From 2b663318cd1773fb8685b1e03295b6bc6889c283 Mon Sep 17 00:00:00 2001
From: simpleoier <netnetchangxk@gmail.com>
Date: Thu, 28 Apr 2022 00:59:22 -0400
Subject: [PATCH 1/2] fix small bugs and add CHiME4 enh_asr1 recipe & results

---
 ci/test_integration_espnet2.sh                |   3 +
 .../enh_asr1/scripts/utils/show_enh_score.sh  |  85 +++++++++++-
 egs2/chime4/enh_asr1/README.md                |  57 ++++++++
 egs2/chime4/enh_asr1/cmd.sh                   | 110 ++++++++++++++++
 egs2/chime4/enh_asr1/conf/chime4.cfg          |   1 +
 .../enh_asr1/conf/decode_asr_transformer.yaml |   7 +
 egs2/chime4/enh_asr1/conf/fbank.conf          |   2 +
 egs2/chime4/enh_asr1/conf/pbs.conf            |  11 ++
 egs2/chime4/enh_asr1/conf/pitch.conf          |   1 +
 egs2/chime4/enh_asr1/conf/queue.conf          |  12 ++
 egs2/chime4/enh_asr1/conf/slurm.conf          |  14 ++
 ..._enh_asr_convtasnet_fbank_transformer.yaml |   1 +
 .../enh_asr1/conf/train_lm_transformer.yaml   |  48 +++++++
 ...it_lr1e-4_accum1_adam_specaug_bypass0.yaml | 124 ++++++++++++++++++
 ...ormer_lr2e-3_accum2_warmup20k_specaug.yaml | 119 +++++++++++++++++
 egs2/chime4/enh_asr1/db.sh                    |   1 +
 egs2/chime4/enh_asr1/enh_asr.sh               |   1 +
 .../CHiME3_simulate_data_patched_parallel.m   |   1 +
 .../enh_asr1/local/bth_chime4_data_prep.sh    |   1 +
 egs2/chime4/enh_asr1/local/chime4_asr_data.sh |   1 +
 egs2/chime4/enh_asr1/local/chime4_enh_data.sh |   1 +
 .../local/clean_chime4_format_data.sh         |   1 +
 .../enh_asr1/local/clean_wsj0_data_prep.sh    |   1 +
 egs2/chime4/enh_asr1/local/cstr_ndx2flist.pl  |   1 +
 egs2/chime4/enh_asr1/local/data.sh            |  89 +++++++++++++
 .../enh_asr1/local/find_noisy_transcripts.pl  |   1 +
 .../chime4/enh_asr1/local/find_transcripts.pl |   1 +
 egs2/chime4/enh_asr1/local/flist2scp.pl       |   1 +
 egs2/chime4/enh_asr1/local/localize.m         |   1 +
 egs2/chime4/enh_asr1/local/make_stft.sh       |   1 +
 egs2/chime4/enh_asr1/local/ndx2flist.pl       |   1 +
 .../enh_asr1/local/normalize_transcript.pl    |   1 +
 egs2/chime4/enh_asr1/local/path.sh            |   0
 .../local/real_enhan_chime4_data_prep.sh      |   1 +
 .../local/real_ext_chime4_data_prep.sh        |   1 +
 .../local/real_noisy_chime4_data_prep.sh      |   1 +
 .../enh_asr1/local/run_beamform_2ch_track.sh  |   1 +
 .../enh_asr1/local/run_beamform_6ch_track.sh  |   1 +
 .../enh_asr1/local/show_enhance_results.sh    |   1 +
 .../local/simu_enhan_chime4_data_prep.sh      |   1 +
 .../local/simu_ext_chime4_data_prep.sh        |   1 +
 .../local/simu_noisy_chime4_data_prep.sh      |   1 +
 egs2/chime4/enh_asr1/local/sym_channel.py     |   1 +
 egs2/chime4/enh_asr1/local/wsj_data_prep.sh   |   1 +
 egs2/chime4/enh_asr1/local/wsj_format_data.sh |   1 +
 egs2/chime4/enh_asr1/path.sh                  |   1 +
 egs2/chime4/enh_asr1/pyscripts                |   1 +
 egs2/chime4/enh_asr1/run.sh                   |  45 +++++++
 egs2/chime4/enh_asr1/scripts                  |   1 +
 egs2/chime4/enh_asr1/steps                    |   1 +
 egs2/chime4/enh_asr1/utils                    |   1 +
 espnet2/enh/espnet_enh_s2t_model.py           |   2 -
 52 files changed, 761 insertions(+), 3 deletions(-)
 mode change 120000 => 100755 egs2/TEMPLATE/enh_asr1/scripts/utils/show_enh_score.sh
 create mode 100644 egs2/chime4/enh_asr1/README.md
 create mode 100644 egs2/chime4/enh_asr1/cmd.sh
 create mode 120000 egs2/chime4/enh_asr1/conf/chime4.cfg
 create mode 100644 egs2/chime4/enh_asr1/conf/decode_asr_transformer.yaml
 create mode 100644 egs2/chime4/enh_asr1/conf/fbank.conf
 create mode 100644 egs2/chime4/enh_asr1/conf/pbs.conf
 create mode 100644 egs2/chime4/enh_asr1/conf/pitch.conf
 create mode 100644 egs2/chime4/enh_asr1/conf/queue.conf
 create mode 100644 egs2/chime4/enh_asr1/conf/slurm.conf
 create mode 120000 egs2/chime4/enh_asr1/conf/train_enh_asr_convtasnet_fbank_transformer.yaml
 create mode 100644 egs2/chime4/enh_asr1/conf/train_lm_transformer.yaml
 create mode 100644 egs2/chime4/enh_asr1/conf/tuning/train_enh_asr_convtasnet_init_noenhloss_wavlm_transformer_init_lr1e-4_accum1_adam_specaug_bypass0.yaml
 create mode 100644 egs2/chime4/enh_asr1/conf/tuning/train_enh_asr_convtasnet_si_snr_fbank_transformer_lr2e-3_accum2_warmup20k_specaug.yaml
 create mode 120000 egs2/chime4/enh_asr1/db.sh
 create mode 120000 egs2/chime4/enh_asr1/enh_asr.sh
 create mode 120000 egs2/chime4/enh_asr1/local/CHiME3_simulate_data_patched_parallel.m
 create mode 120000 egs2/chime4/enh_asr1/local/bth_chime4_data_prep.sh
 create mode 120000 egs2/chime4/enh_asr1/local/chime4_asr_data.sh
 create mode 120000 egs2/chime4/enh_asr1/local/chime4_enh_data.sh
 create mode 120000 egs2/chime4/enh_asr1/local/clean_chime4_format_data.sh
 create mode 120000 egs2/chime4/enh_asr1/local/clean_wsj0_data_prep.sh
 create mode 120000 egs2/chime4/enh_asr1/local/cstr_ndx2flist.pl
 create mode 100755 egs2/chime4/enh_asr1/local/data.sh
 create mode 120000 egs2/chime4/enh_asr1/local/find_noisy_transcripts.pl
 create mode 120000 egs2/chime4/enh_asr1/local/find_transcripts.pl
 create mode 120000 egs2/chime4/enh_asr1/local/flist2scp.pl
 create mode 120000 egs2/chime4/enh_asr1/local/localize.m
 create mode 120000 egs2/chime4/enh_asr1/local/make_stft.sh
 create mode 120000 egs2/chime4/enh_asr1/local/ndx2flist.pl
 create mode 120000 egs2/chime4/enh_asr1/local/normalize_transcript.pl
 create mode 100644 egs2/chime4/enh_asr1/local/path.sh
 create mode 120000 egs2/chime4/enh_asr1/local/real_enhan_chime4_data_prep.sh
 create mode 120000 egs2/chime4/enh_asr1/local/real_ext_chime4_data_prep.sh
 create mode 120000 egs2/chime4/enh_asr1/local/real_noisy_chime4_data_prep.sh
 create mode 120000 egs2/chime4/enh_asr1/local/run_beamform_2ch_track.sh
 create mode 120000 egs2/chime4/enh_asr1/local/run_beamform_6ch_track.sh
 create mode 120000 egs2/chime4/enh_asr1/local/show_enhance_results.sh
 create mode 120000 egs2/chime4/enh_asr1/local/simu_enhan_chime4_data_prep.sh
 create mode 120000 egs2/chime4/enh_asr1/local/simu_ext_chime4_data_prep.sh
 create mode 120000 egs2/chime4/enh_asr1/local/simu_noisy_chime4_data_prep.sh
 create mode 120000 egs2/chime4/enh_asr1/local/sym_channel.py
 create mode 120000 egs2/chime4/enh_asr1/local/wsj_data_prep.sh
 create mode 120000 egs2/chime4/enh_asr1/local/wsj_format_data.sh
 create mode 120000 egs2/chime4/enh_asr1/path.sh
 create mode 120000 egs2/chime4/enh_asr1/pyscripts
 create mode 100755 egs2/chime4/enh_asr1/run.sh
 create mode 120000 egs2/chime4/enh_asr1/scripts
 create mode 120000 egs2/chime4/enh_asr1/steps
 create mode 120000 egs2/chime4/enh_asr1/utils

diff --git a/ci/test_integration_espnet2.sh b/ci/test_integration_espnet2.sh
index 3cf663466e7..dfa6e61e959 100755
--- a/ci/test_integration_espnet2.sh
+++ b/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.
diff --git a/egs2/TEMPLATE/enh_asr1/scripts/utils/show_enh_score.sh b/egs2/TEMPLATE/enh_asr1/scripts/utils/show_enh_score.sh
deleted file mode 120000
index 6d6490d3760..00000000000
--- a/egs2/TEMPLATE/enh_asr1/scripts/utils/show_enh_score.sh
+++ /dev/null
@@ -1 +0,0 @@
-../../../enh1/scripts/utils/show_enh_score.sh
\ No newline at end of file
diff --git a/egs2/TEMPLATE/enh_asr1/scripts/utils/show_enh_score.sh b/egs2/TEMPLATE/enh_asr1/scripts/utils/show_enh_score.sh
new file mode 100755
index 00000000000..e135d73f91f
--- /dev/null
+++ b/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)
diff --git a/egs2/chime4/enh_asr1/README.md b/egs2/chime4/enh_asr1/README.md
new file mode 100644
index 00000000000..ac5df236797
--- /dev/null
+++ b/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|
diff --git a/egs2/chime4/enh_asr1/cmd.sh b/egs2/chime4/enh_asr1/cmd.sh
new file mode 100644
index 00000000000..2aae6919fef
--- /dev/null
+++ b/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
diff --git a/egs2/chime4/enh_asr1/conf/chime4.cfg b/egs2/chime4/enh_asr1/conf/chime4.cfg
new file mode 120000
index 00000000000..5b3477ab5c6
--- /dev/null
+++ b/egs2/chime4/enh_asr1/conf/chime4.cfg
@@ -0,0 +1 @@
+../../asr1/conf/chime4.cfg
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/conf/decode_asr_transformer.yaml b/egs2/chime4/enh_asr1/conf/decode_asr_transformer.yaml
new file mode 100644
index 00000000000..8e7518150a7
--- /dev/null
+++ b/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
diff --git a/egs2/chime4/enh_asr1/conf/fbank.conf b/egs2/chime4/enh_asr1/conf/fbank.conf
new file mode 100644
index 00000000000..82ac7bd0dbc
--- /dev/null
+++ b/egs2/chime4/enh_asr1/conf/fbank.conf
@@ -0,0 +1,2 @@
+--sample-frequency=16000 
+--num-mel-bins=80
diff --git a/egs2/chime4/enh_asr1/conf/pbs.conf b/egs2/chime4/enh_asr1/conf/pbs.conf
new file mode 100644
index 00000000000..119509938ce
--- /dev/null
+++ b/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
diff --git a/egs2/chime4/enh_asr1/conf/pitch.conf b/egs2/chime4/enh_asr1/conf/pitch.conf
new file mode 100644
index 00000000000..e959a19d5b8
--- /dev/null
+++ b/egs2/chime4/enh_asr1/conf/pitch.conf
@@ -0,0 +1 @@
+--sample-frequency=16000
diff --git a/egs2/chime4/enh_asr1/conf/queue.conf b/egs2/chime4/enh_asr1/conf/queue.conf
new file mode 100644
index 00000000000..500582fab31
--- /dev/null
+++ b/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
diff --git a/egs2/chime4/enh_asr1/conf/slurm.conf b/egs2/chime4/enh_asr1/conf/slurm.conf
new file mode 100644
index 00000000000..3b229673638
--- /dev/null
+++ b/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.
diff --git a/egs2/chime4/enh_asr1/conf/train_enh_asr_convtasnet_fbank_transformer.yaml b/egs2/chime4/enh_asr1/conf/train_enh_asr_convtasnet_fbank_transformer.yaml
new file mode 120000
index 00000000000..920b436ba58
--- /dev/null
+++ b/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
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/conf/train_lm_transformer.yaml b/egs2/chime4/enh_asr1/conf/train_lm_transformer.yaml
new file mode 100644
index 00000000000..a502a55381a
--- /dev/null
+++ b/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
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/conf/tuning/train_enh_asr_convtasnet_init_noenhloss_wavlm_transformer_init_lr1e-4_accum1_adam_specaug_bypass0.yaml b/egs2/chime4/enh_asr1/conf/tuning/train_enh_asr_convtasnet_init_noenhloss_wavlm_transformer_init_lr1e-4_accum1_adam_specaug_bypass0.yaml
new file mode 100644
index 00000000000..1eb24dd8134
--- /dev/null
+++ b/egs2/chime4/enh_asr1/conf/tuning/train_enh_asr_convtasnet_init_noenhloss_wavlm_transformer_init_lr1e-4_accum1_adam_specaug_bypass0.yaml
@@ -0,0 +1,124 @@
+# minibatch related
+batch_type: folded
+batch_size: 16  # A6000 x 1
+accum_grad: 1
+grad_clip: 5
+max_epoch: 12
+patience: 10
+# The initialization method for model parameters
+init: xavier_uniform
+val_scheduler_criterion:
+- valid
+- loss
+best_model_criterion:
+-   - valid
+    - acc
+    - max
+-   - train
+    - loss
+    - min
+keep_nbest_models: 10
+num_att_plot: 3
+unused_parameters: true
+freeze_param: [
+    "s2t_model.frontend.upstream",
+]
+init_param: [
+    "../enh1/exp/enh_train_enh_convtasnet_small_raw/valid.loss.ave_1best.pth:encoder:enh_model.encoder",
+    "../enh1/exp/enh_train_enh_convtasnet_small_raw/valid.loss.ave_1best.pth:separator:enh_model.separator",
+    "../enh1/exp/enh_train_enh_convtasnet_small_raw/valid.loss.ave_1best.pth:decoder:enh_model.decoder",
+    "../asr1/exp/asr_train_asr_transformer_wavlm_lr1e-3_specaug_accum1_preenc128_warmup20k_raw_en_char/valid.acc.ave.pth:frontend:s2t_model.frontend",
+    "../asr1/exp/asr_train_asr_transformer_wavlm_lr1e-3_specaug_accum1_preenc128_warmup20k_raw_en_char/valid.acc.ave.pth:preencoder:s2t_model.preencoder",
+    "../asr1/exp/asr_train_asr_transformer_wavlm_lr1e-3_specaug_accum1_preenc128_warmup20k_raw_en_char/valid.acc.ave.pth:encoder:s2t_model.encoder",
+    "../asr1/exp/asr_train_asr_transformer_wavlm_lr1e-3_specaug_accum1_preenc128_warmup20k_raw_en_char/valid.acc.ave.pth:ctc:s2t_model.ctc",
+    "../asr1/exp/asr_train_asr_transformer_wavlm_lr1e-3_specaug_accum1_preenc128_warmup20k_raw_en_char/valid.acc.ave.pth:decoder:s2t_model.decoder",
+]
+
+# network architecture
+enh_encoder: conv
+enh_encoder_conf:
+    channel: 256
+    kernel_size: 40
+    stride: 20
+enh_decoder: conv
+enh_decoder_conf:
+    channel: 256
+    kernel_size: 40
+    stride: 20
+enh_separator: tcn
+enh_separator_conf:
+    num_spk: 1
+    layer: 4
+    stack: 2
+    bottleneck_dim: 256
+    hidden_dim: 512
+    kernel: 3
+    causal: False
+    norm_type: "gLN"
+    nonlinear: relu
+
+frontend: s3prl
+frontend_conf:
+    frontend_conf:
+        upstream: wavlm_large  # Note: If the upstream is changed, please change the input_size in the preencoder.
+    download_dir: ./hub
+    multilayer_feature: true
+
+asr_preencoder: linear
+asr_preencoder_conf:
+    input_size: 1024  # Note: If the upstream is changed, please change this value accordingly.
+    output_size: 128
+
+# encoder related
+asr_encoder: transformer
+asr_encoder_conf:
+    output_size: 256
+    attention_heads: 4
+    linear_units: 2048
+    num_blocks: 12
+    dropout_rate: 0.1
+    attention_dropout_rate: 0.0
+    input_layer: conv2d2
+    normalize_before: true
+
+# decoder related
+asr_decoder: transformer
+asr_decoder_conf:
+    input_layer: embed
+    attention_heads: 4
+    linear_units: 2048
+    num_blocks: 6
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.0
+    self_attention_dropout_rate: 0.0
+    src_attention_dropout_rate: 0.0
+
+asr_model_conf:
+    ctc_weight: 0.3
+    lsm_weight: 0.1
+    length_normalized_loss: false
+    extract_feats_in_collect_stats: false
+
+model_conf:
+    calc_enh_loss: false
+    bypass_enh_prob: 0.0
+
+optim: adam
+optim_conf:
+    lr: 0.0001
+
+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
+    - 100
+    num_freq_mask: 4
+    apply_time_mask: true
+    time_mask_width_range:
+    - 0
+    - 40
+    num_time_mask: 2
diff --git a/egs2/chime4/enh_asr1/conf/tuning/train_enh_asr_convtasnet_si_snr_fbank_transformer_lr2e-3_accum2_warmup20k_specaug.yaml b/egs2/chime4/enh_asr1/conf/tuning/train_enh_asr_convtasnet_si_snr_fbank_transformer_lr2e-3_accum2_warmup20k_specaug.yaml
new file mode 100644
index 00000000000..8e30e5edecb
--- /dev/null
+++ b/egs2/chime4/enh_asr1/conf/tuning/train_enh_asr_convtasnet_si_snr_fbank_transformer_lr2e-3_accum2_warmup20k_specaug.yaml
@@ -0,0 +1,119 @@
+# minibatch related
+batch_type: folded
+batch_size: 16  # A6000 x 1
+accum_grad: 2
+grad_clip: 5
+max_epoch: 50
+patience: 10
+# The initialization method for model parameters
+init: xavier_uniform
+val_scheduler_criterion:
+- valid
+- loss
+best_model_criterion:
+-   - valid
+    - acc
+    - max
+-   - train
+    - loss
+    - min
+keep_nbest_models: 10
+num_att_plot: 0
+
+# network architecture
+enh_encoder: conv
+enh_encoder_conf:
+    channel: 256
+    kernel_size: 40
+    stride: 20
+enh_decoder: conv
+enh_decoder_conf:
+    channel: 256
+    kernel_size: 40
+    stride: 20
+enh_separator: tcn
+enh_separator_conf:
+    num_spk: 1
+    layer: 4
+    stack: 2
+    bottleneck_dim: 256
+    hidden_dim: 512
+    kernel: 3
+    causal: False
+    norm_type: "gLN"
+    nonlinear: relu
+enh_criterions:
+  # The first criterion
+  - name: si_snr
+    conf:
+        eps: 1e-7
+    # the wrapper for the current criterion
+    # for single-talker case, we simplely use fixed_order wrapper
+    wrapper: fixed_order
+    wrapper_conf:
+        weight: 1.0
+
+frontend: default
+frontend_conf:
+    fs: 16000
+    n_fft: 512
+    win_length: 400
+    hop_length: 160
+    frontend_conf: null
+    apply_stft: True
+
+# encoder related
+asr_encoder: transformer
+asr_encoder_conf:
+    output_size: 256
+    attention_heads: 4
+    linear_units: 2048
+    num_blocks: 12
+    dropout_rate: 0.1
+    attention_dropout_rate: 0.0
+    input_layer: conv2d
+    normalize_before: true
+
+# decoder related
+asr_decoder: transformer
+asr_decoder_conf:
+    input_layer: embed
+    attention_heads: 4
+    linear_units: 2048
+    num_blocks: 6
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.0
+    self_attention_dropout_rate: 0.0
+    src_attention_dropout_rate: 0.0
+
+asr_model_conf:
+    ctc_weight: 0.3
+    lsm_weight: 0.1
+    length_normalized_loss: false
+    extract_feats_in_collect_stats: false
+
+model_conf:
+    bypass_enh_prob: 0.0
+
+optim: adam
+optim_conf:
+    lr: 0.002
+scheduler: warmuplr
+scheduler_conf:
+    warmup_steps: 20000
+
+specaug: specaug
+specaug_conf:
+    apply_time_warp: true
+    time_warp_window: 5
+    time_warp_mode: bicubic
+    apply_freq_mask: true
+    freq_mask_width_range:
+    - 0
+    - 30
+    num_freq_mask: 2
+    apply_time_mask: true
+    time_mask_width_range:
+    - 0
+    - 40
+    num_time_mask: 2
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/db.sh b/egs2/chime4/enh_asr1/db.sh
new file mode 120000
index 00000000000..3090b1bc350
--- /dev/null
+++ b/egs2/chime4/enh_asr1/db.sh
@@ -0,0 +1 @@
+../../TEMPLATE/enh_asr1/db.sh
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/enh_asr.sh b/egs2/chime4/enh_asr1/enh_asr.sh
new file mode 120000
index 00000000000..b00d9b13ef7
--- /dev/null
+++ b/egs2/chime4/enh_asr1/enh_asr.sh
@@ -0,0 +1 @@
+../../TEMPLATE/enh_asr1/enh_asr.sh
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/CHiME3_simulate_data_patched_parallel.m b/egs2/chime4/enh_asr1/local/CHiME3_simulate_data_patched_parallel.m
new file mode 120000
index 00000000000..8f939c2e007
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/CHiME3_simulate_data_patched_parallel.m
@@ -0,0 +1 @@
+../../enh1/local/CHiME3_simulate_data_patched_parallel.m
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/bth_chime4_data_prep.sh b/egs2/chime4/enh_asr1/local/bth_chime4_data_prep.sh
new file mode 120000
index 00000000000..f94db52c974
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/bth_chime4_data_prep.sh
@@ -0,0 +1 @@
+../../asr1/local/bth_chime4_data_prep.sh
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/chime4_asr_data.sh b/egs2/chime4/enh_asr1/local/chime4_asr_data.sh
new file mode 120000
index 00000000000..58fbb0a9212
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/chime4_asr_data.sh
@@ -0,0 +1 @@
+../../asr1/local/data.sh
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/chime4_enh_data.sh b/egs2/chime4/enh_asr1/local/chime4_enh_data.sh
new file mode 120000
index 00000000000..d30a4dc12a7
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/chime4_enh_data.sh
@@ -0,0 +1 @@
+../../enh1/local/data.sh
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/clean_chime4_format_data.sh b/egs2/chime4/enh_asr1/local/clean_chime4_format_data.sh
new file mode 120000
index 00000000000..4826e8e382a
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/clean_chime4_format_data.sh
@@ -0,0 +1 @@
+../../enh1/local/clean_chime4_format_data.sh
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/clean_wsj0_data_prep.sh b/egs2/chime4/enh_asr1/local/clean_wsj0_data_prep.sh
new file mode 120000
index 00000000000..5c61d4de024
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/clean_wsj0_data_prep.sh
@@ -0,0 +1 @@
+../../enh1/local/clean_wsj0_data_prep.sh
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/cstr_ndx2flist.pl b/egs2/chime4/enh_asr1/local/cstr_ndx2flist.pl
new file mode 120000
index 00000000000..50660a2b68e
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/cstr_ndx2flist.pl
@@ -0,0 +1 @@
+../../enh1/local/cstr_ndx2flist.pl
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/data.sh b/egs2/chime4/enh_asr1/local/data.sh
new file mode 100755
index 00000000000..dc36d70eae3
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/data.sh
@@ -0,0 +1,89 @@
+#!/usr/bin/env bash
+
+set -e
+set -u
+set -o pipefail
+
+log() {
+    local fname=${BASH_SOURCE[1]##*/}
+    echo -e "$(date '+%Y-%m-%dT%H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
+}
+SECONDS=0
+
+help_message=$(cat << EOF
+Usage: $0 --extra-annotations <path> [--stage <stage>] [--stop_stage <stop_stage>] [--nj <nj>]
+
+  required argument:
+    --extra-annotations: path to a directory containing extra annotations for CHiME4
+                         This is required for preparing et05_simu_isolated_1ch_track.
+    NOTE:
+        You can download it manually from
+            http://spandh.dcs.shef.ac.uk/chime_challenge/CHiME4/download.html
+        Then unzip the downloaded file to CHiME4_diff;
+        You will then find the extra annotations in CHiME4_diff/CHiME3/data/annotations
+
+  optional argument:
+    [--stage]: 1 (default) or 2
+    [--stop_stage]: 1 or 2 (default)
+    [--nj]: number of parallel pool workers in MATLAB
+EOF
+)
+
+
+stage=0
+stop_stage=100
+extra_annotations=
+local_data_opts=
+train_dev=dt05_multi_isolated_1ch_track
+log "$0 $*"
+. utils/parse_options.sh
+
+
+if [ $# -ne 0 ] || [ -z "${extra_annotations}" ]; then
+    echo "${help_message}"
+    exit 2
+fi
+
+. ./path.sh || exit 1;
+. ./cmd.sh || exit 1;
+. ./db.sh || exit 1;
+
+
+if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
+    log "stage 0: Enh data preparation"
+    local/chime4_enh_data.sh --extra_annotations ${extra_annotations} ${local_data_opts}
+fi
+
+if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
+    log "stage 1: ASR data preparation"
+    local/chime4_asr_data.sh --stage 0 --stop-stage 1 ${local_data_opts}
+fi
+
+if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
+    log "stage 2: Enh_ASR data preparation: combine enh and asr data"
+
+    # dummy spk1.scp
+    for dset in tr05_real_noisy train_si284 dt05_real_isolated_1ch_track et05_real_isolated_1ch_track dt05_real_beamformit_2mics dt05_simu_beamformit_2mics et05_real_beamformit_2mics et05_simu_beamformit_2mics dt05_real_beamformit_5mics dt05_simu_beamformit_5mics et05_real_beamformit_5mics et05_simu_beamformit_5mics; do
+        cp data/${dset}/wav.scp data/${dset}/spk1.scp
+    done
+    cp data/tr05_simu_isolated_1ch_track/spk1.scp data/tr05_simu_noisy
+
+    # utt2category
+    <data/tr05_simu_noisy/wav.scp awk '{print($1, "SIMU")}' > data/tr05_simu_noisy/utt2category
+    <data/tr05_real_noisy/wav.scp awk '{print($1, "REAL")}' > data/tr05_real_noisy/utt2category
+    <data/train_si284/wav.scp awk '{print($1, "CLEAN")}' > data/train_si284/utt2category
+    <data/dt05_simu_isolated_1ch_track/wav.scp awk '{print($1, "SIMU")}' > data/dt05_simu_isolated_1ch_track/utt2category
+    <data/dt05_real_isolated_1ch_track/wav.scp awk '{print($1, "REAL")}' > data/dt05_real_isolated_1ch_track/utt2category
+
+    utils/combine_data.sh --extra_files "utt2category spk1.scp" \
+        data/tr05_multi_noisy data/tr05_simu_noisy data/tr05_real_noisy 
+    utils/combine_data.sh --extra_files "utt2category spk1.scp" \
+        data/tr05_multi_noisy_si284 data/tr05_multi_noisy data/train_si284
+    utils/combine_data.sh --extra_files "utt2category spk1.scp" data/${train_dev} \
+        data/dt05_simu_isolated_1ch_track data/dt05_real_isolated_1ch_track
+fi
+
+if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then
+    log "stage 3: Srctexts preparation"
+    local/chime4_asr_data.sh --stage 2 --stop-stage 2
+fi
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/find_noisy_transcripts.pl b/egs2/chime4/enh_asr1/local/find_noisy_transcripts.pl
new file mode 120000
index 00000000000..ae475b3b32d
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/find_noisy_transcripts.pl
@@ -0,0 +1 @@
+../../enh1/local/find_noisy_transcripts.pl
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/find_transcripts.pl b/egs2/chime4/enh_asr1/local/find_transcripts.pl
new file mode 120000
index 00000000000..5e58a9d0c0e
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/find_transcripts.pl
@@ -0,0 +1 @@
+../../enh1/local/find_transcripts.pl
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/flist2scp.pl b/egs2/chime4/enh_asr1/local/flist2scp.pl
new file mode 120000
index 00000000000..c44f94660eb
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/flist2scp.pl
@@ -0,0 +1 @@
+../../enh1/local/flist2scp.pl
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/localize.m b/egs2/chime4/enh_asr1/local/localize.m
new file mode 120000
index 00000000000..f93a989f0ad
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/localize.m
@@ -0,0 +1 @@
+../../enh1/local/localize.m
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/make_stft.sh b/egs2/chime4/enh_asr1/local/make_stft.sh
new file mode 120000
index 00000000000..cf9038f4ea2
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/make_stft.sh
@@ -0,0 +1 @@
+../../asr1/local/make_stft.sh
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/ndx2flist.pl b/egs2/chime4/enh_asr1/local/ndx2flist.pl
new file mode 120000
index 00000000000..5f79e7991f9
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/ndx2flist.pl
@@ -0,0 +1 @@
+../../asr1/local/ndx2flist.pl
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/normalize_transcript.pl b/egs2/chime4/enh_asr1/local/normalize_transcript.pl
new file mode 120000
index 00000000000..1be067e3703
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/normalize_transcript.pl
@@ -0,0 +1 @@
+../../enh1/local/normalize_transcript.pl
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/path.sh b/egs2/chime4/enh_asr1/local/path.sh
new file mode 100644
index 00000000000..e69de29bb2d
diff --git a/egs2/chime4/enh_asr1/local/real_enhan_chime4_data_prep.sh b/egs2/chime4/enh_asr1/local/real_enhan_chime4_data_prep.sh
new file mode 120000
index 00000000000..13c906eba90
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/real_enhan_chime4_data_prep.sh
@@ -0,0 +1 @@
+../../asr1/local/real_enhan_chime4_data_prep.sh
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/real_ext_chime4_data_prep.sh b/egs2/chime4/enh_asr1/local/real_ext_chime4_data_prep.sh
new file mode 120000
index 00000000000..6620a1d2eb4
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/real_ext_chime4_data_prep.sh
@@ -0,0 +1 @@
+../../enh1/local/real_ext_chime4_data_prep.sh
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/real_noisy_chime4_data_prep.sh b/egs2/chime4/enh_asr1/local/real_noisy_chime4_data_prep.sh
new file mode 120000
index 00000000000..86d5a8cca3b
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/real_noisy_chime4_data_prep.sh
@@ -0,0 +1 @@
+../../enh1/local/real_noisy_chime4_data_prep.sh
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/run_beamform_2ch_track.sh b/egs2/chime4/enh_asr1/local/run_beamform_2ch_track.sh
new file mode 120000
index 00000000000..eb7894626ea
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/run_beamform_2ch_track.sh
@@ -0,0 +1 @@
+../../asr1/local/run_beamform_2ch_track.sh
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/run_beamform_6ch_track.sh b/egs2/chime4/enh_asr1/local/run_beamform_6ch_track.sh
new file mode 120000
index 00000000000..d8609c18f57
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/run_beamform_6ch_track.sh
@@ -0,0 +1 @@
+../../asr1/local/run_beamform_6ch_track.sh
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/show_enhance_results.sh b/egs2/chime4/enh_asr1/local/show_enhance_results.sh
new file mode 120000
index 00000000000..7be0ac655cd
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/show_enhance_results.sh
@@ -0,0 +1 @@
+../../asr1/local/show_enhance_results.sh
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/simu_enhan_chime4_data_prep.sh b/egs2/chime4/enh_asr1/local/simu_enhan_chime4_data_prep.sh
new file mode 120000
index 00000000000..f1227dc8071
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/simu_enhan_chime4_data_prep.sh
@@ -0,0 +1 @@
+../../asr1/local/simu_enhan_chime4_data_prep.sh
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/simu_ext_chime4_data_prep.sh b/egs2/chime4/enh_asr1/local/simu_ext_chime4_data_prep.sh
new file mode 120000
index 00000000000..58b7195ba04
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/simu_ext_chime4_data_prep.sh
@@ -0,0 +1 @@
+../../enh1/local/simu_ext_chime4_data_prep.sh
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/simu_noisy_chime4_data_prep.sh b/egs2/chime4/enh_asr1/local/simu_noisy_chime4_data_prep.sh
new file mode 120000
index 00000000000..da4d7f621c7
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/simu_noisy_chime4_data_prep.sh
@@ -0,0 +1 @@
+../../enh1/local/simu_noisy_chime4_data_prep.sh
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/sym_channel.py b/egs2/chime4/enh_asr1/local/sym_channel.py
new file mode 120000
index 00000000000..9901c190202
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/sym_channel.py
@@ -0,0 +1 @@
+../../asr1/local/sym_channel.py
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/wsj_data_prep.sh b/egs2/chime4/enh_asr1/local/wsj_data_prep.sh
new file mode 120000
index 00000000000..2ba8ba465af
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/wsj_data_prep.sh
@@ -0,0 +1 @@
+../../asr1/local/wsj_data_prep.sh
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/local/wsj_format_data.sh b/egs2/chime4/enh_asr1/local/wsj_format_data.sh
new file mode 120000
index 00000000000..036fb8b8689
--- /dev/null
+++ b/egs2/chime4/enh_asr1/local/wsj_format_data.sh
@@ -0,0 +1 @@
+../../asr1/local/wsj_format_data.sh
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/path.sh b/egs2/chime4/enh_asr1/path.sh
new file mode 120000
index 00000000000..f2720c6899b
--- /dev/null
+++ b/egs2/chime4/enh_asr1/path.sh
@@ -0,0 +1 @@
+../../TEMPLATE/enh_asr1/path.sh
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/pyscripts b/egs2/chime4/enh_asr1/pyscripts
new file mode 120000
index 00000000000..008f9bd4bc5
--- /dev/null
+++ b/egs2/chime4/enh_asr1/pyscripts
@@ -0,0 +1 @@
+../../TEMPLATE/enh_asr1/pyscripts
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/run.sh b/egs2/chime4/enh_asr1/run.sh
new file mode 100755
index 00000000000..c42213e9441
--- /dev/null
+++ b/egs2/chime4/enh_asr1/run.sh
@@ -0,0 +1,45 @@
+#!/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
+
+
+extra_annotations=
+
+train_set=tr05_multi_noisy_si284 # tr05_multi_noisy (original training data) or tr05_multi_noisy_si284 (add si284 data)
+valid_set=dt05_multi_isolated_1ch_track
+test_sets="\
+dt05_real_isolated_1ch_track dt05_simu_isolated_1ch_track et05_real_isolated_1ch_track et05_simu_isolated_1ch_track \
+dt05_real_beamformit_2mics dt05_simu_beamformit_2mics et05_real_beamformit_2mics et05_simu_beamformit_2mics \
+dt05_real_beamformit_5mics dt05_simu_beamformit_5mics et05_real_beamformit_5mics et05_simu_beamformit_5mics \
+"
+
+enh_asr_config=conf/train_enh_asr_convtasnet_fbank_transformer.yaml
+inference_config=conf/decode_asr_transformer.yaml
+lm_config=conf/train_lm_transformer.yaml
+
+
+use_word_lm=false
+word_vocab_size=65000
+
+./enh_asr.sh \
+    --lang en \
+    --spk_num 1 \
+    --ref_channel 3 \
+    --local_data_opts "--extra-annotations ${extra_annotations}" \
+    --nlsyms_txt data/nlsyms.txt \
+    --token_type char \
+    --feats_type raw \
+    --feats_normalize utt_mvn \
+    --enh_asr_config "${enh_asr_config}" \
+    --inference_config "${inference_config}" \
+    --lm_config "${lm_config}" \
+    --use_word_lm ${use_word_lm} \
+    --word_vocab_size ${word_vocab_size} \
+    --train_set "${train_set}" \
+    --valid_set "${valid_set}" \
+    --test_sets "${test_sets}" \
+    --bpe_train_text "data/${train_set}/text" \
+    --lm_train_text "data/${train_set}/text data/local/other_text/text" "$@"
diff --git a/egs2/chime4/enh_asr1/scripts b/egs2/chime4/enh_asr1/scripts
new file mode 120000
index 00000000000..6c0f28ef23c
--- /dev/null
+++ b/egs2/chime4/enh_asr1/scripts
@@ -0,0 +1 @@
+../../TEMPLATE/enh_asr1/scripts
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/steps b/egs2/chime4/enh_asr1/steps
new file mode 120000
index 00000000000..91f2d234e20
--- /dev/null
+++ b/egs2/chime4/enh_asr1/steps
@@ -0,0 +1 @@
+../../../tools/kaldi/egs/wsj/s5/steps
\ No newline at end of file
diff --git a/egs2/chime4/enh_asr1/utils b/egs2/chime4/enh_asr1/utils
new file mode 120000
index 00000000000..f49247da827
--- /dev/null
+++ b/egs2/chime4/enh_asr1/utils
@@ -0,0 +1 @@
+../../../tools/kaldi/egs/wsj/s5/utils
\ No newline at end of file
diff --git a/espnet2/enh/espnet_enh_s2t_model.py b/espnet2/enh/espnet_enh_s2t_model.py
index a966f88a98b..51746f9fbc1 100644
--- a/espnet2/enh/espnet_enh_s2t_model.py
+++ b/espnet2/enh/espnet_enh_s2t_model.py
@@ -32,7 +32,6 @@ def __init__(
         self,
         enh_model: ESPnetEnhancementModel,
         s2t_model: Union[ESPnetASRModel, ESPnetSTModel],
-        permutation_by_enh: bool = True,
         calc_enh_loss: bool = True,
         bypass_enh_prob: float = 0,  # 0 means do not bypass enhancement for all data
     ):
@@ -44,7 +43,6 @@ def __init__(
 
         self.bypass_enh_prob = bypass_enh_prob
 
-        self.permutation_by_enh = permutation_by_enh
         self.calc_enh_loss = calc_enh_loss
         self.extract_feats_in_collect_stats = (
             self.s2t_model.extract_feats_in_collect_stats

From de81cf979fd61ab13e0ab0fe0432fbbaa4776be3 Mon Sep 17 00:00:00 2001
From: simpleoier <netnetchangxk@gmail.com>
Date: Thu, 28 Apr 2022 11:54:10 -0400
Subject: [PATCH 2/2] update egs2/chime4/enh_asr1/README.md and related enh1,
 asr1 configs.

---
 ...-3_specaug_accum1_preenc128_warmup20k.yaml | 90 +++++++++++++++++++
 .../tuning/train_enh_convtasnet_small.yaml    | 64 +++++++++++++
 egs2/chime4/enh_asr1/README.md                | 76 ++++++++++++----
 3 files changed, 212 insertions(+), 18 deletions(-)
 create mode 100644 egs2/chime4/asr1/conf/tuning/train_asr_transformer_wavlm_lr1e-3_specaug_accum1_preenc128_warmup20k.yaml
 create mode 100644 egs2/chime4/enh1/conf/tuning/train_enh_convtasnet_small.yaml

diff --git a/egs2/chime4/asr1/conf/tuning/train_asr_transformer_wavlm_lr1e-3_specaug_accum1_preenc128_warmup20k.yaml b/egs2/chime4/asr1/conf/tuning/train_asr_transformer_wavlm_lr1e-3_specaug_accum1_preenc128_warmup20k.yaml
new file mode 100644
index 00000000000..cee2e0c896d
--- /dev/null
+++ b/egs2/chime4/asr1/conf/tuning/train_asr_transformer_wavlm_lr1e-3_specaug_accum1_preenc128_warmup20k.yaml
@@ -0,0 +1,90 @@
+# minibatch related
+batch_type: folded
+batch_size: 32
+accum_grad: 1
+grad_clip: 5
+max_epoch: 50
+patience: none
+# The initialization method for model parameters
+init: xavier_uniform
+val_scheduler_criterion:
+- valid
+- loss
+best_model_criterion:
+-   - valid
+    - acc
+    - max
+keep_nbest_models: 10
+unused_parameters: true
+# SSL-based frontend is fixed during training for training efficiency,
+# however, the gradients are backprogated through frontend to the enhancement.
+freeze_param: [
+    "frontend.upstream"
+]
+
+# network architecture
+frontend: s3prl
+frontend_conf:
+    frontend_conf:
+        upstream: wavlm_large  # Note: If the upstream is changed, please change the input_size in the preencoder.
+    download_dir: ./hub
+    multilayer_feature: True
+
+preencoder: linear
+preencoder_conf:
+    input_size: 1024  # Note: If the upstream is changed, please change this value accordingly.
+    output_size: 128
+
+# encoder related
+encoder: transformer
+encoder_conf:
+    output_size: 256
+    attention_heads: 4
+    linear_units: 2048
+    num_blocks: 12
+    dropout_rate: 0.1
+    attention_dropout_rate: 0.0
+    input_layer: conv2d2
+    normalize_before: true
+
+# decoder related
+decoder: transformer
+decoder_conf:
+    input_layer: embed
+    attention_heads: 4
+    linear_units: 2048
+    num_blocks: 6
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.0
+    self_attention_dropout_rate: 0.0
+    src_attention_dropout_rate: 0.0
+
+model_conf:
+    ctc_weight: 0.3
+    lsm_weight: 0.1
+    length_normalized_loss: false
+    extract_feats_in_collect_stats: false
+
+optim: adam
+optim_conf:
+    lr: 0.001
+scheduler: warmuplr
+scheduler_conf:
+    warmup_steps: 20000
+
+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
+    - 100
+    num_freq_mask: 4
+    apply_time_mask: true
+    time_mask_width_range:
+    - 0
+    - 40
+    num_time_mask: 2
+
diff --git a/egs2/chime4/enh1/conf/tuning/train_enh_convtasnet_small.yaml b/egs2/chime4/enh1/conf/tuning/train_enh_convtasnet_small.yaml
new file mode 100644
index 00000000000..7c73d4c868f
--- /dev/null
+++ b/egs2/chime4/enh1/conf/tuning/train_enh_convtasnet_small.yaml
@@ -0,0 +1,64 @@
+optim: adam
+init: xavier_uniform
+max_epoch: 100
+batch_type: folded
+batch_size: 32
+iterator_type: chunk
+chunk_length: 32000
+num_workers: 4
+optim_conf:
+    lr: 1.0e-03
+    eps: 1.0e-08
+    weight_decay: 1.0e-05
+patience: 4
+val_scheduler_criterion:
+- valid
+- loss
+best_model_criterion:
+-   - valid
+    - si_snr
+    - max
+-   - valid
+    - loss
+    - min
+keep_nbest_models: 1
+scheduler: reducelronplateau
+scheduler_conf:
+    mode: min
+    factor: 0.5
+    patience: 3
+model_conf:
+    loss_type: si_snr
+encoder: conv
+encoder_conf:
+    channel: 256
+    kernel_size: 40
+    stride: 20
+decoder: conv
+decoder_conf:
+    channel: 256
+    kernel_size: 40
+    stride: 20
+separator: tcn
+separator_conf:
+    num_spk: 1
+    layer: 4
+    stack: 2
+    bottleneck_dim: 256
+    hidden_dim: 512
+    kernel: 3
+    causal: False
+    norm_type: "gLN"
+    nonlinear: relu
+criterions:
+  # The first criterion
+  - name: si_snr
+    conf:
+        eps: 1e-7
+    # the wrapper for the current criterion
+    # for single-talker case, we simplely use fixed_order wrapper
+    wrapper:
+      - type: fixed_order
+        wrapper_conf:
+          weight: 1.0
+
diff --git a/egs2/chime4/enh_asr1/README.md b/egs2/chime4/enh_asr1/README.md
index ac5df236797..f01c087f211 100644
--- a/egs2/chime4/enh_asr1/README.md
+++ b/egs2/chime4/enh_asr1/README.md
@@ -7,51 +7,91 @@
 - 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_init_noenhloss_wavlm_transformer_init_lr1e-4_accum1_adam_specaug_bypass0_raw_en_char
+- Pretrained model: https://huggingface.co/espnet/simpleoier_chime4_enh_asr_convtasnet_init_noenhloss_wavlm_transformer_init_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_isolated_1ch_track|1640|27119|98.3|1.3|0.4|0.2|1.9|21.8|
+|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|98.5|1.2|0.3|0.2|1.7|19.6|
+|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|98.6|1.1|0.3|0.2|1.5|18.7|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/dt05_simu_isolated_1ch_track|1640|27120|97.2|2.1|0.7|0.3|3.1|28.9|
+|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|97.9|1.5|0.5|0.2|2.3|25.2|
+|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|98.4|1.2|0.4|0.1|1.7|19.9|
+|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|96.7|2.6|0.7|0.4|3.7|31.6|
+|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|97.4|2.0|0.6|0.3|2.9|27.3|
+|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|97.8|1.8|0.4|0.2|2.5|24.3|
+|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|94.6|3.7|1.6|0.5|5.9|37.3|
+|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|96.6|2.5|1.0|0.3|3.7|32.5|
+|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|97.5|1.9|0.7|0.3|2.9|28.9|
+
+### 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_isolated_1ch_track|1640|160390|99.4|0.2|0.4|0.2|0.8|21.8|
+|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|99.5|0.2|0.3|0.2|0.7|19.6|
+|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|99.6|0.1|0.3|0.2|0.6|18.7|
+|decode_asr_transformer_normalize_output_wavtrue_lm_lm_train_lm_transformer_en_char_valid.loss.ave_enh_asr_model_valid.acc.ave/dt05_simu_isolated_1ch_track|1640|160400|98.8|0.5|0.7|0.3|1.5|28.9|
+|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|99.2|0.3|0.5|0.2|1.1|25.2|
+|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|99.5|0.2|0.3|0.1|0.7|19.9|
+|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|98.6|0.6|0.8|0.4|1.8|31.7|
+|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|98.9|0.4|0.7|0.3|1.4|27.3|
+|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|99.1|0.4|0.5|0.2|1.1|24.3|
+|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|97.0|1.2|1.9|0.6|3.7|37.3|
+|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|98.2|0.6|1.1|0.4|2.1|32.5|
+|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|98.8|0.4|0.8|0.3|1.5|28.9|
+
+### Enhancement
+
+|dataset|STOI|SDR|SI_SNR|
+|---|---|---|---|
+|dt05_simu_isolated_1ch_track|0.86|4.97|1.77|
+|et05_simu_isolated_1ch_track|0.85|5.45|0.88|
+
 
 ## enh_asr_train_enh_asr_convtasnet_fbank_transformer_raw_en_char
+- Pretrained model: https://huggingface.co/espnet/simpleoier_chime4_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_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_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_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/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_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_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_isolated_1ch_track|1320|21416|82.8|13.1|4.1|1.9|19.1|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_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_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_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/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_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_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_isolated_1ch_track|1320|126812|90.3|4.8|4.9|2.2|11.9|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|
+|dataset|STOI|SDR|SI_SNR|
+|---|---|---|---|
+|dt05_simu_isolated_1ch_track|0.87|7.14|4.51|
+|et05_simu_isolated_1ch_track|0.85|7.47|3.02|