
+    num_blocks: 6
+    linear_units: 2048
+    dropout_rate: 0.1
+
+extra_asr_decoder: transformer
+extra_asr_decoder_conf:
+    input_layer: embed
+    num_blocks: 2
+    linear_units: 2048
+    dropout_rate: 0.1
+
+extra_mt_decoder: transformer
+extra_mt_decoder_conf:
+    input_layer: embed
+    num_blocks: 2
+    linear_units: 2048
+    dropout_rate: 0.1
+
+# loss related
+model_conf:
+    asr_weight: 0.3
+    mt_weight: 0.0
+    mtlalpha: 1.0
+    lsm_weight: 0.1
+    length_normalized_loss: false
+
+# optimization related
+optim: adam
+accum_grad: 2
+grad_clip: 3
+max_epoch: 50
+optim_conf:
+    lr: 2.5
+scheduler: noamlr
+scheduler_conf:
+    model_size: 256
+    warmup_steps: 25000
+
+# minibatch related
+batch_type: folded
+batch_bins: 64
+
+best_model_criterion:
+-   - valid
+    - acc
+    - max
+keep_nbest_models: 10
+
+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
diff --git a/egs2/covost2/st1/db.sh b/egs2/covost2/st1/db.sh
new file mode 120000
index 00000000000..39ab78266e5
--- /dev/null
+++ b/egs2/covost2/st1/db.sh
@@ -0,0 +1 @@
+../../TEMPLATE/st1/db.sh
\ No newline at end of file
diff --git a/egs2/covost2/st1/local/data.sh b/egs2/covost2/st1/local/data.sh
new file mode 100755
index 00000000000..ca0912e743d
--- /dev/null
+++ b/egs2/covost2/st1/local/data.sh
@@ -0,0 +1,78 @@
+#!/usr/bin/env bash
+
+# Copyright 2021 Carnegie Mellon University (Jiatong Shi)
+#  Apache 2.0  (http://www.apache.org/licenses/LICENSE-2.0)
+
+. ./path.sh || exit 1;
+. ./cmd.sh || exit 1;
+. ./db.sh || exit 1;
+
+# general configuration
+stage=0       # start from 0 if you need to start from data preparation
+stop_stage=100
+SECONDS=0
+src_lang=es
+tgt_lang=en
+
+ . utils/parse_options.sh || exit 1;
+
+
+log() {
+    local fname=${BASH_SOURCE[1]##*/}
+    echo -e "$(date '+%Y-%m-%dT%H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
+}
+
+if [ -z "${COVOST2}" ]; then
+    log "Fill the value of 'COVOST2' of db.sh"
+    exit 1
+fi
+mkdir -p ${COVOST2}
+
+if [ -z "${COMMONVOICE}" ]; then
+    log "Fill the value of 'COMMONVOICE' of db.sh"
+    exit 1
+fi
+mkdir -p ${COMMONVOICE}
+
+# 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
+
+
+
+if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
+    log "stage 0: Data Downloading"
+
+    # base url for downloads.
+    data_url=https://voice-prod-bundler-ee1969a6ce8178826482b88e843c335139bd3fb4.s3.amazonaws.com/cv-corpus-4-2019-12-10/${src_lang}.tar.gz
+
+    # Download CommonVoice
+    mkdir -p ${COMMONVOICE}/${src_lang}
+    local/download_and_untar_commonvoice.sh ${COMMONVOICE}/${src_lang} ${data_url} ${src_lang}.tar.gz
+
+    # Download translation
+    if [[ ${src_lang} != en ]]; then
+        wget --no-check-certificate https://dl.fbaipublicfiles.com/covost/covost_v2.${src_lang}_${tgt_lang}.tsv.tar.gz \
+            -P ${COVOST2}
+        tar -xzf ${COVOST2}/covost_v2.${src_lang}_${tgt_lang}.tsv.tar.gz -C ${COVOST2}
+    fi
+    wget --no-check-certificate https://dl.fbaipublicfiles.com/covost/covost2.zip \
+          -P ${COVOST2}
+    unzip ${COVOST2}/covost2.zip -d ${COVOST2}
+    # NOTE: some non-English target languages lack translation from English
+fi
+
+if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
+    log "stage 1: Data Preparation"
+    # use underscore-separated names in data directories.
+    local/data_prep_commonvoice.pl "${COMMONVOICE}/${src_lang}" validated data/validated.${src_lang}
+
+    # text preprocessing (tokenization, case, punctuation marks etc.)
+    local/data_prep_covost2.sh ${COVOST2} ${src_lang} ${tgt_lang} || exit 1;
+    # NOTE: train/dev/test splits are different from original CommonVoice
+fi
+
+
+log "Successfully finished. [elapsed=${SECONDS}s]"
diff --git a/egs2/covost2/st1/local/data_prep_commonvoice.pl b/egs2/covost2/st1/local/data_prep_commonvoice.pl
new file mode 120000
index 00000000000..5e053ed441f
--- /dev/null
+++ b/egs2/covost2/st1/local/data_prep_commonvoice.pl
@@ -0,0 +1 @@
+../../asr1/local/data_prep_commonvoice.pl
\ No newline at end of file
diff --git a/egs2/covost2/st1/local/data_prep_covost2.sh b/egs2/covost2/st1/local/data_prep_covost2.sh
new file mode 120000
index 00000000000..bc9b98f915d
--- /dev/null
+++ b/egs2/covost2/st1/local/data_prep_covost2.sh
@@ -0,0 +1 @@
+../../asr1/local/data_prep_covost2.sh
\ No newline at end of file
diff --git a/egs2/covost2/st1/local/download_and_untar_commonvoice.sh b/egs2/covost2/st1/local/download_and_untar_commonvoice.sh
new file mode 120000
index 00000000000..021e6f9a4f1
--- /dev/null
+++ b/egs2/covost2/st1/local/download_and_untar_commonvoice.sh
@@ -0,0 +1 @@
+../../asr1/local/download_and_untar_commonvoice.sh
\ No newline at end of file
diff --git a/egs2/covost2/st1/local/path.sh b/egs2/covost2/st1/local/path.sh
new file mode 120000
index 00000000000..d404bdb839c
--- /dev/null
+++ b/egs2/covost2/st1/local/path.sh
@@ -0,0 +1 @@
+../../asr1/local/path.sh
\ No newline at end of file
diff --git a/egs2/covost2/st1/local/process_tsv.py b/egs2/covost2/st1/local/process_tsv.py
new file mode 120000
index 00000000000..2710e325be5
--- /dev/null
+++ b/egs2/covost2/st1/local/process_tsv.py
@@ -0,0 +1 @@
+../../asr1/local/process_tsv.py
\ No newline at end of file
diff --git a/egs2/covost2/st1/path.sh b/egs2/covost2/st1/path.sh
new file mode 120000
index 00000000000..8e43dca7d4d
--- /dev/null
+++ b/egs2/covost2/st1/path.sh
@@ -0,0 +1 @@
+../../TEMPLATE/st1/path.sh
\ No newline at end of file
diff --git a/egs2/covost2/st1/pyscripts b/egs2/covost2/st1/pyscripts
new file mode 120000
index 00000000000..f8eb6b2ab7b
--- /dev/null
+++ b/egs2/covost2/st1/pyscripts
@@ -0,0 +1 @@
+../../TEMPLATE/st1/pyscripts
\ No newline at end of file
diff --git a/egs2/covost2/st1/run.sh b/egs2/covost2/st1/run.sh
new file mode 100755
index 00000000000..5cd66dbaf53
--- /dev/null
+++ b/egs2/covost2/st1/run.sh
@@ -0,0 +1,114 @@
+#!/usr/bin/env 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
+
+# language related
+src_lang=es
+tgt_lang=en
+# English (en)
+# French (fr)
+# German (de)
+# Spanish (es)
+# Catalan (ca)
+# Italian (it)
+# Russian (ru)
+# Chinese (zh-CN)
+# Portuguese (pt)
+# Persian (fa)
+# Estonian (et)
+# Mongolian (mn)
+# Dutch (nl)
+# Turkish (tr)
+# Arabic (ar)
+# Swedish (sv-SE)
+# Latvian (lv)
+# Slovenian (sl)
+# Tamil (ta)
+# Japanese (ja)
+# Indonesian (id)
+# Welsh (cy)
+
+src_nbpe=1000
+tgt_nbpe=1000
+src_case=lc.rm
+tgt_case=lc.rm
+
+train_set=train.${src_lang}-${tgt_lang}
+train_dev=dev.${src_lang}-${tgt_lang}
+test_set="test.${src_lang}-${tgt_lang} dev.${src_lang}-${tgt_lang} "
+
+st_config=conf/train_st.yaml
+inference_config=conf/decode_st.yaml
+
+# verify language directions
+is_exist=false
+is_low_resource=false
+if [[ ${src_lang} == en ]]; then
+    tgt_langs=de_ca_zh-CN_fa_et_mn_tr_ar_sv-SE_lv_sl_ta_ja_id_cy
+    for lang in $(echo ${tgt_langs} | tr '_' ' '); do
+        if [[ ${lang} == "${tgt_lang}" ]]; then
+            is_exist=true
+            break
+        fi
+    done
+else
+    lr_src_langs=it_ru_zh-CN_pt_fa_et_mn_nl_tr_ar_sv-SE_lv_sl_ta_ja_id_cy
+    for lang in $(echo ${lr_src_langs} | tr '_' ' '); do
+        if [[ ${lang} == "${src_lang}" ]]; then
+            is_low_resource=true
+            break
+        fi
+    done
+    src_langs=fr_de_es_ca_it_ru_zh-CN_pt_fa_et_mn_nl_tr_ar_sv-SE_lv_sl_ta_ja_id_cy
+    for lang in $(echo ${src_langs} | tr '_' ' '); do
+        if [[ ${lang} == "${src_lang}" ]]; then
+            is_exist=true
+            break
+        fi
+    done
+fi
+if [[ ${is_exist} == false ]]; then
+    echo "No language direction: ${src_lang} to ${tgt_lang}" && exit 1;
+fi
+
+if [ ${is_low_resource} = true ]; then
+    speed_perturb_factors="0.9 1.0 1.1"
+else
+    speed_perturb_factors="0.8 0.9 1.0 1.1 1.2"
+fi
+
+if [ ${src_lang} == ja ] || [ ${src_lang} == zh-CN ]; then
+    src_nbpe=4000
+fi
+
+if [ ${tgt_lang} == ja ] || [ ${tgt_lang} == zh-CN ]; then
+    tgt_nbpe=4000
+fi
+
+./st.sh \
+    --ngpu 1 \
+    --local_data_opts "--stage 0 --src_lang ${src_lang} --tgt_lang ${tgt_lang}" \
+    --speed_perturb_factors "${speed_perturb_factors}" \
+    --use_lm false \
+    --feats_type raw \
+    --audio_format "flac.ark" \
+    --token_joint false \
+    --src_lang ${src_lang} \
+    --tgt_lang ${tgt_lang} \
+    --src_token_type "bpe" \
+    --src_nbpe $src_nbpe \
+    --tgt_token_type "bpe" \
+    --tgt_nbpe $tgt_nbpe \
+    --src_case ${src_case} \
+    --tgt_case ${tgt_case} \
+    --st_config "${st_config}" \
+    --inference_config "${inference_config}" \
+    --train_set "${train_set}" \
+    --valid_set "${train_dev}" \
+    --test_sets "${test_set}" \
+    --src_bpe_train_text "data/${train_set}/text.${src_case}.${src_lang}" \
+    --tgt_bpe_train_text "data/${train_set}/text.${tgt_case}.${tgt_lang}" \
+    --lm_train_text "data/${train_set}/text.${tgt_case}.${tgt_lang}"  "$@"
diff --git a/egs2/covost2/st1/scripts b/egs2/covost2/st1/scripts
new file mode 120000
index 00000000000..c5b75bb5b0a
--- /dev/null
+++ b/egs2/covost2/st1/scripts
@@ -0,0 +1 @@
+../../TEMPLATE/st1/scripts
\ No newline at end of file
diff --git a/egs2/covost2/st1/st.sh b/egs2/covost2/st1/st.sh
new file mode 120000
index 00000000000..5c7465739e3
--- /dev/null
+++ b/egs2/covost2/st1/st.sh
@@ -0,0 +1 @@
+../../TEMPLATE/st1/st.sh
\ No newline at end of file
diff --git a/egs2/covost2/st1/steps b/egs2/covost2/st1/steps
new file mode 120000
index 00000000000..91f2d234e20
--- /dev/null
+++ b/egs2/covost2/st1/steps
@@ -0,0 +1 @@
+../../../tools/kaldi/egs/wsj/s5/steps
\ No newline at end of file
diff --git a/egs2/covost2/st1/utils b/egs2/covost2/st1/utils
new file mode 120000
index 00000000000..f49247da827
--- /dev/null
+++ b/egs2/covost2/st1/utils
@@ -0,0 +1 @@
+../../../tools/kaldi/egs/wsj/s5/utils
\ No newline at end of file