fix small bugs and add CHiME4 enh_asr1 recipe & results

ci/test_integration_espnet2.sh

@@ -134,6 +134,9 @@ if python3 -c 'import torch as t; from distutils.version import LooseVersion as
     for f in egs2/*/ssl1/conf/train*.yaml; do
         ${python} -m espnet2.bin.hubert_train --config "${f}" --iterator_type none --normalize none --dry_run true --output_dir out --token_list dummy_token_list
     done
+    for f in egs2/*/enh_asr1/conf/train_enh_asr*.yaml; do
+        ${python} -m espnet2.bin.enh_s2t_train --config "${f}" --iterator_type none --dry_run true --output_dir out --token_list dummy_token_list
+    done
 fi
 
 # These files must be same each other.

egs2/TEMPLATE/enh_asr1/scripts/utils/show_enh_score.sh

@@ -0,0 +1,84 @@
+#!/usr/bin/env bash
+mindepth=0
+maxdepth=1
+
+. utils/parse_options.sh
+
+if [ $# -gt 1 ]; then
+    echo "Usage: $0 --mindepth 0 --maxdepth 1 [exp]" 1>&2
+    echo ""
+    echo "Show the system environments and the evaluation results in Markdown format."
+    echo 'The default of <exp> is "exp/".'
+    exit 1
+fi
+
+[ -f ./path.sh ] && . ./path.sh
+set -euo pipefail
+if [ $# -eq 1 ]; then
+    exp=$(realpath "$1")
+else
+    exp=exp
+fi
+
+
+cat << EOF
+<!-- Generated by $0 -->
+# RESULTS
+## Environments
+- date: \`$(LC_ALL=C date)\`
+EOF
+
+python3 << EOF
+import sys, espnet, torch
+pyversion = sys.version.replace('\n', ' ')
+
+print(f"""- python version: \`{pyversion}\`
+- espnet version: \`espnet {espnet.__version__}\`
+- pytorch version: \`pytorch {torch.__version__}\`""")
+EOF
+
+cat << EOF
+- Git hash: \`$(git rev-parse HEAD)\`
+  - Commit date: \`$(git log -1 --format='%cd')\`
+
+EOF
+
+
+while IFS= read -r expdir; do
+    if ls "${expdir}"/*/scoring_enh/result_stoi.txt &> /dev/null; then
+        echo -e "\n## $(basename ${expdir})\n"
+        [ -e "${expdir}"/config.yaml ] && grep ^config "${expdir}"/config.yaml
+        metrics=()
+        heading="\n|dataset|"
+        sep="|---|"
+        for type in pesq estoi stoi sar sdr sir si_snr; do
+            if ls "${expdir}"/*/scoring_enh/result_${type}.txt &> /dev/null; then
+                metrics+=("$type")
+                heading+="${type^^}|"
+                sep+="---|"
+            fi
+        done
+        echo -e "${heading}\n${sep}"
+
+        setnames=()
+        for dirname in "${expdir}"/*/scoring_enh/result_stoi.txt; do
+            dset=$(echo $dirname | sed -e "s#${expdir}/\([^/]*\)/scoring_enh/result_stoi.txt#\1#g")
+            setnames+=("$dset")
+        done
+        for dset in "${setnames[@]}"; do
+            line="|${dset}|"
+            for ((i=0; i<${#metrics[@]}; i++)); do
+                type=${metrics[$i]}
+                if [ -f "${expdir}"/${dset}/scoring_enh/result_${type}.txt ]; then
+                    score=$(head -n1 "${expdir}"/${dset}/scoring_enh/result_${type}.txt)
+                else
+                    score=""
+                fi
+                line+="${score}|"
+            done
+            echo $line
+        done
+        echo ""
+    fi
+
+done < <(find ${exp} -mindepth ${mindepth} -maxdepth ${maxdepth} -type d)

egs2/chime4/enh_asr1/README.md

@@ -0,0 +1,57 @@
+<!-- Generated by scripts/utils/show_asr_result.sh -->
+# RESULTS
+## Environments
+- date: `Thu Apr 28 00:09:17 EDT 2022`
+- python version: `3.7.11 (default, Jul 27 2021, 14:32:16)  [GCC 7.5.0]`
+- espnet version: `espnet 202204`
+- pytorch version: `pytorch 1.8.1`
+- Git hash: `44971ff962aae30c962226f1ba3d87de057ac00e`
+  - Commit date: `Wed Apr 27 10:13:03 2022 -0400`
+- Pretrained model: https://huggingface.co/espnet/simpleoier_chime4_enh_asr_train_enh_asr_convtasnet_fbank_transformer_raw_en_char
+
+## enh_asr_train_enh_asr_convtasnet_fbank_transformer_raw_en_char
+### WER
+
+|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
+|---|---|---|---|---|---|---|---|---|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/dt05_real_beamformit_2mics|1640|27119|93.0|5.2|1.8|0.6|7.7|53.3|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/dt05_real_beamformit_5mics|1640|27119|93.9|4.5|1.6|0.5|6.7|49.9|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/dt05_real_isolated_1ch_track|1640|27119|91.8|6.0|2.2|0.8|9.0|57.7|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/dt05_simu_beamformit_2mics|1640|27120|92.2|6.0|1.9|0.7|8.6|55.5|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/dt05_simu_beamformit_5mics|1640|27120|93.6|4.9|1.5|0.6|7.1|51.6|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/dt05_simu_isolated_1ch_track|1640|27120|89.9|7.6|2.4|1.0|11.1|59.7|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/et05_real_beamformit_2mics|1320|21409|86.7|9.7|3.5|1.3|14.5|64.7|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/et05_real_beamformit_5mics|1320|21409|89.2|7.9|2.9|1.0|11.8|61.2|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/et05_real_isolated_1ch_track|1320|21409|84.6|11.4|4.0|1.5|17.0|69.4|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/et05_simu_beamformit_2mics|1320|21416|86.0|10.5|3.5|1.5|15.5|67.5|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/et05_simu_beamformit_5mics|1320|21416|88.1|8.9|3.1|1.2|13.1|64.8|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/et05_simu_isolated_1ch_track|1320|21416|82.8|13.1|4.1|1.9|19.1|69.4|
+
+### CER
+
+|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
+|---|---|---|---|---|---|---|---|---|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/dt05_real_beamformit_2mics|1640|160390|96.6|1.4|2.0|0.6|4.0|53.3|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/dt05_real_beamformit_5mics|1640|160390|97.1|1.1|1.8|0.5|3.4|49.9|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/dt05_real_isolated_1ch_track|1640|160390|95.9|1.7|2.3|0.8|4.8|57.7|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/dt05_simu_beamformit_2mics|1640|160400|95.9|1.7|2.3|0.7|4.8|55.5|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/dt05_simu_beamformit_5mics|1640|160400|96.8|1.4|1.9|0.6|3.8|51.6|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/dt05_simu_isolated_1ch_track|1640|160400|94.7|2.5|2.9|1.0|6.3|59.7|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/et05_real_beamformit_2mics|1320|126796|92.8|3.2|4.0|1.2|8.4|64.7|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/et05_real_beamformit_5mics|1320|126796|94.3|2.4|3.3|1.0|6.6|61.2|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/et05_real_isolated_1ch_track|1320|126796|91.5|3.8|4.6|1.6|10.0|69.4|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/et05_simu_beamformit_2mics|1320|126812|92.2|3.5|4.2|1.7|9.5|67.5|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/et05_simu_beamformit_5mics|1320|126812|93.7|2.7|3.5|1.4|7.7|64.8|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/et05_simu_isolated_1ch_track|1320|126812|90.3|4.8|4.9|2.2|11.9|69.4|
+
+### TER
+
+|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
+|---|---|---|---|---|---|---|---|---|
+
+### Enhancement
+
+|dataset|STOI|SAR|SDR|SI_SNR|
+|---|---|---|---|---|
+|dt05_simu_isolated_1ch_track|0.87|7.14|7.14|4.51|
+|et05_simu_isolated_1ch_track|0.85|7.47|7.47|3.02|

egs2/chime4/enh_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

egs2/chime4/enh_asr1/conf/chime4.cfg

@@ -0,0 +1 @@
+../../asr1/conf/chime4.cfg

egs2/chime4/enh_asr1/conf/decode_asr_transformer.yaml

@@ -0,0 +1,7 @@
+batch_size: 0
+beam-size: 10
+penalty: 0.0
+maxlenratio: 0.0
+minlenratio: 0.0
+ctc-weight: 0.3
+lm-weight: 1.0

egs2/chime4/enh_asr1/conf/fbank.conf

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

egs2/chime4/enh_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

egs2/chime4/enh_asr1/conf/pitch.conf

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

egs2/chime4/enh_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

egs2/chime4/enh_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.

egs2/chime4/enh_asr1/conf/train_enh_asr_convtasnet_fbank_transformer.yaml

@@ -0,0 +1 @@
+tuning/train_enh_asr_convtasnet_si_snr_fbank_transformer_lr2e-3_accum2_warmup20k_specaug.yaml

egs2/chime4/enh_asr1/conf/train_lm_transformer.yaml

@@ -0,0 +1,48 @@
+# network architecture
+# encoder related
+encoder: transformer
+encoder_conf:
+    input_layer: conv2d
+    num_blocks: 12
+    linear_units: 2048
+    dropout_rate: 0.1
+    output_size: 256
+    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
+
+# hybrid CTC/attention
+model_conf:
+    ctc_weight: 0.3
+    lsm_weight: 0.1
+    length_normalized_loss: false
+
+# optimization related
+optim: adam
+accum_grad: 2
+grad_clip: 5
+patience: 10
+max_epoch: 100
+optim_conf:
+    lr: 0.005
+scheduler: warmuplr
+scheduler_conf:
+    warmup_steps: 20000
+
+# minibatch related
+batch_type: folded
+batch_size: 32
+
+# criterion
+best_model_criterion:
+-  - valid
+   - acc
+   - max
+keep_nbest_models: 10