[Recipe] Add iwslt22 low resource speech translation task for egs2

egs2/README.md

@@ -63,6 +63,7 @@ See: https://espnet.github.io/espnet/espnet2_tutorial.html#recipes-using-espnet2
 | iwslt14                 | IWSLT14 MT shared task                                                                                                           | MT                      | DEU->ENG             | http://dl.fbaipublicfiles.com/fairseq/data/iwslt14/de-en.tgz                                                  |              |
 | iwslt21_low_resource    | ALFFA, IARPA Babel, Gamayun, IWSLT 2021                                                                                          | ASR                     | SWA                   | http://www.openslr.org/25/ https://catalog.ldc.upenn.edu/LDC2017S05 https://gamayun.translatorswb.org/data/ https://iwslt.org/2021/low-resource |              |
 | iwslt22_dialect         | IWSLT2022 dialectal speech translation shared task                                                                               | ASR/ST                  | ARA->Tunisian ARA     | https://github.com/kevinduh/iwslt22-dialect.git                                                              |              |
+| iwslt22_low_resource | IWSLT2022 Low-resource speech translation track task                                                                               | ST                  | Tamasheq->FrenchPermalink     | https://github.com/mzboito/IWSLT2022_Tamasheq_data.git |
 | jdcinal                 | Japanese Dialogue Corpus of Information Navigation and Attentive Listening Annotated with Extended ISO-24617-2 Dialogue Act Tags | SLU                     | JPN                   | http://www.lrec-conf.org/proceedings/lrec2018/pdf/464.pdf http://tts.speech.cs.cmu.edu/awb/infomation_navigation_and_attentive_listening_0.2.zip |              |
 | jkac                    | J-KAC: Japanese Kamishibai and audiobook corpus                                                                                  | TTS                     | JPN                  | https://sites.google.com/site/shinnosuketakamichi/research-topics/j-kac_corpus                               |              |
 | jmd                     | JMD: Japanese multi-dialect corpus for speech synthesis                                                                          | TTS                     | JPN                  | https://sites.google.com/site/shinnosuketakamichi/research-topics/jmd_corpus                                 |              |

egs2/TEMPLATE/asr1/db.sh

@@ -137,6 +137,7 @@ CMU_ARCTIC=downloads
 CMU_INDIC=downloads
 INDIC_SPEECH=downloads
 IWSLT22_DIALECT=
+IWSLT22_LOW_RESOURCE=downloads
 JKAC=
 MUCS_SUBTASK1=downloads
 MUCS_SUBTASK2=downloads

egs2/iwslt22_low_resource/st1/RESULTS.md

@@ -0,0 +1,26 @@
+<!-- Generated by scripts/utils/show_translation_result.sh -->
+# RESULTS
+## Environments
+- date: `Fri Mar 10 16:10:20 CST 2023`
+- python version: `3.8.16 (default, Jan 17 2023, 23:13:24)  [GCC 11.2.0]`
+- espnet version: `espnet 202301`
+- pytorch version: `pytorch 1.13.1`
+- Git hash: `ff841366229d539eb74d23ac999cae7c0cc62cad`
+  - Commit date: `Mon Feb 20 12:23:15 2023 -0500`
+
+## st_wav2vec-transformer-warmup-15k
+
+### BLEU
+
+|dataset|score|verbose_score|
+|---|---|---|
+|decode_pen2_st_model_valid.acc.ave/test|2.6|22.5/4.5/1.8/0.8 (BP = 0.736 ratio = 0.765 hyp_len = 17223 ref_len = 22504)|
+
+## st_full_wav2vec-transformer-warmup-15k
+
+### BLEU
+
+|dataset|score|verbose_score|
+|---|---|---|
+|decode_pen2_st_model_valid.acc.ave/test|3.6|24.7/5.4/2.1/1.0 (BP = 0.894 ratio = 0.899 hyp_len = 20241 ref_len = 22504)|
+

egs2/iwslt22_low_resource/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

egs2/iwslt22_low_resource/st1/conf/decode_st_transformer.yaml

@@ -0,0 +1 @@
+tuning/decode_pen2.yaml

egs2/iwslt22_low_resource/st1/conf/fbank.conf

@@ -0,0 +1,2 @@
+--sample-frequency=16000 
+--num-mel-bins=80

egs2/iwslt22_low_resource/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

egs2/iwslt22_low_resource/st1/conf/pitch.conf

@@ -0,0 +1 @@
+--sample-frequency=16000

egs2/iwslt22_low_resource/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

egs2/iwslt22_low_resource/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.

egs2/iwslt22_low_resource/st1/conf/train_st_transformer.yaml

@@ -0,0 +1 @@
+tuning/train_st_transformer_warmup15k.yaml

egs2/iwslt22_low_resource/st1/conf/tuning/decode_pen2.yaml

@@ -0,0 +1,6 @@
+batch_size: 1
+beam_size: 10
+penalty: 0.2
+maxlenratio: 0.0
+minlenratio: 0.0
+lm_weight: 0.0

egs2/iwslt22_low_resource/st1/conf/tuning/train_st_transformer_warmup15k.yaml

@@ -0,0 +1,81 @@
+batch_type: numel
+batch_bins: 2000000
+accum_grad: 32 # RTX 3090 Ti X 1
+max_epoch: 80
+patience: none
+best_model_criterion:
+-   - valid
+    - acc
+    - max
+keep_nbest_models: 10
+
+# 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
+
+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: 0.1
+    src_attention_dropout_rate: 0.1
+
+model_conf:
+    asr_weight: 0.0
+    mt_weight: 0.0
+    mtlalpha: 0.0
+    lsm_weight: 0.1
+    length_normalized_loss: false
+    extract_feats_in_collect_stats: false
+
+optim: adam
+grad_clip: 3
+optim_conf:
+    lr: 12.5
+scheduler: noamlr
+scheduler_conf:
+    model_size: 256
+    warmup_steps: 15000
+
+frontend: s3prl
+frontend_conf:
+    frontend_conf:
+        upstream: hf_wav2vec2_custom  # Note: If the upstream is changed, please change the input_size in the preencoder.
+        path_or_url: LIA-AvignonUniversity/IWSLT2022-tamasheq-only
+    download_dir: ./hub
+    multilayer_feature: True
+
+preencoder: linear
+preencoder_conf:
+   input_size: 768  # Note: If the upstream is changed, please change this value accordingly.
+   output_size: 80
+
+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
+
+freeze_param: [
+  "frontend.upstream"
+]
+

egs2/iwslt22_low_resource/st1/db.sh

@@ -0,0 +1 @@
+../../TEMPLATE/st1/db.sh

egs2/iwslt22_low_resource/st1/local/data.sh

@@ -0,0 +1,101 @@
+#!/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
+
+. ./db.sh || exit 1;
+. ./path.sh || exit 1;
+. ./cmd.sh || exit 1;
+
+log() {
+    local fname=${BASH_SOURCE[1]##*/}
+    echo -e "$(date '+%Y-%m-%dT%H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
+}
+SECONDS=0
+
+stage=1
+stop_stage=100000
+splits_dir=data/iwslt22_taq_fra_split
+
+log "$0 $*"
+. utils/parse_options.sh
+
+if [ -z "${IWSLT22_LOW_RESOURCE}" ]; then
+    log "Fill the value of 'IWSLT22_LOW_RESOURCE' of db.sh"
+    exit 1
+fi
+
+if [ $# -ne 0 ]; then
+    log "Error: No positional arguments are required."
+    exit 2
+fi
+
+if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ] && [ ! -d "${splits_dir}" ]; then
+    log "stage 1: Official splits from IWSLT Low Resource Speech Translation"
+
+    git clone https://github.com/mzboito/IWSLT2022_Tamasheq_data.git ${splits_dir}
+
+    # train comprises 17 hours of clean speech in Tamasheq, translated to the French language
+    # train_full comprises a 19 hour version of this corpus,
+    # including 2 additional hours of data that was labeled by annotators as potentially noisy
+    mkdir -p data/train/org
+    mkdir -p data/train_full/org
+    mkdir -p data/valid/org
+    mkdir -p data/test/org
+
+    for set in train valid test
+    do
+        cp -r ${splits_dir}/taq_fra_clean/${set}/* data/${set}/org
+    done
+    cp -r ${splits_dir}/taq_fra_full/train/* data/train_full/org
+
+fi
+
+if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
+    log "stage 2: Data Preparation"
+
+    for set in train train_full valid test
+    do
+        python local/preprocess.py --out data/${set} --data data/${set}/org
+
+        cp data/${set}/text.fr data/${set}/text
+
+        utils/utt2spk_to_spk2utt.pl data/${set}/utt2spk > data/${set}/spk2utt
+        utils/fix_data_dir.sh --utt_extra_files "text.fr" data/${set}
+        utils/validate_data_dir.sh --no-feats data/${set} || exit 1
+    done
+fi
+
+if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then
+    log "stage 3: Normalize Transcripts"
+
+    # check extra module installation
+    if ! command -v tokenizer.perl > /dev/null; then
+        echo "Error: it seems that moses is not installed." >&2
+        echo "Error: please install moses as follows." >&2
+        echo "Error: cd ${MAIN_ROOT}/tools && make moses.done" >&2
+        exit 1
+    fi
+
+    for set in train train_full valid test
+    do
+        cut -d ' ' -f 2- data/${set}/text.fr > data/${set}/fr.org
+        cut -d ' ' -f 1 data/${set}/text.fr > data/${set}/uttlist
+
+        # tokenize
+        tokenizer.perl -l fr -q < data/${set}/fr.org > data/${set}/fr.tok
+        paste -d ' ' data/${set}/uttlist data/${set}/fr.tok > data/${set}/text.tc.fr
+
+        # remove empty lines that were previously only punctuation
+        # small to use fix_data_dir as is, where it does reduce lines based on extra files
+        <"data/${set}/text.tc.fr" awk ' { if( NF != 1 ) print $0; } ' >"data/${set}/text"
+        utils/fix_data_dir.sh --utt_extra_files "text.tc.fr text.fr" data/${set}
+        cp data/${set}/text.tc.fr data/${set}/text
+        utils/fix_data_dir.sh --utt_extra_files "text.tc.fr text.fr" data/${set}
+        utils/validate_data_dir.sh --no-feats data/${set} || exit 1
+    done
+fi
+
+log "Successfully finished. [elapsed=${SECONDS}s]"