Merge remote-tracking branch 'upstream/master' into fix_init

.github/workflows/ci.yaml

@@ -36,7 +36,7 @@ jobs:
       - uses: actions/cache@v1
         with:
           path: ~/.cache/pip
-          key: ${{ runner.os }}-pip-${{ matrix.python-version }}-${{ matrix.pytorch-version }}-${{ hashFiles('**/setup.py') }}
+          key: ${{ runner.os }}-pip-${{ matrix.python-version }}-${{ matrix.pytorch-version }}-${{ hashFiles('**/setup.py') }}-${{ hashFiles('**/Makefile') }}
       - uses: actions/setup-python@v1
         with:
           python-version: ${{ matrix.python-version }}

.github/workflows/test_import.yaml

@@ -19,10 +19,6 @@ jobs:
         pytorch-version: [1.11.0]
     steps:
     - uses: actions/checkout@v2
-    - uses: actions/cache@v1
-      with:
-        path: ~/.cache/pip
-        key: ${{ runner.os }}-pip-${{ matrix.python-version }}-${{ matrix.pytorch-version }}-${{ hashFiles('**/setup.py') }}
     - name: Set up Python
       uses: actions/setup-python@v1
       with:
@@ -35,10 +31,14 @@ jobs:
       env:
         TH_VERSION: ${{ matrix.pytorch-version }}
       run: |
+        python3 -m pip install "numpy<=1.21.4"
+        python3 -m pip freeze
         ./tools/installers/install_torch.sh false ${TH_VERSION} CPU
         ./tools/installers/install_chainer.sh CPU
         python3 setup.py bdist_wheel
         python3 -m pip install dist/espnet-*.whl
+        # log
+        python3 -m pip freeze
     - name: Import all modules (Try1)
       run: |
         python3 ./ci/test_import_all.py
@@ -47,6 +47,8 @@ jobs:
         TH_VERSION: ${{ matrix.pytorch-version }}
       run: |
         python3 -m pip install "$(ls dist/espnet-*.whl)[all]"
+        # log
+        python3 -m pip freeze
     - name: Import all modules (Try2)
       run: |
         python3 ./ci/test_import_all.py

ci/test_integration_espnet2.sh

@@ -53,6 +53,30 @@ if python3 -c "import k2" &> /dev/null; then
        --asr-args "--model_conf extract_feats_in_collect_stats=false --max_epoch=1"
 fi
 
+if python3 -c "from warprnnt_pytorch import RNNTLoss" &> /dev/null; then
+    echo "==== [ESPnet2] ASR Transducer (standalone) ==="
+
+    for t in ${token_types}; do
+	asr_tag="transducer_${t}"
+
+	echo "==== [Conformer-RNN-T] feats_type=raw, token_types=${t}, model_conf.extract_feats_in_collect_stats=False, normalize=utt_mvn ==="
+	./run.sh --asr_task "asr_transducer" --ngpu 0 --stage 10 --stop-stage 13 --skip-upload false --feats-type "raw" --token-type ${t} \
+		 --feats_normalize "utterance_mvn" --lm-args "--max_epoch=1" --python "${python}" --inference_asr_model "valid.loss.best.pth" \
+		 --asr-tag "${asr_tag}_conformer" --asr-args "--model_conf extract_feats_in_collect_stats=false --max_epoch=1 \
+		 --encoder_conf body_conf='[{'block_type': 'conformer', 'hidden_size': 30, 'linear_size': 30, 'heads': 2, 'conv_mod_kernel_size': 3}]' \
+		 --decoder_conf='{'embed_size': 30, 'hidden_size': 30}' --joint_network_conf joint_space_size=30"
+
+	echo "==== [Streaming Conformer-RNN-T] feats_type=raw, token_types=${t}, model_conf.extract_feats_in_collect_stats=False, normalize=utt_mvn ==="
+	./run.sh --asr_task "asr_transducer" --ngpu 0 --stage 10 --stop-stage 13 --skip-upload false --feats-type "raw" --token-type ${t} \
+		 --feats_normalize "utterance_mvn" --lm-args "--max_epoch=1" --python "${python}" --inference_asr_model "valid.loss.best.pth" \
+		 --asr-tag "${asr_tag}_conformer_streaming" --asr-args "--model_conf extract_feats_in_collect_stats=false --max_epoch=1 \
+		 --encoder_conf main_conf='{'dynamic_chunk_training': True}' \
+		 --encoder_conf body_conf='[{'block_type': 'conformer', 'hidden_size': 30, 'linear_size': 30, 'heads': 2, 'conv_mod_kernel_size': 3}]' \
+		 --decoder_conf='{'embed_size': 30, 'hidden_size': 30}' --joint_network_conf joint_space_size=30 " \
+		 --inference-args "--streaming true --chunk_size 2 --left_context 2 --right_context 0"
+    done
+fi
+
 # Remove generated files in order to reduce the disk usage
 rm -rf exp dump data
 cd "${cwd}"

egs2/README.md

@@ -8,6 +8,7 @@ See: https://espnet.github.io/espnet/espnet2_tutorial.html#recipes-using-espnet2
 
 | Directory name          | Corpus name                                                                                                                      | Task                    | Language              | URL                                                                                                          | Note         |
 |-------------------------|----------------------------------------------------------------------------------------------------------------------------------|-------------------------| --------------------- | ------------------------------------------------------------------------------------------------------------ | ------------ |
+| aesrc2020               | Accented English Speech Recognition Challenge 2020                                                                               | ASR                     | ENG                   | https://arxiv.org/abs/2102.10233                                                                             |              |
 | aidatatang_200zh        | Aidatatang_200zh A free Chinese Mandarin speech corpus                                                                           | ASR                     | CMN                   | http://www.openslr.org/resources/62                                                                          |              |
 | aishell                 | AISHELL-ASR0009-OS1 Open Source Mandarin Speech Corpus                                                                           | ASR                     | CMN                   | http://www.aishelltech.com/kysjcp                                                                            |              |
 | aishell2                | AISHELL-2 Open Source Mandarin Speech Corpus                                                                                     | ASR                     | CMN                   | https://www.aishelltech.com/aishell_2                                                                        |              |

egs2/TEMPLATE/asr1/asr.sh

@@ -83,6 +83,7 @@ num_splits_lm=1   # Number of splitting for lm corpus.
 word_vocab_size=10000 # Size of word vocabulary.
 
 # ASR model related
+asr_task=asr   # ASR task mode. Either 'asr' or 'asr_transducer'.
 asr_tag=       # Suffix to the result dir for asr model training.
 asr_exp=       # Specify the directory path for ASR experiment.
                # If this option is specified, asr_tag is ignored.
@@ -203,6 +204,7 @@ Options:
     --num_splits_lm   # Number of splitting for lm corpus (default="${num_splits_lm}").
 
     # ASR model related
+    --asr_task         # ASR task mode. Either 'asr' or 'asr_transducer'. (default="${asr_task}").
     --asr_tag          # Suffix to the result dir for asr model training (default="${asr_tag}").
     --asr_exp          # Specify the directory path for ASR experiment.
                        # If this option is specified, asr_tag is ignored (default="${asr_exp}").
@@ -974,7 +976,7 @@ if ! "${skip_train}"; then
 
         # shellcheck disable=SC2046,SC2086
         ${train_cmd} JOB=1:"${_nj}" "${_logdir}"/stats.JOB.log \
-            ${python} -m espnet2.bin.asr_train \
+            ${python} -m espnet2.bin.${asr_task}_train \
                 --collect_stats true \
                 --use_preprocessor true \
                 --bpemodel "${bpemodel}" \
@@ -1101,7 +1103,7 @@ if ! "${skip_train}"; then
             --num_nodes "${num_nodes}" \
             --init_file_prefix "${asr_exp}"/.dist_init_ \
             --multiprocessing_distributed true -- \
-            ${python} -m espnet2.bin.asr_train \
+            ${python} -m espnet2.bin.${asr_task}_train \
                 --use_preprocessor true \
                 --bpemodel "${bpemodel}" \
                 --token_type "${token_type}" \
@@ -1191,24 +1193,24 @@ if ! "${skip_eval}"; then
         # 2. Generate run.sh
         log "Generate '${asr_exp}/${inference_tag}/run.sh'. You can resume the process from stage 12 using this script"
         mkdir -p "${asr_exp}/${inference_tag}"; echo "${run_args} --stage 12 \"\$@\"; exit \$?" > "${asr_exp}/${inference_tag}/run.sh"; chmod +x "${asr_exp}/${inference_tag}/run.sh"
-        if "${use_k2}"; then
-          # Now only _nj=1 is verified if using k2
-          asr_inference_tool="espnet2.bin.asr_inference_k2"
-
-          _opts+="--is_ctc_decoding ${k2_ctc_decoding} "
-          _opts+="--use_nbest_rescoring ${use_nbest_rescoring} "
-          _opts+="--num_paths ${num_paths} "
-          _opts+="--nll_batch_size ${nll_batch_size} "
-          _opts+="--k2_config ${k2_config} "
-        else
-          if "${use_streaming}"; then
-              asr_inference_tool="espnet2.bin.asr_inference_streaming"
-          elif "${use_maskctc}"; then
-              asr_inference_tool="espnet2.bin.asr_inference_maskctc"
-          else
-              asr_inference_tool="espnet2.bin.asr_inference"
-          fi
-        fi
+
+	inference_bin_tag=""
+	if [ ${asr_task} == "asr" ]; then
+            if "${use_k2}"; then
+		# Now only _nj=1 is verified if using k2
+		inference_bin_tag="_k2"
+
+		_opts+="--is_ctc_decoding ${k2_ctc_decoding} "
+		_opts+="--use_nbest_rescoring ${use_nbest_rescoring} "
+		_opts+="--num_paths ${num_paths} "
+		_opts+="--nll_batch_size ${nll_batch_size} "
+		_opts+="--k2_config ${k2_config} "
+	    elif "${use_streaming}"; then
+		inference_bin_tag="_streaming"
+	    elif "${use_maskctc}"; then
+		inference_bin_tag="_maskctc"
+            fi
+	fi
 
         for dset in ${test_sets}; do
             _data="${data_feats}/${dset}"
@@ -1250,7 +1252,7 @@ if ! "${skip_eval}"; then
             rm -f "${_logdir}/*.log"
             # shellcheck disable=SC2046,SC2086
             ${_cmd} --gpu "${_ngpu}" JOB=1:"${_nj}" "${_logdir}"/asr_inference.JOB.log \
-                ${python} -m ${asr_inference_tool} \
+                ${python} -m espnet2.bin.${asr_task}_inference${inference_bin_tag} \
                     --batch_size ${batch_size} \
                     --ngpu "${_ngpu}" \
                     --data_path_and_name_and_type "${_data}/${_scp},speech,${_type}" \
@@ -1261,7 +1263,7 @@ if ! "${skip_eval}"; then
                     ${_opts} ${inference_args} || { cat $(grep -l -i error "${_logdir}"/asr_inference.*.log) ; exit 1; }
 
             # 3. Calculate and report RTF based on decoding logs
-            if [ $asr_inference_tool == "espnet2.bin.asr_inference" ]; then
+            if [ ${asr_task} == "asr" ] && [ -z ${inference_bin_tag} ]; then
                 log "Calculating RTF & latency... log: '${_logdir}/calculate_rtf.log'"
                 rm -f "${_logdir}"/calculate_rtf.log
                 _fs=$(python3 -c "import humanfriendly as h;print(h.parse_size('${fs}'))")

egs2/aesrc2020/asr1/README.md

@@ -0,0 +1,34 @@
+<!-- Generated by scripts/utils/show_asr_result.sh -->
+# RESULTS
+## Environments
+- date: `Sat Aug 20 06:55:57 EDT 2022`
+- python version: `3.8.13 (default, Mar 28 2022, 11:38:47)  [GCC 7.5.0]`
+- espnet version: `espnet 202207`
+- pytorch version: `pytorch 1.8.1`
+- Git hash: `c892feb2ba248c85b683bf3cdef6c8f7ce85449a`
+  - Commit date: `Thu Aug 18 11:54:56 2022 -0400`
+
+## asr_train_asr_raw_en_bpe5000_sp
+
+- Model link: [huggingface](https://huggingface.co/espnet/brianyan918_aesrc2020_asr_conformer)
+- ASR config: [./conf/train_asr.yaml](./conf/train_asr.yaml)
+- Inference config: [./conf/decode_asr.yaml](./conf/decode_asr.yaml)
+
+### WER
+
+|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
+|---|---|---|---|---|---|---|---|---|
+|decode_asr_asr_model_valid.acc.ave/test|15299|146813|94.8|4.7|0.6|0.7|5.9|31.9|
+
+### CER
+
+|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
+|---|---|---|---|---|---|---|---|---|
+|decode_asr_asr_model_valid.acc.ave/test|15299|708064|97.9|1.2|0.9|0.7|2.8|31.9|
+
+### TER
+
+|dataset|Snt|Wrd|Corr|Sub|Del|Ins|Err|S.Err|
+|---|---|---|---|---|---|---|---|---|
+|decode_asr_asr_model_valid.acc.ave/test|15299|165136|93.2|4.3|2.6|0.6|7.4|31.9|
+

egs2/aesrc2020/asr1/asr.sh

@@ -0,0 +1 @@
+../../TEMPLATE/asr1/asr.sh

egs2/aesrc2020/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/aesrc2020/asr1/conf/decode_asr.yaml

@@ -0,0 +1,6 @@
+beam_size: 20
+ctc_weight: 0.3
+lm_weight: 0.0
+maxlenratio: 0.0
+minlenratio: 0.0
+penalty: 0.0

egs2/aesrc2020/asr1/conf/fbank.conf

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

egs2/aesrc2020/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/aesrc2020/asr1/conf/pitch.conf

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

egs2/aesrc2020/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/aesrc2020/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/aesrc2020/asr1/conf/train_asr.yaml

@@ -0,0 +1 @@
+tuning/train_asr_conformer_lr2e-3_warmup15k_amp_nondeterministic.yaml