test_integration_espnet2.sh

#!/usr/bin/env bash

set -euo pipefail

source tools/activate_python.sh
PYTHONPATH="${PYTHONPATH:-}:$(pwd)/tools/s3prl"
export PYTHONPATH
python="coverage run --append"
cwd=$(pwd)

gen_dummy_coverage(){
    # To avoid a problem when parallel running for `coverage run`.
    # Please put this command after cd ./egs2/foo/bar
    touch empty.py; ${python} empty.py
}

#### Make sure chainer-independent ####
python3 -m pip uninstall -y chainer

# [ESPnet2] test asr recipe
cd ./egs2/mini_an4/asr1
gen_dummy_coverage
echo "==== [ESPnet2] ASR ==="
./run.sh --stage 1 --stop-stage 1
feats_types="raw fbank_pitch"
token_types="bpe char"
for t in ${feats_types}; do
    ./run.sh --stage 2 --stop-stage 4 --feats-type "${t}" --python "${python}"
done
cp -r dump/raw data/
./run.sh --stage 2 --stop-stage 4 --feats-type "raw_copy" \
    --train_set raw/train_nodev --valid_set raw/train_dev --test_sets raw/test --python "${python}"
for t in ${token_types}; do
    ./run.sh --stage 5 --stop-stage 5 --token-type "${t}" --python "${python}"
done
use_lm=true
for t in ${feats_types}; do
    for t2 in ${token_types}; do
        echo "==== feats_type=${t}, token_types=${t2} ==="
        ./run.sh --use_lm ${use_lm} --ngpu 0 --stage 6 --stop-stage 13 --skip-upload false --feats-type "${t}" --token-type "${t2}" --python "${python}" --asr-args "--num_workers 0"
    done
    use_lm=false
    echo "==== feats_type=raw_copy, token_types=bpe ==="
    cp -r dump/raw data/
    ./run.sh --use_lm ${use_lm} --ngpu 0 --stage 4 --stop-stage 13 --skip-upload false --feats-type "raw_copy" --token-type "${t2}" \
        --train_set raw/train_nodev --valid_set raw/train_dev --test_sets raw/test --python "${python}" --asr-args "--num_workers 0"
done
echo "==== feats_type=raw, token_types=bpe, model_conf.extract_feats_in_collect_stats=False, normalize=utt_mvn ==="
./run.sh --ngpu 0 --stage 10 --stop-stage 13 --skip-upload false --feats-type "raw" --token-type "bpe" \
    --feats_normalize "utterance_mvn" --python "${python}" \
    --asr-args "--model_conf extract_feats_in_collect_stats=false --num_workers 0"

echo "==== feats_type=raw, token_types=bpe, model_conf.extract_feats_in_collect_stats=False, normalize=utt_mvn, with data augmentation ==="
./run.sh --ngpu 0 --stage 10 --stop-stage 13 --skip-upload false --feats-type "raw" --token-type "bpe" \
    --asr_config "conf/train_asr_rnn_data_aug_debug.yaml" \
    --feats_normalize "utterance_mvn" --python "${python}" \
    --asr-args "--model_conf extract_feats_in_collect_stats=false --num_workers 0"

echo "==== use_streaming, feats_type=raw, token_types=bpe, model_conf.extract_feats_in_collect_stats=False, normalize=utt_mvn ==="
./run.sh --use_streaming true --ngpu 0 --stage 10 --stop-stage 13 --skip-upload false --feats-type "raw" --token-type "bpe" \
    --feats_normalize "utterance_mvn"  --python "${python}" \
    --asr_config "" --asr-tag "train_raw_bpe_streaming" \
    --asr-args "--model_conf extract_feats_in_collect_stats=false --encoder=contextual_block_transformer
                --encoder_conf='{'block_size': 40, 'hop_size': 16, 'look_ahead': 16, 'output_size': 2, 'attention_heads': 2, 'linear_units': 2, 'num_blocks': 1}'
                --decoder=transformer --decoder_conf='{'attention_heads': 2, 'linear_units': 2, 'num_blocks': 1}'
                --max_epoch 1 --num_iters_per_epoch 1 --batch_size 2 --batch_type folded --num_workers 0"

echo "==== Transducer, feats_type=raw, token_types=bpe ==="
./run.sh --asr-tag "espnet_model_transducer" --ngpu 0 --stage 10 --stop-stage 13 --skip-upload false \
    --feats-type "raw" --token-type "bpe" --python "${python}" \
    --asr-args "--decoder transducer --decoder_conf hidden_size=2 --model_conf ctc_weight=0.0 --joint_net_conf joint_space_size=2 --num_workers 0 \
    --best_model_criterion '(valid, loss, min)'" --inference_asr_model "valid.loss.best.pth"

if [ "$(python3 -c "import torch; print(torch.cuda.is_available())")" == "True" ]; then
    echo "==== Multi-Blank Transducer, feats_type=raw, token_types=bpe ==="
    ./run.sh --asr-tag "espnet_model_multi_blank_transducer" --ngpu 1 --stage 10 --stop-stage 13 --skip-upload false \
        --feats-type "raw" --token-type "bpe" --python "${python}" \
        --asr-tag "train_multi_black_transducer" \
        --asr_args "--decoder transducer --decoder_conf hidden_size=2 --model_conf ctc_weight=0.0 --joint_net_conf joint_space_size=2 \
                    --best_model_criterion '(valid, loss, min)' --model_conf transducer_multi_blank_durations=[2] \
                    --max_epoch 1 --num_iters_per_epoch 1 --batch_size 2 --batch_type folded --num_workers 0" \
        --inference_asr_model "valid.loss.best.pth" --inference_config "conf/decode_multi_blank_transducer_debug.yaml"
fi

if python3 -c "import k2" &> /dev/null; then
    echo "==== use_k2, num_paths > nll_batch_size, feats_type=raw, token_types=bpe, model_conf.extract_feats_in_collect_stats=False, normalize=utt_mvn ==="
    ./run.sh --num_paths 4 --nll_batch_size 2 --use_k2 true --ngpu 0 --stage 12 --stop-stage 13 --skip-upload false --feats-type "raw" --token-type "bpe" \
        --feats_normalize "utterance_mvn" --python "${python}" --asr-args "--model_conf extract_feats_in_collect_stats=false --num_workers 0"

    echo "==== use_k2, num_paths == nll_batch_size, feats_type=raw, token_types=bpe, model_conf.extract_feats_in_collect_stats=False, normalize=utt_mvn ==="
    ./run.sh --num_paths 2 --nll_batch_size 2 --use_k2 true --ngpu 0 --stage 12 --stop-stage 13 --skip-upload false --feats-type "raw" --token-type "bpe" \
       --feats_normalize "utterance_mvn" --python "${python}" --asr-args "--model_conf extract_feats_in_collect_stats=false --num_workers 0"
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_config "" --asr_task "asr_transducer" --ngpu 0 --stage 10 --stop-stage 13 --skip-upload false --feats-type "raw" --token-type ${t} \
            --feats_normalize "utterance_mvn" --python "${python}" --inference_asr_model "valid.loss.best.pth" \
            --asr-tag "${asr_tag}_conformer" \
            --asr-args "--model_conf extract_feats_in_collect_stats=false \
                        --encoder_conf body_conf='[{'block_type': 'conformer', 'hidden_size': 2, 'linear_size': 4, 'heads': 2, 'conv_mod_kernel_size': 3}]' \
                        --decoder_conf='{'embed_size': 4, 'hidden_size': 4}' --joint_network_conf joint_space_size=4 \
                        --max_epoch 1 --num_iters_per_epoch 1 --batch_size 2 --batch_type folded --num_workers 0"

        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_config "" --asr_task "asr_transducer" --ngpu 0 --stage 10 --stop-stage 13 --skip-upload false --feats-type "raw" --token-type ${t} \
            --feats_normalize "utterance_mvn" --python "${python}" --inference_asr_model "valid.loss.best.pth" \
            --asr-tag "${asr_tag}_conformer_streaming" \
            --asr-args "--model_conf extract_feats_in_collect_stats=false \
                        --encoder_conf main_conf='{'dynamic_chunk_training': True}' \
                        --encoder_conf body_conf='[{'block_type': 'conformer', 'hidden_size': 2, 'linear_size': 4, 'heads': 2, 'conv_mod_kernel_size': 3}]' \
                        --decoder_conf='{'embed_size': 4, 'hidden_size': 4}' --joint_network_conf joint_space_size=4 \
                        --max_epoch 1 --num_iters_per_epoch 1 --batch_size 2 --batch_type folded --num_workers 0" \
            --inference-args "--streaming true --decoding_window 160 --left_context 2"
    done
fi

echo "==== [PIT_ASR] feats_type=raw, token_types=bpe, model_conf.extract_feats_in_collect_stats=False, normalize=utt_mvn ==="
for i in $(seq 2); do
    for rr in raw raw/org; do
        cp dump/${rr}/train_nodev/text dump/${rr}/train_nodev/text_spk${i}
        cp dump/${rr}/train_dev/text dump/${rr}/train_dev/text_spk${i}
    done
    cp dump/raw/test/text dump/raw/test/text_spk${i}
    cp dump/raw/test_seg/text dump/raw/test_seg/text_spk${i}
done
./run_multispkr.sh --ngpu 0 --stage 10 --stop-stage 13 --skip-upload false --feats-type "raw" --token-type "bpe" \
    --feats_normalize "utterance_mvn" --python "${python}" \
    --asr_config "" \
    --asr_tag "train_multispkr_raw_en_bpe30" \
    --asr-args "--model_conf extract_feats_in_collect_stats=false \
                --ctc_conf reduce=False --encoder transformer_multispkr \
                --encoder_conf '{'num_blocks': 2, 'num_blocks_sd': 2, 'num_inf': 2, 'output_size': 2, 'attention_heads': 2, 'linear_units': 2}' \
                --decoder rnn \
                --model pit_espnet --model_conf '{'num_inf': 2, 'num_ref': 2}' \
                --preprocessor multi --preprocessor_conf text_name='['text', 'text_spk2']' \
                --max_epoch 1 --num_iters_per_epoch 1 --batch_size 2 --batch_type folded --num_workers 0" \
    --inference-args "--multi_asr true"

# Remove generated files in order to reduce the disk usage
rm -rf exp dump data
cd "${cwd}"

# [ESPnet2] test tts recipe
cd ./egs2/mini_an4/tts1
gen_dummy_coverage
echo "==== [ESPnet2] TTS ==="
./run.sh --ngpu 0 --stage 1 --stop-stage 7 --skip-upload false --python "${python}" --train-args "--num_workers 0"
# Remove generated files in order to reduce the disk usage
rm -rf exp dump data

# [ESPnet2] test gan-tts recipe
# NOTE(kan-bayashi): pytorch 1.4 - 1.6 works but 1.6 has a problem with CPU,
#   so we test this recipe using only pytorch > 1.6 here.
#   See also: https://github.com/pytorch/pytorch/issues/42446
if python3 -c 'import torch as t; from packaging.version import parse as L; assert L(t.__version__) > L("1.6")' &> /dev/null; then
    ./run.sh --fs 22050 --tts_task gan_tts --feats_extract linear_spectrogram --feats_normalize none --inference_model latest.pth \
        --ngpu 0 --stop-stage 7 --skip-upload false --python "${python}" \
        --train-config "" --train-args "--max_epoch 1 --num_iters_per_epoch 1 --batch_size 1 --batch_type folded --num_workers 0"
    rm -rf exp dump data
fi
cd "${cwd}"

# [ESPnet2] test enh recipe
if python -c 'import torch as t; from packaging.version import parse as L; assert L(t.__version__) >= L("1.2.0")' &> /dev/null; then
    cd ./egs2/mini_an4/enh1
    gen_dummy_coverage
    echo "==== [ESPnet2] ENH ==="
    ./run.sh --stage 1 --stop-stage 1 --python "${python}"
    feats_types="raw"
    for t in ${feats_types}; do
        echo "==== feats_type=${t} with preprocessor ==="
        ./run.sh --ngpu 0 --stage 2 --stop-stage 10 --skip-upload false --feats-type "${t}" --ref-num 1 --python "${python}" \
            --extra_wav_list "rirs.scp noises.scp" --enh_config ./conf/train_with_preprocessor_debug.yaml --enh-args "--num_workers 0"
        ./run.sh --ngpu 0 --stage 5 --stop-stage 10 --skip-upload false --feats-type "${t}" --ref-num 1 --python "${python}" \
            --enh_config conf/train_with_data_aug_debug.yaml --enh-args "--num_workers 0"
        ./run.sh --ngpu 0 --stage 2 --stop-stage 10 --skip-upload false --feats-type "${t}" --ref-num 2 --python "${python}" \
            --enh_config conf/train_with_dynamic_mixing_debug.yaml --enh-args "--num_workers 0"
    done
    rm data/**/utt2category 2>/dev/null || true
    rm -r dump
    for t in ${feats_types}; do
        echo "==== feats_type=${t} without preprocessor ==="
        ./run.sh --ngpu 0 --stage 2 --stop-stage 10 --skip-upload false --feats-type "${t}" --ref-num 1 --python "${python}" --enh-args "--num_workers 0"
    done
    # Remove generated files in order to reduce the disk usage
    rm -rf exp dump data
    cd "${cwd}"
fi

# [ESPnet2] test enh_tse recipe
if python -c 'import torch as t; from packaging.version import parse as L; assert L(t.__version__) >= L("1.2.0")' &> /dev/null; then
    cd ./egs2/mini_an4/tse1
    gen_dummy_coverage
    echo "==== [ESPnet2] ENH_TSE ==="
    feats_types="raw"
    for t in ${feats_types}; do
        echo "==== feats_type=${t} ==="
        ./run.sh --ngpu 0 --stage 1 --stop-stage 10 --skip-upload false --feats-type "${t}" --ref-num 1 --python "${python}" --enh-args "--num_workers 0"
        ./run.sh --ngpu 0 --stage 1 --stop-stage 10 --skip-upload false --feats-type "${t}" --ref-num 1 --python "${python}" \
            --local_data_opts "--random-enrollment true" --enh_config ./conf/train_random_enrollment_debug.yaml --enh-args "--num_workers 0"
    done
    # Remove generated files in order to reduce the disk usage
    rm -rf exp dump data
    cd "${cwd}"
fi

# [ESPnet2] test ssl1 recipe
if python3 -c 'import torch as t; from packaging.version import parse as L; assert L(t.__version__) >= L("1.12.0")' &> /dev/null; then
    cd ./egs2/mini_an4/ssl1
    gen_dummy_coverage
    echo "==== [ESPnet2] SSL1/HUBERT ==="
    ./run.sh --ngpu 0 --stage 1 --stop-stage 7 --feats-type "raw" --token_type "word" --skip_upload_hf false --python "${python}" --hubert-args "--num_workers 0"
    # Remove generated files in order to reduce the disk usage
    rm -rf exp dump data
    cd "${cwd}"
fi

# [ESPnet2] test enh_asr1 recipe
if python -c 'import torch as t; from packaging.version import parse as L; assert L(t.__version__) >= L("1.2.0")' &> /dev/null; then
    cd ./egs2/mini_an4/enh_asr1
    gen_dummy_coverage
    echo "==== [ESPnet2] ENH_ASR ==="
    ./run.sh --ngpu 0 --stage 0 --stop-stage 15 --skip-upload_hf false --feats-type "raw" --spk-num 1 --enh_asr_args "--enh_separator_conf num_spk=1 --num_workers 0" --python "${python}"
    # Remove generated files in order to reduce the disk usage
    rm -rf exp dump data
    cd "${cwd}"
fi

# [ESPnet2] test st recipe
cd ./egs2/mini_an4/st1
echo "==== [ESPnet2] ST ==="
./run.sh --stage 1 --stop-stage 1
feats_types="raw fbank_pitch"
token_types="bpe char"
for t in ${feats_types}; do
    ./run.sh --stage 2 --stop-stage 4 --feats-type "${t}" --python "${python}"
done
for t in ${token_types}; do
    ./run.sh --stage 5 --stop-stage 5 --tgt_token_type "${t}" --src_token_type "${t}" --python "${python}"
done
use_lm=true
for t in ${feats_types}; do
    for t2 in ${token_types}; do
        echo "==== feats_type=${t}, token_types=${t2} ==="
        ./run.sh --use_lm ${use_lm} --ngpu 0 --stage 6 --stop-stage 13 --skip-upload false --feats-type "${t}" --tgt_token_type "${t2}" --src_token_type "${t2}" --python "${python}" --st-args "--num_workers 0"
    done
    use_lm=false
done
echo "==== feats_type=raw, token_types=bpe, model_conf.extract_feats_in_collect_stats=False, normalize=utt_mvn ==="
./run.sh --ngpu 0 --stage 10 --stop-stage 13 --skip-upload false --feats-type "raw" --tgt_token_type "bpe" --src_token_type "bpe" \
    --feats_normalize "utterance_mvn" --python "${python}" --st-args "--model_conf extract_feats_in_collect_stats=false --num_workers 0"

echo "==== use_streaming, feats_type=raw, token_types=bpe, model_conf.extract_feats_in_collect_stats=False, normalize=utt_mvn ==="
./run.sh --use_streaming true --ngpu 0 --stage 10 --stop-stage 13 --skip-upload false --feats-type "raw" --tgt_token_type "bpe" --src_token_type "bpe" \
    --feats_normalize "utterance_mvn" --python "${python}" \
    --st-config conf/train_st_streaming_debug.yaml --st-args "--model_conf extract_feats_in_collect_stats=false --num_workers 0"

# Remove generated files in order to reduce the disk usage
rm -rf exp dump data
cd "${cwd}"

# [ESPnet2] test asr2 recipe
cd ./egs2/mini_an4/asr2
gen_dummy_coverage
echo "==== [ESPnet2] ASR2 ==="
./run.sh --ngpu 0 --stage 1 --stop-stage 15 --skip-upload false --use-lm false --python "${python}" --asr-args "--num_workers 0"
# Remove generated files in order to reduce the disk usage
rm -rf exp dump data
cd "${cwd}"

# [ESPnet2] test spk1 recipe
cd ./egs2/mini_an4/spk1
gen_dummy_coverage
echo "==== [ESPnet2] SPK ==="
./run.sh --ngpu 0 --stage 0 --stop-stage 4 --feats-type "raw" --python "${python}" --spk-args "--num_workers 0"
./run.sh --ngpu 0 --stage 4 --stop-stage 4 --feats-type "raw" --python "${python}" --spk_config conf/train_rawnet3_dataaug_debug.yaml --spk-args "--num_workers 0"
./run.sh --ngpu 0 --stage 4 --stop-stage 4 --feats-type "raw" --python "${python}" --spk_config conf/train_rawnet3_sampler.yaml --spk-args "--num_workers 0"
# Remove generated files in order to reduce the disk usage
rm -rf exp dump data
cd "${cwd}"

echo "=== report ==="
coverage combine egs2/*/*/.coverage
coverage report
coverage xml