Upload 43 files (#1)

- Upload 43 files (297c44b8ce3b4be589aa860e75f4371ef3a928b6)


Co-authored-by: Axel Chiu <Axelisme@users.noreply.huggingface.co>

ASR-model/TransformerLM_seg_char/env.log

@@ -0,0 +1,195 @@
+SpeechBrain system description
+==============================
+Python version:
+3.8.10 (default, Jun  2 2021, 10:49:15) 
+[GCC 9.4.0]
+==============================
+Installed Python packages:
+appdirs==1.4.4
+argon2-cffi==20.1.0
+async-generator==1.10
+attrs==19.3.0
+Automat==0.8.0
+autopep8==1.5.7
+backcall==0.2.0
+backports.entry-points-selectable==1.1.0
+black==19.10b0
+bleach==3.3.1
+blessings==1.7
+blinker==1.4
+bottle==0.12.19
+certifi==2019.11.28
+cffi==1.14.6
+cfgv==3.3.0
+chardet==3.0.4
+Click==7.0
+cloud-init==21.2
+colorama==0.4.3
+command-not-found==0.3
+configobj==5.0.6
+constantly==15.1.0
+cryptography==2.8
+cupshelpers==1.0
+cycler==0.10.0
+d2l==0.16.6
+datasets==1.11.0
+dbus-python==1.2.16
+debugpy==1.3.0
+decorator==5.0.9
+defer==1.0.6
+defusedxml==0.7.1
+dill==0.3.4
+distlib==0.3.2
+distro==1.4.0
+distro-info===0.23ubuntu1
+entrypoints==0.3
+filelock==3.0.12
+flake8==3.7.9
+fsspec==2021.7.0
+gpustat==0.6.0
+gpuview==0.4.0
+httplib2==0.14.0
+huggingface-hub==0.0.16
+hyperlink==19.0.0
+HyperPyYAML==1.0.0
+identify==2.2.11
+idna==2.8
+importlib-metadata==1.5.0
+incremental==16.10.1
+ipykernel==6.0.2
+ipython==7.25.0
+ipython-genutils==0.2.0
+ipywidgets==7.6.3
+jedi==0.18.0
+Jinja2==2.10.1
+joblib==1.0.1
+jsonpatch==1.22
+jsonpointer==2.0
+jsonschema==3.2.0
+jupyter==1.0.0
+jupyter-client==6.1.12
+jupyter-console==6.4.0
+jupyter-core==4.7.1
+jupyterlab-pygments==0.1.2
+jupyterlab-widgets==1.0.0
+keyring==18.0.1
+kiwisolver==1.3.1
+language-selector==0.1
+launchpadlib==1.10.13
+lazr.restfulclient==0.14.2
+lazr.uri==1.0.3
+macaroonbakery==1.3.1
+MarkupSafe==1.1.0
+matplotlib==3.4.2
+matplotlib-inline==0.1.2
+mccabe==0.6.1
+mistune==0.8.4
+more-itertools==4.2.0
+multiprocess==0.70.12.2
+nbclient==0.5.3
+nbconvert==6.1.0
+nbformat==5.1.3
+nest-asyncio==1.5.1
+netifaces==0.10.4
+nodeenv==1.6.0
+notebook==6.4.0
+numpy==1.21.2
+nvidia-ml-py3==7.352.0
+oauthlib==3.1.0
+packaging==21.0
+pandas==1.3.0
+pandocfilters==1.4.3
+parso==0.8.2
+pathspec==0.9.0
+pexpect==4.6.0
+pickleshare==0.7.5
+Pillow==8.3.1
+platformdirs==2.0.2
+pluggy==0.13.1
+pre-commit==2.15.0
+prometheus-client==0.11.0
+prompt-toolkit==3.0.19
+protobuf==3.6.1
+psutil==5.8.0
+ptyprocess==0.7.0
+py==1.10.0
+pyarrow==5.0.0
+pyasn1==0.4.2
+pyasn1-modules==0.2.1
+pycairo==1.16.2
+pycodestyle==2.5.0
+pycparser==2.20
+pycups==1.9.73
+pyflakes==2.1.1
+Pygments==2.9.0
+PyGObject==3.36.0
+PyHamcrest==1.9.0
+PyJWT==1.7.1
+pymacaroons==0.13.0
+PyMySQL==1.0.2
+PyNaCl==1.3.0
+pyOpenSSL==19.0.0
+pyparsing==2.4.7
+pyRFC3339==1.1
+pyrsistent==0.15.5
+pyserial==3.4
+pytest==5.4.1
+python-apt==2.0.0+ubuntu0.20.4.5
+python-dateutil==2.8.2
+python-debian===0.1.36ubuntu1
+pytube==10.9.3
+pytz==2019.3
+PyYAML==5.3.1
+pyzmq==22.1.0
+qtconsole==5.1.1
+QtPy==1.9.0
+regex==2021.7.6
+requests==2.22.0
+requests-unixsocket==0.2.0
+ruamel.yaml==0.17.10
+ruamel.yaml.clib==0.2.6
+scipy==1.7.1
+screen-resolution-extra==0.0.0
+SecretStorage==2.3.1
+Send2Trash==1.7.1
+sentencepiece==0.1.96
+service-identity==18.1.0
+simplejson==3.16.0
+six==1.14.0
+sos==4.1
+-e git+https://github.com/speechbrain/speechbrain.git@1d194bfc51ae20b9e38596d220cdf0f4977e69de#egg=speechbrain
+ssh-import-id==5.10
+supervisor==4.1.0
+systemd-python==234
+terminado==0.10.1
+testpath==0.5.0
+toml==0.10.2
+torch==1.8.1
+torchaudio==0.8.1
+torchvision==0.10.0
+tornado==6.1
+tqdm==4.62.2
+traitlets==5.0.5
+Twisted==18.9.0
+typed-ast==1.4.3
+typing-extensions==3.10.0.0
+ubuntu-advantage-tools==27.2
+ufw==0.36
+unattended-upgrades==0.1
+urllib3==1.25.8
+virtualenv==20.6.0
+wadllib==1.3.3
+wcwidth==0.2.5
+webencodings==0.5.1
+widgetsnbextension==3.5.1
+xkit==0.0.0
+xxhash==2.0.2
+yamllint==1.23.0
+zipp==1.0.0
+zope.interface==4.7.1
+==============================
+Git revision:
+e3e51338
+==============================
+Cuda version:
+10.2

ASR-model/TransformerLM_seg_char/hyperparams.yaml

@@ -0,0 +1,95 @@
+# Generated 2021-10-05 from:
+# /mnt/md0/user_wayne/speechbrain/recipes/MATBN/LM/hparams/TransformerLM_seg_char.yaml
+# yamllint disable
+output_folder: results/TransformerLM_seg_char
+save_folder: results/TransformerLM_seg_char/save
+train_log: results/TransformerLM_seg_char/train_log.txt
+num_workers: 4
+
+data_folder: results/prepare_seg
+
+tokenizer_file: results/tokenizer_seg_bpe5k_char/5000_char.model
+
+tokenizer: &id001 !new:sentencepiece.SentencePieceProcessor
+
+pretrainer: !new:speechbrain.utils.parameter_transfer.Pretrainer
+  collect_in: results/TransformerLM_seg_char/tokenizer
+  loadables:
+    tokenizer: *id001
+  paths:
+    tokenizer: results/tokenizer_seg_bpe5k_char/5000_char.model
+
+train_logger: !new:speechbrain.utils.train_logger.FileTrainLogger
+  save_file: results/TransformerLM_seg_char/train_log.txt
+
+# Training parameters
+number_of_epochs: 20
+batch_size: 64
+lr: 1
+accumulation_steps: 2
+ckpt_interval_minutes: 15
+
+epoch_counter: &id004 !new:speechbrain.utils.epoch_loop.EpochCounter
+
+  limit: 20
+
+# Dataloader options
+train_dataloader_opts:
+  batch_size: 64
+  num_workers: 4
+  shuffle: true
+  pin_memory: true
+
+valid_dataloader_opts:
+  batch_size: 64
+  num_workers: 4
+
+test_dataloader_opts:
+  batch_size: 64
+  num_workers: 4
+
+# Model parameters
+d_model: 576
+
+# Outputs
+output_neurons: 5000
+blank_index: 0
+bos_index: 1
+eos_index: 2
+unk_index: 0
+pad_index: 0
+
+model: &id002 !new:speechbrain.lobes.models.transformer.TransformerLM.TransformerLM
+
+  vocab: 5000
+  d_model: 576
+  nhead: 6
+  num_encoder_layers: 6
+  num_decoder_layers: 0
+  d_ffn: 1538
+  dropout: 0.2
+  activation: !name:torch.nn.GELU
+  normalize_before: false
+
+modules:
+  model: *id002
+lr_annealing: &id003 !new:speechbrain.nnet.schedulers.NoamScheduler
+  lr_initial: 1
+  n_warmup_steps: 1000
+  model_size: 576
+
+checkpointer: !new:speechbrain.utils.checkpoints.Checkpointer
+  checkpoints_dir: results/TransformerLM_seg_char/save
+  recoverables:
+    model: *id002
+    scheduler: *id003
+    counter: *id004
+log_softmax: !new:speechbrain.nnet.activations.Softmax
+  apply_log: true
+
+optimizer: !name:torch.optim.Adam
+  lr: 0
+  betas: (0.9, 0.98)
+  eps: 0.000000001
+
+compute_cost: !name:speechbrain.nnet.losses.nll_loss

ASR-model/TransformerLM_seg_char/log.txt

@@ -0,0 +1,276 @@
+2021-10-05 19:59:24,051 - speechbrain.core - INFO - Beginning experiment!
+2021-10-05 19:59:24,052 - speechbrain.core - INFO - Experiment folder: results/TransformerLM_seg_char
+2021-10-05 19:59:24,902 - speechbrain.utils.superpowers - DEBUG - appdirs==1.4.4
+argon2-cffi==20.1.0
+async-generator==1.10
+attrs==19.3.0
+Automat==0.8.0
+autopep8==1.5.7
+backcall==0.2.0
+backports.entry-points-selectable==1.1.0
+black==19.10b0
+bleach==3.3.1
+blessings==1.7
+blinker==1.4
+bottle==0.12.19
+certifi==2019.11.28
+cffi==1.14.6
+cfgv==3.3.0
+chardet==3.0.4
+Click==7.0
+cloud-init==21.2
+colorama==0.4.3
+command-not-found==0.3
+configobj==5.0.6
+constantly==15.1.0
+cryptography==2.8
+cupshelpers==1.0
+cycler==0.10.0
+d2l==0.16.6
+datasets==1.11.0
+dbus-python==1.2.16
+debugpy==1.3.0
+decorator==5.0.9
+defer==1.0.6
+defusedxml==0.7.1
+dill==0.3.4
+distlib==0.3.2
+distro==1.4.0
+distro-info===0.23ubuntu1
+entrypoints==0.3
+filelock==3.0.12
+flake8==3.7.9
+fsspec==2021.7.0
+gpustat==0.6.0
+gpuview==0.4.0
+httplib2==0.14.0
+huggingface-hub==0.0.16
+hyperlink==19.0.0
+HyperPyYAML==1.0.0
+identify==2.2.11
+idna==2.8
+importlib-metadata==1.5.0
+incremental==16.10.1
+ipykernel==6.0.2
+ipython==7.25.0
+ipython-genutils==0.2.0
+ipywidgets==7.6.3
+jedi==0.18.0
+Jinja2==2.10.1
+joblib==1.0.1
+jsonpatch==1.22
+jsonpointer==2.0
+jsonschema==3.2.0
+jupyter==1.0.0
+jupyter-client==6.1.12
+jupyter-console==6.4.0
+jupyter-core==4.7.1
+jupyterlab-pygments==0.1.2
+jupyterlab-widgets==1.0.0
+keyring==18.0.1
+kiwisolver==1.3.1
+language-selector==0.1
+launchpadlib==1.10.13
+lazr.restfulclient==0.14.2
+lazr.uri==1.0.3
+macaroonbakery==1.3.1
+MarkupSafe==1.1.0
+matplotlib==3.4.2
+matplotlib-inline==0.1.2
+mccabe==0.6.1
+mistune==0.8.4
+more-itertools==4.2.0
+multiprocess==0.70.12.2
+nbclient==0.5.3
+nbconvert==6.1.0
+nbformat==5.1.3
+nest-asyncio==1.5.1
+netifaces==0.10.4
+nodeenv==1.6.0
+notebook==6.4.0
+numpy==1.21.2
+nvidia-ml-py3==7.352.0
+oauthlib==3.1.0
+packaging==21.0
+pandas==1.3.0
+pandocfilters==1.4.3
+parso==0.8.2
+pathspec==0.9.0
+pexpect==4.6.0
+pickleshare==0.7.5
+Pillow==8.3.1
+platformdirs==2.0.2
+pluggy==0.13.1
+pre-commit==2.15.0
+prometheus-client==0.11.0
+prompt-toolkit==3.0.19
+protobuf==3.6.1
+psutil==5.8.0
+ptyprocess==0.7.0
+py==1.10.0
+pyarrow==5.0.0
+pyasn1==0.4.2
+pyasn1-modules==0.2.1
+pycairo==1.16.2
+pycodestyle==2.5.0
+pycparser==2.20
+pycups==1.9.73
+pyflakes==2.1.1
+Pygments==2.9.0
+PyGObject==3.36.0
+PyHamcrest==1.9.0
+PyJWT==1.7.1
+pymacaroons==0.13.0
+PyMySQL==1.0.2
+PyNaCl==1.3.0
+pyOpenSSL==19.0.0
+pyparsing==2.4.7
+pyRFC3339==1.1
+pyrsistent==0.15.5
+pyserial==3.4
+pytest==5.4.1
+python-apt==2.0.0+ubuntu0.20.4.5
+python-dateutil==2.8.2
+python-debian===0.1.36ubuntu1
+pytube==10.9.3
+pytz==2019.3
+PyYAML==5.3.1
+pyzmq==22.1.0
+qtconsole==5.1.1
+QtPy==1.9.0
+regex==2021.7.6
+requests==2.22.0
+requests-unixsocket==0.2.0
+ruamel.yaml==0.17.10
+ruamel.yaml.clib==0.2.6
+scipy==1.7.1
+screen-resolution-extra==0.0.0
+SecretStorage==2.3.1
+Send2Trash==1.7.1
+sentencepiece==0.1.96
+service-identity==18.1.0
+simplejson==3.16.0
+six==1.14.0
+sos==4.1
+-e git+https://github.com/speechbrain/speechbrain.git@1d194bfc51ae20b9e38596d220cdf0f4977e69de#egg=speechbrain
+ssh-import-id==5.10
+supervisor==4.1.0
+systemd-python==234
+terminado==0.10.1
+testpath==0.5.0
+toml==0.10.2
+torch==1.8.1
+torchaudio==0.8.1
+torchvision==0.10.0
+tornado==6.1
+tqdm==4.62.2
+traitlets==5.0.5
+Twisted==18.9.0
+typed-ast==1.4.3
+typing-extensions==3.10.0.0
+ubuntu-advantage-tools==27.2
+ufw==0.36
+unattended-upgrades==0.1
+urllib3==1.25.8
+virtualenv==20.6.0
+wadllib==1.3.3
+wcwidth==0.2.5
+webencodings==0.5.1
+widgetsnbextension==3.5.1
+xkit==0.0.0
+xxhash==2.0.2
+yamllint==1.23.0
+zipp==1.0.0
+zope.interface==4.7.1
+
+
+2021-10-05 19:59:24,907 - speechbrain.utils.superpowers - DEBUG - e3e51338
+
+
+2021-10-05 19:59:24,936 - speechbrain.utils.parameter_transfer - DEBUG - Collecting files (or symlinks) for pretraining in results/TransformerLM_seg_char/tokenizer.
+2021-10-05 19:59:24,937 - speechbrain.pretrained.fetching - INFO - Fetch 5000_char.model: Linking to local file in /mnt/md0/user_wayne/speechbrain/recipes/MATBN/results/tokenizer_seg_bpe5k_char/5000_char.model.
+2021-10-05 19:59:24,937 - speechbrain.utils.parameter_transfer - INFO - Loading pretrained files for: tokenizer
+2021-10-05 19:59:25,190 - speechbrain.core - INFO - Info: ckpt_interval_minutes arg from hparam file is used
+2021-10-05 19:59:27,052 - speechbrain.core - INFO - 24.7M trainable parameters in LM
+2021-10-05 19:59:27,053 - speechbrain.utils.checkpoints - INFO - Would load a checkpoint here, but none found yet.
+2021-10-05 19:59:27,053 - speechbrain.utils.epoch_loop - INFO - Going into epoch 1
+2021-10-05 20:03:07,476 - speechbrain.utils.train_logger - INFO - epoch: 1, lr: 4.99e-04 - train loss: 4.23 - valid loss: 3.19
+2021-10-05 20:03:07,790 - speechbrain.utils.checkpoints - INFO - Saved an end-of-epoch checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-03-07+00
+2021-10-05 20:03:07,792 - speechbrain.utils.epoch_loop - INFO - Going into epoch 2
+2021-10-05 20:06:50,852 - speechbrain.utils.train_logger - INFO - epoch: 2, lr: 1.00e-03 - train loss: 2.95 - valid loss: 2.77
+2021-10-05 20:06:51,157 - speechbrain.utils.checkpoints - INFO - Saved an end-of-epoch checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-06-50+00
+2021-10-05 20:06:51,186 - speechbrain.utils.checkpoints - INFO - Deleted checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-03-07+00
+2021-10-05 20:06:51,186 - speechbrain.utils.epoch_loop - INFO - Going into epoch 3
+2021-10-05 20:10:34,626 - speechbrain.utils.train_logger - INFO - epoch: 3, lr: 1.23e-03 - train loss: 2.73 - valid loss: 2.64
+2021-10-05 20:10:34,936 - speechbrain.utils.checkpoints - INFO - Saved an end-of-epoch checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-10-34+00
+2021-10-05 20:10:34,967 - speechbrain.utils.checkpoints - INFO - Deleted checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-06-50+00
+2021-10-05 20:10:34,967 - speechbrain.utils.epoch_loop - INFO - Going into epoch 4
+2021-10-05 20:14:18,480 - speechbrain.utils.train_logger - INFO - epoch: 4, lr: 1.07e-03 - train loss: 2.60 - valid loss: 2.54
+2021-10-05 20:14:18,787 - speechbrain.utils.checkpoints - INFO - Saved an end-of-epoch checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-14-18+00
+2021-10-05 20:14:18,820 - speechbrain.utils.checkpoints - INFO - Deleted checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-10-34+00
+2021-10-05 20:14:18,821 - speechbrain.utils.epoch_loop - INFO - Going into epoch 5
+2021-10-05 20:18:02,905 - speechbrain.utils.train_logger - INFO - epoch: 5, lr: 9.56e-04 - train loss: 2.49 - valid loss: 2.48
+2021-10-05 20:18:03,213 - speechbrain.utils.checkpoints - INFO - Saved an end-of-epoch checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-18-02+00
+2021-10-05 20:18:03,247 - speechbrain.utils.checkpoints - INFO - Deleted checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-14-18+00
+2021-10-05 20:18:03,247 - speechbrain.utils.epoch_loop - INFO - Going into epoch 6
+2021-10-05 20:21:47,703 - speechbrain.utils.train_logger - INFO - epoch: 6, lr: 8.73e-04 - train loss: 2.41 - valid loss: 2.42
+2021-10-05 20:21:48,011 - speechbrain.utils.checkpoints - INFO - Saved an end-of-epoch checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-21-47+00
+2021-10-05 20:21:48,048 - speechbrain.utils.checkpoints - INFO - Deleted checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-18-02+00
+2021-10-05 20:21:48,048 - speechbrain.utils.epoch_loop - INFO - Going into epoch 7
+2021-10-05 20:25:32,303 - speechbrain.utils.train_logger - INFO - epoch: 7, lr: 8.08e-04 - train loss: 2.35 - valid loss: 2.38
+2021-10-05 20:25:32,612 - speechbrain.utils.checkpoints - INFO - Saved an end-of-epoch checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-25-32+00
+2021-10-05 20:25:32,650 - speechbrain.utils.checkpoints - INFO - Deleted checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-21-47+00
+2021-10-05 20:25:32,651 - speechbrain.utils.epoch_loop - INFO - Going into epoch 8
+2021-10-05 20:29:17,400 - speechbrain.utils.train_logger - INFO - epoch: 8, lr: 7.56e-04 - train loss: 2.29 - valid loss: 2.35
+2021-10-05 20:29:17,706 - speechbrain.utils.checkpoints - INFO - Saved an end-of-epoch checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-29-17+00
+2021-10-05 20:29:17,745 - speechbrain.utils.checkpoints - INFO - Deleted checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-25-32+00
+2021-10-05 20:29:17,746 - speechbrain.utils.epoch_loop - INFO - Going into epoch 9
+2021-10-05 20:33:02,633 - speechbrain.utils.train_logger - INFO - epoch: 9, lr: 7.13e-04 - train loss: 2.24 - valid loss: 2.33
+2021-10-05 20:33:02,943 - speechbrain.utils.checkpoints - INFO - Saved an end-of-epoch checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-33-02+00
+2021-10-05 20:33:02,985 - speechbrain.utils.checkpoints - INFO - Deleted checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-29-17+00
+2021-10-05 20:33:02,985 - speechbrain.utils.epoch_loop - INFO - Going into epoch 10
+2021-10-05 20:36:47,808 - speechbrain.utils.train_logger - INFO - epoch: 10, lr: 6.76e-04 - train loss: 2.19 - valid loss: 2.32
+2021-10-05 20:36:48,118 - speechbrain.utils.checkpoints - INFO - Saved an end-of-epoch checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-36-47+00
+2021-10-05 20:36:48,161 - speechbrain.utils.checkpoints - INFO - Deleted checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-33-02+00
+2021-10-05 20:36:48,161 - speechbrain.utils.epoch_loop - INFO - Going into epoch 11
+2021-10-05 20:40:33,763 - speechbrain.utils.train_logger - INFO - epoch: 11, lr: 6.45e-04 - train loss: 2.15 - valid loss: 2.30
+2021-10-05 20:40:34,072 - speechbrain.utils.checkpoints - INFO - Saved an end-of-epoch checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-40-33+00
+2021-10-05 20:40:34,117 - speechbrain.utils.checkpoints - INFO - Deleted checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-36-47+00
+2021-10-05 20:40:34,117 - speechbrain.utils.epoch_loop - INFO - Going into epoch 12
+2021-10-05 20:44:19,873 - speechbrain.utils.train_logger - INFO - epoch: 12, lr: 6.17e-04 - train loss: 2.11 - valid loss: 2.30
+2021-10-05 20:44:20,181 - speechbrain.utils.checkpoints - INFO - Saved an end-of-epoch checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-44-19+00
+2021-10-05 20:44:20,228 - speechbrain.utils.checkpoints - INFO - Deleted checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-40-33+00
+2021-10-05 20:44:20,228 - speechbrain.utils.epoch_loop - INFO - Going into epoch 13
+2021-10-05 20:48:05,429 - speechbrain.utils.train_logger - INFO - epoch: 13, lr: 5.93e-04 - train loss: 2.08 - valid loss: 2.29
+2021-10-05 20:48:05,739 - speechbrain.utils.checkpoints - INFO - Saved an end-of-epoch checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-48-05+00
+2021-10-05 20:48:05,788 - speechbrain.utils.checkpoints - INFO - Deleted checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-44-19+00
+2021-10-05 20:48:05,789 - speechbrain.utils.epoch_loop - INFO - Going into epoch 14
+2021-10-05 20:51:51,235 - speechbrain.utils.train_logger - INFO - epoch: 14, lr: 5.71e-04 - train loss: 2.04 - valid loss: 2.29
+2021-10-05 20:51:51,545 - speechbrain.utils.checkpoints - INFO - Saved an end-of-epoch checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-51-51+00
+2021-10-05 20:51:51,596 - speechbrain.utils.checkpoints - INFO - Deleted checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-48-05+00
+2021-10-05 20:51:51,596 - speechbrain.utils.epoch_loop - INFO - Going into epoch 15
+2021-10-05 20:55:37,264 - speechbrain.utils.train_logger - INFO - epoch: 15, lr: 5.52e-04 - train loss: 2.01 - valid loss: 2.28
+2021-10-05 20:55:37,574 - speechbrain.utils.checkpoints - INFO - Saved an end-of-epoch checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-55-37+00
+2021-10-05 20:55:37,626 - speechbrain.utils.checkpoints - INFO - Deleted checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-51-51+00
+2021-10-05 20:55:37,627 - speechbrain.utils.epoch_loop - INFO - Going into epoch 16
+2021-10-05 20:59:22,956 - speechbrain.utils.train_logger - INFO - epoch: 16, lr: 5.34e-04 - train loss: 1.98 - valid loss: 2.28
+2021-10-05 20:59:23,263 - speechbrain.utils.checkpoints - INFO - Saved an end-of-epoch checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-59-22+00
+2021-10-05 20:59:23,293 - speechbrain.utils.epoch_loop - INFO - Going into epoch 17
+2021-10-05 21:03:08,924 - speechbrain.utils.train_logger - INFO - epoch: 17, lr: 5.18e-04 - train loss: 1.95 - valid loss: 2.29
+2021-10-05 21:03:09,230 - speechbrain.utils.checkpoints - INFO - Saved an end-of-epoch checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+21-03-08+00
+2021-10-05 21:03:09,301 - speechbrain.utils.checkpoints - INFO - Deleted checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-59-22+00
+2021-10-05 21:03:09,301 - speechbrain.utils.epoch_loop - INFO - Going into epoch 18
+2021-10-05 21:06:55,028 - speechbrain.utils.train_logger - INFO - epoch: 18, lr: 5.04e-04 - train loss: 1.92 - valid loss: 2.29
+2021-10-05 21:06:55,335 - speechbrain.utils.checkpoints - INFO - Saved an end-of-epoch checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+21-06-55+00
+2021-10-05 21:06:55,408 - speechbrain.utils.checkpoints - INFO - Deleted checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+21-03-08+00
+2021-10-05 21:06:55,409 - speechbrain.utils.epoch_loop - INFO - Going into epoch 19
+2021-10-05 21:10:40,930 - speechbrain.utils.train_logger - INFO - epoch: 19, lr: 4.90e-04 - train loss: 1.90 - valid loss: 2.30
+2021-10-05 21:10:41,237 - speechbrain.utils.checkpoints - INFO - Saved an end-of-epoch checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+21-10-40+00
+2021-10-05 21:10:41,312 - speechbrain.utils.checkpoints - INFO - Deleted checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+21-06-55+00
+2021-10-05 21:10:41,312 - speechbrain.utils.epoch_loop - INFO - Going into epoch 20
+2021-10-05 21:14:27,278 - speechbrain.utils.train_logger - INFO - epoch: 20, lr: 4.78e-04 - train loss: 1.87 - valid loss: 2.30
+2021-10-05 21:14:27,586 - speechbrain.utils.checkpoints - INFO - Saved an end-of-epoch checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+21-14-27+00
+2021-10-05 21:14:27,663 - speechbrain.utils.checkpoints - INFO - Deleted checkpoint in results/TransformerLM_seg_char/save/CKPT+2021-10-05+21-10-40+00
+2021-10-05 21:14:27,664 - speechbrain.utils.checkpoints - INFO - Loading a checkpoint from results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-55-37+00
+2021-10-05 21:14:27,706 - root - DEBUG - SaveableDataLoader was requested to load a checkpoint, but the DataLoader has already been iterated. The DataLoader file will be ignored. This is normal in evaluation, when a checkpoint is loaded just to retrieve the best model.
+2021-10-05 21:14:31,760 - speechbrain.utils.train_logger - INFO - Epoch loaded: 15 - test loss: 2.68

ASR-model/TransformerLM_seg_char/save/CKPT+2021-10-05+20-55-37+00/CKPT.yaml

@@ -0,0 +1,4 @@
+# yamllint disable
+end-of-epoch: true
+loss: 2.282208064707314
+unixtime: 1633438537.264693

ASR-model/TransformerLM_seg_char/save/CKPT+2021-10-05+20-55-37+00/brain.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9e24193f36931b7f57932532efbdcf64971f42732383ba6808825f77db258f6
+size 28

ASR-model/TransformerLM_seg_char/save/CKPT+2021-10-05+20-55-37+00/counter.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e629fa6598d732768f7c726b4b621285f9c3b85303900aa912017db7617d8bdb
+size 2

ASR-model/TransformerLM_seg_char/save/CKPT+2021-10-05+20-55-37+00/dataloader-TRAIN.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c78961d3d782d8a85d9344eedae027f43ce6b9fd35c8f355861a39e0d0ddecc5
+size 3

ASR-model/TransformerLM_seg_char/save/CKPT+2021-10-05+20-55-37+00/model.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7a86f5034ce4797ae0fd5c91b5aa1d9cf370c4b00e12cbb30ec43f484b9aed89
+size 104726057

ASR-model/TransformerLM_seg_char/save/CKPT+2021-10-05+20-55-37+00/optimizer.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1cfecb22673760d25f6bfc67321a0ff1e9f46cbfdab6e6671fea25a6397e7d3e
+size 197914775

ASR-model/TransformerLM_seg_char/save/CKPT+2021-10-05+20-55-37+00/scheduler.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31a2c1b304e48d83777679d5d8dae37370952827c011eceb6b0e55ed8f848574
+size 431

ASR-model/TransformerLM_seg_char/save/CKPT+2021-10-05+21-14-27+00/CKPT.yaml

@@ -0,0 +1,4 @@
+# yamllint disable
+end-of-epoch: true
+loss: 2.3031107332648304
+unixtime: 1633439667.279289

ASR-model/TransformerLM_seg_char/save/CKPT+2021-10-05+21-14-27+00/brain.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9e24193f36931b7f57932532efbdcf64971f42732383ba6808825f77db258f6
+size 28

ASR-model/TransformerLM_seg_char/save/CKPT+2021-10-05+21-14-27+00/counter.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f5ca38f748a1d6eaf726b8a42fb575c3c71f1864a8143301782de13da2d9202b
+size 2

ASR-model/TransformerLM_seg_char/save/CKPT+2021-10-05+21-14-27+00/dataloader-TRAIN.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c78961d3d782d8a85d9344eedae027f43ce6b9fd35c8f355861a39e0d0ddecc5
+size 3

ASR-model/TransformerLM_seg_char/save/CKPT+2021-10-05+21-14-27+00/model.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5610f4428ab9e40a6a69f2a009d6fbd1b2ac25470b6a47f139172ca264c6ea1a
+size 104726057

ASR-model/TransformerLM_seg_char/save/CKPT+2021-10-05+21-14-27+00/optimizer.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1e61f107260662bbb7b6b74433922eb0ad516163c7a478ebe6466037eed8c489
+size 197914775

ASR-model/TransformerLM_seg_char/save/CKPT+2021-10-05+21-14-27+00/scheduler.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:60e6d1257c2e1fc1ea9c16b54b3f5d6690602634979718bde831b1084b107c84
+size 431

ASR-model/TransformerLM_seg_char/train.py

@@ -0,0 +1,150 @@
+import sys
+
+import torch
+import speechbrain as sb
+from speechbrain.dataio import dataset
+from speechbrain.utils.distributed import run_on_main
+from hyperpyyaml import load_hyperpyyaml
+
+
+class LM(sb.core.Brain):
+    def compute_forward(self, batch, stage):
+        batch = batch.to(self.device)
+        tokens_bos, _ = batch.tokens_bos
+        logits = self.hparams.model(tokens_bos)
+        pred = self.hparams.log_softmax(logits)
+        return pred
+
+    def compute_objectives(self, predictions, batch, stage):
+        batch = batch.to(self.device)
+        tokens_eos, tokens_len = batch.tokens_eos
+        loss = self.hparams.compute_cost(
+            predictions, tokens_eos, length=tokens_len
+        )
+        return loss
+
+    def fit_batch(self, batch):
+        predictions = self.compute_forward(batch, sb.Stage.TRAIN)
+        loss = self.compute_objectives(predictions, batch, sb.Stage.TRAIN)
+
+        (loss / self.hparams.accumulation_steps).backward()
+
+        if self.step % self.hparams.accumulation_steps == 0:
+            self.check_gradients(loss)
+
+            self.optimizer.step()
+            self.optimizer.zero_grad()
+
+            if isinstance(
+                self.hparams.lr_annealing, sb.nnet.schedulers.NoamScheduler
+            ) or isinstance(
+                self.hparams.lr_annealing,
+                sb.nnet.schedulers.CyclicCosineScheduler,
+            ):
+                self.hparams.lr_annealing(self.optimizer)
+
+        return loss
+
+    def on_stage_end(self, stage, stage_loss, epoch):
+        stage_stats = {"loss": stage_loss}
+        if stage == sb.Stage.TRAIN:
+            self.train_stats = stage_stats
+
+        if stage == sb.Stage.VALID and sb.utils.distributed.if_main_process():
+            if not (
+                isinstance(
+                    self.hparams.lr_annealing, sb.nnet.schedulers.NoamScheduler
+                )
+                or isinstance(
+                    self.hparams.lr_annealing,
+                    sb.nnet.schedulers.CyclicCosineScheduler,
+                )
+            ):
+                old_lr, new_lr = self.hparams.lr_annealing(stage_loss)
+                sb.nnet.schedulers.update_learning_rate(self.optimizer, new_lr)
+            else:
+                old_lr = self.hparams.lr_annealing.current_lr
+
+            self.hparams.train_logger.log_stats(
+                stats_meta={"epoch": epoch, "lr": old_lr},
+                train_stats=self.train_stats,
+                valid_stats=stage_stats,
+            )
+            self.checkpointer.save_and_keep_only(
+                meta=stage_stats, min_keys=["loss"],
+            )
+
+        if stage == sb.Stage.TEST and sb.utils.distributed.if_main_process():
+            self.hparams.train_logger.log_stats(
+                stats_meta={"Epoch loaded": self.hparams.epoch_counter.current},
+                test_stats=stage_stats,
+            )
+
+
+def dataio_prepare(hparams):
+    @sb.utils.data_pipeline.takes("transcription")
+    @sb.utils.data_pipeline.provides(
+        "transcription", "tokens_bos", "tokens_eos"
+    )
+    def transcription_pipline(transcription):
+        yield transcription
+        tokens_list = hparams["tokenizer"].encode_as_ids(transcription)
+        tokens_bos = torch.LongTensor([hparams["bos_index"]] + (tokens_list))
+        yield tokens_bos
+        tokens_eos = torch.LongTensor(tokens_list + [hparams["eos_index"]])
+        yield tokens_eos
+
+    data_folder = hparams["data_folder"]
+    datasets = {}
+    for dataset_name in ["train", "dev", "test"]:
+        json_path = f"{data_folder}/{dataset_name}.json"
+        datasets[dataset_name] = dataset.DynamicItemDataset.from_json(
+            json_path=json_path,
+            replacements={"data_root": data_folder},
+            dynamic_items=[transcription_pipline],
+            output_keys=["transcription", "tokens_bos", "tokens_eos"],
+        )
+
+    return datasets
+
+
+if __name__ == "__main__":
+    hparams_file_path, run_opts, overrides = sb.parse_arguments(sys.argv[1:])
+    with open(hparams_file_path) as hparams_file:
+        hparams = load_hyperpyyaml(hparams_file, overrides)
+
+    sb.utils.distributed.ddp_init_group(run_opts)
+
+    sb.create_experiment_directory(
+        experiment_directory=hparams["output_folder"],
+        hyperparams_to_save=hparams_file_path,
+        overrides=overrides,
+    )
+
+    run_on_main(hparams["pretrainer"].collect_files)
+    hparams["pretrainer"].load_collected(device=run_opts["device"])
+
+    datasets = dataio_prepare(hparams)
+
+    lm_brain = LM(
+        modules=hparams["modules"],
+        opt_class=hparams["optimizer"],
+        hparams=hparams,
+        run_opts=run_opts,
+        checkpointer=hparams["checkpointer"],
+    )
+
+    lm_brain.fit(
+        lm_brain.hparams.epoch_counter,
+        datasets["train"],
+        datasets["dev"],
+        train_loader_kwargs=hparams["train_dataloader_opts"],
+        valid_loader_kwargs=hparams["valid_dataloader_opts"],
+    )
+
+    # evaluation
+    lm_brain.evaluate(
+        datasets["test"],
+        min_key="loss",
+        test_loader_kwargs=hparams["test_dataloader_opts"],
+    )

ASR-model/TransformerLM_seg_char/train_log.txt

@@ -0,0 +1,21 @@
+epoch: 1, lr: 4.99e-04 - train loss: 4.23 - valid loss: 3.19
+epoch: 2, lr: 1.00e-03 - train loss: 2.95 - valid loss: 2.77
+epoch: 3, lr: 1.23e-03 - train loss: 2.73 - valid loss: 2.64
+epoch: 4, lr: 1.07e-03 - train loss: 2.60 - valid loss: 2.54
+epoch: 5, lr: 9.56e-04 - train loss: 2.49 - valid loss: 2.48
+epoch: 6, lr: 8.73e-04 - train loss: 2.41 - valid loss: 2.42
+epoch: 7, lr: 8.08e-04 - train loss: 2.35 - valid loss: 2.38
+epoch: 8, lr: 7.56e-04 - train loss: 2.29 - valid loss: 2.35
+epoch: 9, lr: 7.13e-04 - train loss: 2.24 - valid loss: 2.33
+epoch: 10, lr: 6.76e-04 - train loss: 2.19 - valid loss: 2.32
+epoch: 11, lr: 6.45e-04 - train loss: 2.15 - valid loss: 2.30
+epoch: 12, lr: 6.17e-04 - train loss: 2.11 - valid loss: 2.30
+epoch: 13, lr: 5.93e-04 - train loss: 2.08 - valid loss: 2.29
+epoch: 14, lr: 5.71e-04 - train loss: 2.04 - valid loss: 2.29
+epoch: 15, lr: 5.52e-04 - train loss: 2.01 - valid loss: 2.28
+epoch: 16, lr: 5.34e-04 - train loss: 1.98 - valid loss: 2.28
+epoch: 17, lr: 5.18e-04 - train loss: 1.95 - valid loss: 2.29
+epoch: 18, lr: 5.04e-04 - train loss: 1.92 - valid loss: 2.29
+epoch: 19, lr: 4.90e-04 - train loss: 1.90 - valid loss: 2.30
+epoch: 20, lr: 4.78e-04 - train loss: 1.87 - valid loss: 2.30
+Epoch loaded: 15 - test loss: 2.68

ASR-model/asr_transformer_seg_char_ctc0.3/env.log

@@ -0,0 +1,195 @@
+SpeechBrain system description
+==============================
+Python version:
+3.8.10 (default, Sep 28 2021, 16:10:42) 
+[GCC 9.3.0]
+==============================
+Installed Python packages:
+appdirs==1.4.4
+argon2-cffi==20.1.0
+async-generator==1.10
+attrs==19.3.0
+Automat==0.8.0
+autopep8==1.5.7
+backcall==0.2.0
+backports.entry-points-selectable==1.1.0
+black==19.10b0
+bleach==3.3.1
+blessings==1.7
+blinker==1.4
+bottle==0.12.19
+certifi==2019.11.28
+cffi==1.14.6
+cfgv==3.3.0
+chardet==3.0.4
+Click==7.0
+cloud-init==21.2
+colorama==0.4.3
+command-not-found==0.3
+configobj==5.0.6
+constantly==15.1.0
+cryptography==2.8
+cupshelpers==1.0
+cycler==0.10.0
+d2l==0.16.6
+datasets==1.11.0
+dbus-python==1.2.16
+debugpy==1.3.0
+decorator==5.0.9
+defer==1.0.6
+defusedxml==0.7.1
+dill==0.3.4
+distlib==0.3.2
+distro==1.4.0
+distro-info===0.23ubuntu1
+entrypoints==0.3
+filelock==3.0.12
+flake8==3.7.9
+fsspec==2021.7.0
+gpustat==0.6.0
+gpuview==0.4.0
+httplib2==0.14.0
+huggingface-hub==0.0.16
+hyperlink==19.0.0
+HyperPyYAML==1.0.0
+identify==2.2.11
+idna==2.8
+importlib-metadata==1.5.0
+incremental==16.10.1
+ipykernel==6.0.2
+ipython==7.25.0
+ipython-genutils==0.2.0
+ipywidgets==7.6.3
+jedi==0.18.0
+Jinja2==2.10.1
+joblib==1.0.1
+jsonpatch==1.22
+jsonpointer==2.0
+jsonschema==3.2.0
+jupyter==1.0.0
+jupyter-client==6.1.12
+jupyter-console==6.4.0
+jupyter-core==4.7.1
+jupyterlab-pygments==0.1.2
+jupyterlab-widgets==1.0.0
+keyring==18.0.1
+kiwisolver==1.3.1
+language-selector==0.1
+launchpadlib==1.10.13
+lazr.restfulclient==0.14.2
+lazr.uri==1.0.3
+macaroonbakery==1.3.1
+MarkupSafe==1.1.0
+matplotlib==3.4.2
+matplotlib-inline==0.1.2
+mccabe==0.6.1
+mistune==0.8.4
+more-itertools==4.2.0
+multiprocess==0.70.12.2
+nbclient==0.5.3
+nbconvert==6.1.0
+nbformat==5.1.3
+nest-asyncio==1.5.1
+netifaces==0.10.4
+nodeenv==1.6.0
+notebook==6.4.0
+numpy==1.21.2
+nvidia-ml-py3==7.352.0
+oauthlib==3.1.0
+packaging==21.0
+pandas==1.3.0
+pandocfilters==1.4.3
+parso==0.8.2
+pathspec==0.9.0
+pexpect==4.6.0
+pickleshare==0.7.5
+Pillow==8.3.1
+platformdirs==2.0.2
+pluggy==0.13.1
+pre-commit==2.15.0
+prometheus-client==0.11.0
+prompt-toolkit==3.0.19
+protobuf==3.6.1
+psutil==5.8.0
+ptyprocess==0.7.0
+py==1.10.0
+pyarrow==5.0.0
+pyasn1==0.4.2
+pyasn1-modules==0.2.1
+pycairo==1.16.2
+pycodestyle==2.5.0
+pycparser==2.20
+pycups==1.9.73
+pyflakes==2.1.1
+Pygments==2.9.0
+PyGObject==3.36.0
+PyHamcrest==1.9.0
+PyJWT==1.7.1
+pymacaroons==0.13.0
+PyMySQL==1.0.2
+PyNaCl==1.3.0
+pyOpenSSL==19.0.0
+pyparsing==2.4.7
+pyRFC3339==1.1
+pyrsistent==0.15.5
+pyserial==3.4
+pytest==5.4.1
+python-apt==2.0.0+ubuntu0.20.4.5
+python-dateutil==2.8.2
+python-debian===0.1.36ubuntu1
+pytube==10.9.3
+pytz==2019.3
+PyYAML==5.3.1
+pyzmq==22.1.0
+qtconsole==5.1.1
+QtPy==1.9.0
+regex==2021.7.6
+requests==2.22.0
+requests-unixsocket==0.2.0
+ruamel.yaml==0.17.10
+ruamel.yaml.clib==0.2.6
+scipy==1.7.1
+screen-resolution-extra==0.0.0
+SecretStorage==2.3.1
+Send2Trash==1.7.1
+sentencepiece==0.1.96
+service-identity==18.1.0
+simplejson==3.16.0
+six==1.14.0
+sos==4.1
+-e git+https://github.com/speechbrain/speechbrain.git@1d194bfc51ae20b9e38596d220cdf0f4977e69de#egg=speechbrain
+ssh-import-id==5.10
+supervisor==4.1.0
+systemd-python==234
+terminado==0.10.1
+testpath==0.5.0
+toml==0.10.2
+torch==1.8.1
+torchaudio==0.8.1
+torchvision==0.10.0
+tornado==6.1
+tqdm==4.62.2
+traitlets==5.0.5
+Twisted==18.9.0
+typed-ast==1.4.3
+typing-extensions==3.10.0.0
+ubuntu-advantage-tools==27.2
+ufw==0.36
+unattended-upgrades==0.1
+urllib3==1.25.8
+virtualenv==20.6.0
+wadllib==1.3.3
+wcwidth==0.2.5
+webencodings==0.5.1
+widgetsnbextension==3.5.1
+xkit==0.0.0
+xxhash==2.0.2
+yamllint==1.23.0
+zipp==1.0.0
+zope.interface==4.7.1
+==============================
+Git revision:
+e3e51338
+==============================
+Cuda version:
+10.2

ASR-model/asr_transformer_seg_char_ctc0.3/hyperparams.yaml

@@ -0,0 +1,241 @@
+# Generated 2021-10-11 from:
+# /mnt/md0/user_wayne/speechbrain/recipes/MATBN/ASR/hparams/transformer_seg_LM_char_decode.yaml
+# yamllint disable
+output_folder: results/asr_transformer_seg_char
+cer_file: results/asr_transformer_seg_char/cer.txt
+train_log: results/asr_transformer_seg_char/train_log.txt
+save_folder: results/asr_transformer_seg_char/save
+ckpt_interval_minutes: 15
+num_workers: 4
+
+data_folder: results/prepare_seg
+tokenizer_file: results/tokenizer_seg_bpe5k_char/5000_char.model
+lm_file: results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-55-37+00/model.ckpt
+
+tokenizer: &id001 !new:sentencepiece.SentencePieceProcessor
+
+pretrainer: !new:speechbrain.utils.parameter_transfer.Pretrainer
+  collect_in: results/asr_transformer_seg_char/save
+  loadables:
+    lm: &id003 !new:speechbrain.lobes.models.transformer.TransformerLM.TransformerLM
+                                                                        # yamllint disable-line rule:line-length
+      vocab: 5000
+      d_model: 576
+      nhead: 6
+      num_encoder_layers: 6
+      num_decoder_layers: 0
+      d_ffn: 1538
+      dropout: 0.2
+      activation: !name:torch.nn.GELU
+      normalize_before: false
+
+    tokenizer: *id001
+  paths:
+    lm: results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-55-37+00/model.ckpt
+    tokenizer: results/tokenizer_seg_bpe5k_char/5000_char.model
+
+train_logger: !new:speechbrain.utils.train_logger.FileTrainLogger
+  save_file: results/asr_transformer_seg_char/train_log.txt
+
+# Feature parameters
+sample_rate: 16000
+n_fft: 400
+n_mels: 80
+hop_length: 10
+
+compute_features: !new:speechbrain.lobes.features.Fbank
+  sample_rate: 16000
+  n_fft: 400
+  n_mels: 80
+  hop_length: 10
+
+speed_perturb: !new:speechbrain.processing.speech_augmentation.SpeedPerturb
+  orig_freq: 16000
+  speeds: [90, 100, 110]
+
+# Training parameters
+number_of_epochs: 60
+batch_size: 1
+ctc_weight: 0.3
+gradient_accumulation: 32
+gradient_clipping: 5.0
+loss_reduction: batchmean
+sorting: random
+
+# stages related parameters
+stage_one_epochs: 40
+lr_adam: 1.0
+lr_sgd: 0.000025
+
+# Dataloader options
+train_dataloader_opts:
+  batch_size: 1
+  num_workers: 4
+  shuffle: true
+
+valid_dataloader_opts:
+  batch_size: 1
+  num_workers: 4
+
+test_dataloader_opts:
+  batch_size: 1
+  num_workers: 4
+
+####################### Model parameters ###########################
+# Transformer
+d_model: 256
+nhead: 4
+num_encoder_layers: 12
+num_decoder_layers: 6
+d_ffn: 2048
+transformer_dropout: 0.1
+activation: &id002 !name:torch.nn.GELU
+output_neurons: 5000
+vocab_size: 5000
+
+# Outputs
+blank_index: 0
+label_smoothing: 0.1
+pad_index: 0
+bos_index: 1
+eos_index: 2
+unk_index: 0
+
+# Decoding parameters
+min_decode_ratio: 0.0
+max_decode_ratio: 1.0 # 1.0
+valid_search_interval: 10
+valid_beam_size: 10
+test_beam_size: 10
+ctc_weight_decode: 0.30
+lm_weight: 0.9
+
+############################## models ################################
+
+CNN: &id004 !new:speechbrain.lobes.models.convolution.ConvolutionFrontEnd
+  input_shape: (8, 10, 80)
+  num_blocks: 2
+  num_layers_per_block: 1
+  out_channels: (256, 256)
+  kernel_sizes: (3, 3)
+  strides: (2, 2)
+  residuals: (False, False)
+
+Transformer: &id005 !new:speechbrain.lobes.models.transformer.TransformerASR.TransformerASR
+                                                                          # yamllint disable-line rule:line-length
+  input_size: 5120
+  tgt_vocab: 5000
+  d_model: 256
+  nhead: 4
+  num_encoder_layers: 12
+  num_decoder_layers: 6
+  d_ffn: 2048
+  dropout: 0.1
+  activation: *id002
+  normalize_before: true
+
+lm_model: *id003
+ctc_lin: &id007 !new:speechbrain.nnet.linear.Linear
+  input_size: 256
+  n_neurons: 5000
+
+seq_lin: &id006 !new:speechbrain.nnet.linear.Linear
+  input_size: 256
+  n_neurons: 5000
+
+modules:
+  CNN: *id004
+  Transformer: *id005
+  seq_lin: *id006
+  ctc_lin: *id007
+  normalize: &id010 !new:speechbrain.processing.features.InputNormalization
+
+    norm_type: global
+    update_until_epoch: 4
+
+model: &id008 !new:torch.nn.ModuleList
+- [*id004, *id005, *id006, *id007]
+Adam: !name:torch.optim.Adam
+  lr: 0
+  betas: (0.9, 0.98)
+  eps: 0.000000001
+
+SGD: !name:torch.optim.SGD
+  lr: 0.000025
+  momentum: 0.99
+  nesterov: true
+
+valid_search: !new:speechbrain.decoders.S2STransformerBeamSearch
+  modules: [*id005, *id006, *id007]
+  bos_index: 1
+  eos_index: 2
+  blank_index: 0
+  min_decode_ratio: 0.0
+  max_decode_ratio: 1.0
+  beam_size: 10
+  ctc_weight: 0.30
+  using_eos_threshold: false
+  length_normalization: true
+
+test_search: !new:speechbrain.decoders.S2STransformerBeamSearch
+  modules: [*id005, *id006, *id007]
+  bos_index: 1
+  eos_index: 2
+  blank_index: 0
+  min_decode_ratio: 0.0
+  max_decode_ratio: 1.0
+  beam_size: 10
+  ctc_weight: 0.30
+  lm_weight: 0.9
+  lm_modules: *id003
+  temperature: 1.15
+  temperature_lm: 1.15
+  using_eos_threshold: false
+  length_normalization: true
+
+log_softmax: !new:torch.nn.LogSoftmax
+  dim: -1
+
+ctc_cost: !name:speechbrain.nnet.losses.ctc_loss
+  blank_index: 0
+  reduction: batchmean
+
+seq_cost: !name:speechbrain.nnet.losses.kldiv_loss
+  label_smoothing: 0.1
+  reduction: batchmean
+
+noam_annealing: &id009 !new:speechbrain.nnet.schedulers.NoamScheduler
+  lr_initial: 1.0
+  n_warmup_steps: 25000
+  model_size: 256
+
+checkpointer: !new:speechbrain.utils.checkpoints.Checkpointer
+  checkpoints_dir: results/asr_transformer_seg_char/save
+  recoverables:
+    model: *id008
+    noam_scheduler: *id009
+    normalizer: *id010
+    counter: &id011 !new:speechbrain.utils.epoch_loop.EpochCounter
+
+      limit: 60
+
+epoch_counter: *id011
+normalize: *id010
+augmentation: !new:speechbrain.lobes.augment.SpecAugment
+  time_warp: true
+  time_warp_window: 5
+  time_warp_mode: bicubic
+  freq_mask: true
+  n_freq_mask: 2
+  time_mask: true
+  n_time_mask: 2
+  replace_with_zero: false
+  freq_mask_width: 30
+  time_mask_width: 40
+
+remove_spaces: true
+split_tokens: &id012 !apply:operator.not_ [true]
+
+cer_computer: !name:speechbrain.utils.metric_stats.ErrorRateStats
+  split_tokens: *id012
+acc_computer: !name:speechbrain.utils.Accuracy.AccuracyStats

ASR-model/asr_transformer_seg_char_ctc0.3/results/asr_transformer_seg_char/env.log

@@ -0,0 +1,109 @@
+SpeechBrain system description
+==============================
+Python version:
+3.8.16 (default, Jun 12 2023, 18:09:05) 
+[GCC 11.2.0]
+==============================
+Installed Python packages:
+antlr4-python3-runtime==4.8
+anyio==3.7.0
+astroid==2.15.5
+bitarray==2.7.6
+certifi==2023.5.7
+cffi==1.15.1
+cfgv==3.3.1
+charset-normalizer==3.1.0
+click==7.1.2
+cmake==3.26.4
+colorama==0.4.6
+Cython==0.29.36
+dill==0.3.6
+distlib==0.3.6
+exceptiongroup==1.1.2
+fairseq @ git+https://github.com/facebookresearch/fairseq.git@31fba013a070eaff69dec8642e68e7134d60ab0f
+fastapi==0.98.0
+filelock==3.12.2
+flake8==6.0.0
+fsspec==2023.6.0
+h11==0.14.0
+httpcore==0.17.2
+httptools==0.1.2
+httpx==0.24.1
+huggingface-hub==0.15.1
+hydra-core==1.0.7
+HyperPyYAML==1.2.1
+identify==2.5.24
+idna==3.4
+importlib-resources==5.12.0
+iniconfig==2.0.0
+isort==5.12.0
+jieba==0.42.1
+Jinja2==3.1.2
+joblib==1.3.1
+lazy-object-proxy==1.9.0
+lit==16.0.6
+lxml==4.9.2
+MarkupSafe==2.1.3
+mccabe==0.7.0
+mock==5.0.2
+mpmath==1.3.0
+networkx==3.1
+nodeenv==1.8.0
+numpy==1.24.4
+nvidia-cublas-cu11==11.10.3.66
+nvidia-cuda-cupti-cu11==11.7.101
+nvidia-cuda-nvrtc-cu11==11.7.99
+nvidia-cuda-runtime-cu11==11.7.99
+nvidia-cudnn-cu11==8.5.0.96
+nvidia-cufft-cu11==10.9.0.58
+nvidia-curand-cu11==10.2.10.91
+nvidia-cusolver-cu11==11.4.0.1
+nvidia-cusparse-cu11==11.7.4.91
+nvidia-nccl-cu11==2.14.3
+nvidia-nvtx-cu11==11.7.91
+omegaconf==2.0.6
+packaging==23.1
+platformdirs==3.8.0
+pluggy==1.2.0
+portalocker==2.7.0
+pre-commit==3.3.3
+pycodestyle==2.10.0
+pycparser==2.21
+pydantic==1.10.9
+pyflakes==3.0.1
+pylint==2.17.4
+pytest==7.4.0
+python-multipart==0.0.6
+PyYAML==6.0
+regex==2023.6.3
+requests==2.31.0
+ruamel.yaml==0.17.28
+ruamel.yaml.clib==0.2.7
+sacrebleu==2.3.1
+scikit-learn==1.3.0
+scipy==1.10.1
+sentencepiece==0.1.99
+sniffio==1.3.0
+speechbrain==0.5.14
+starlette==0.27.0
+subword-nmt==0.3.8
+sympy==1.12
+tabulate==0.9.0
+threadpoolctl==3.1.0
+tomli==2.0.1
+tomlkit==0.11.8
+torch==2.0.1
+torchaudio==2.0.2
+tqdm==4.65.0
+triton==2.0.0
+typing_extensions==4.7.1
+urllib3==2.0.3
+uvicorn==0.11.3
+uvloop==0.17.0
+virtualenv==20.23.1
+websockets==8.1
+wrapt==1.15.0
+zipp==3.15.0
+==============================
+Could not get git revision==============================
+CUDA not available

ASR-model/asr_transformer_seg_char_ctc0.3/results/asr_transformer_seg_char/hyperparams.yaml

@@ -0,0 +1,244 @@
+# Generated 2023-07-05 from:
+# /mnt/md0/user_sinica_axel/MATBN/results/asr_transformer_seg_char_ctc0.3/hyperparams.yaml
+# yamllint disable
+# Generated 2021-10-11 from:
+# /mnt/md0/user_wayne/speechbrain/recipes/MATBN/ASR/hparams/transformer_seg_LM_char_decode.yaml
+# yamllint disable
+output_folder: results/asr_transformer_seg_char
+cer_file: results/asr_transformer_seg_char/cer.txt
+train_log: results/asr_transformer_seg_char/train_log.txt
+save_folder: results/asr_transformer_seg_char/save
+ckpt_interval_minutes: 15
+num_workers: 4
+
+data_folder: results/prepare_seg
+tokenizer_file: results/tokenizer_seg_bpe5k_char/5000_char.model
+lm_file: results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-55-37+00/model.ckpt
+
+tokenizer: &id001 !new:sentencepiece.SentencePieceProcessor
+
+pretrainer: !new:speechbrain.utils.parameter_transfer.Pretrainer
+  collect_in: results/asr_transformer_seg_char/save
+  loadables:
+    lm: &id003 !new:speechbrain.lobes.models.transformer.TransformerLM.TransformerLM
+                                                                        # yamllint disable-line rule:line-length
+      vocab: 5000
+      d_model: 576
+      nhead: 6
+      num_encoder_layers: 6
+      num_decoder_layers: 0
+      d_ffn: 1538
+      dropout: 0.2
+      activation: !name:torch.nn.GELU
+      normalize_before: false
+
+    tokenizer: *id001
+  paths:
+    lm: results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-55-37+00/model.ckpt
+    tokenizer: results/tokenizer_seg_bpe5k_char/5000_char.model
+
+train_logger: !new:speechbrain.utils.train_logger.FileTrainLogger
+  save_file: results/asr_transformer_seg_char/train_log.txt
+
+# Feature parameters
+sample_rate: 16000
+n_fft: 400
+n_mels: 80
+hop_length: 10
+
+compute_features: !new:speechbrain.lobes.features.Fbank
+  sample_rate: 16000
+  n_fft: 400
+  n_mels: 80
+  hop_length: 10
+
+speed_perturb: !new:speechbrain.processing.speech_augmentation.SpeedPerturb
+  orig_freq: 16000
+  speeds: [90, 100, 110]
+
+# Training parameters
+number_of_epochs: 60
+batch_size: 1
+ctc_weight: 0.3
+gradient_accumulation: 32
+gradient_clipping: 5.0
+loss_reduction: batchmean
+sorting: random
+
+# stages related parameters
+stage_one_epochs: 40
+lr_adam: 1.0
+lr_sgd: 0.000025
+
+# Dataloader options
+train_dataloader_opts:
+  batch_size: 1
+  num_workers: 4
+  shuffle: true
+
+valid_dataloader_opts:
+  batch_size: 1
+  num_workers: 4
+
+test_dataloader_opts:
+  batch_size: 1
+  num_workers: 4
+
+####################### Model parameters ###########################
+# Transformer
+d_model: 256
+nhead: 4
+num_encoder_layers: 12
+num_decoder_layers: 6
+d_ffn: 2048
+transformer_dropout: 0.1
+activation: &id002 !name:torch.nn.GELU
+output_neurons: 5000
+vocab_size: 5000
+
+# Outputs
+blank_index: 0
+label_smoothing: 0.1
+pad_index: 0
+bos_index: 1
+eos_index: 2
+unk_index: 0
+
+# Decoding parameters
+min_decode_ratio: 0.0
+max_decode_ratio: 1.0 # 1.0
+valid_search_interval: 10
+valid_beam_size: 10
+test_beam_size: 10
+ctc_weight_decode: 0.30
+lm_weight: 0.9
+
+############################## models ################################
+
+CNN: &id004 !new:speechbrain.lobes.models.convolution.ConvolutionFrontEnd
+  input_shape: (8, 10, 80)
+  num_blocks: 2
+  num_layers_per_block: 1
+  out_channels: (256, 256)
+  kernel_sizes: (3, 3)
+  strides: (2, 2)
+  residuals: (False, False)
+
+Transformer: &id005 !new:speechbrain.lobes.models.transformer.TransformerASR.TransformerASR
+                                                                          # yamllint disable-line rule:line-length
+  input_size: 5120
+  tgt_vocab: 5000
+  d_model: 256
+  nhead: 4
+  num_encoder_layers: 12
+  num_decoder_layers: 6
+  d_ffn: 2048
+  dropout: 0.1
+  activation: *id002
+  normalize_before: true
+
+lm_model: *id003
+ctc_lin: &id007 !new:speechbrain.nnet.linear.Linear
+  input_size: 256
+  n_neurons: 5000
+
+seq_lin: &id006 !new:speechbrain.nnet.linear.Linear
+  input_size: 256
+  n_neurons: 5000
+
+modules:
+  CNN: *id004
+  Transformer: *id005
+  seq_lin: *id006
+  ctc_lin: *id007
+  normalize: &id010 !new:speechbrain.processing.features.InputNormalization
+
+    norm_type: global
+    update_until_epoch: 4
+
+model: &id008 !new:torch.nn.ModuleList
+- [*id004, *id005, *id006, *id007]
+Adam: !name:torch.optim.Adam
+  lr: 0
+  betas: (0.9, 0.98)
+  eps: 0.000000001
+
+SGD: !name:torch.optim.SGD
+  lr: 0.000025
+  momentum: 0.99
+  nesterov: true
+
+valid_search: !new:speechbrain.decoders.S2STransformerBeamSearch
+  modules: [*id005, *id006, *id007]
+  bos_index: 1
+  eos_index: 2
+  blank_index: 0
+  min_decode_ratio: 0.0
+  max_decode_ratio: 1.0
+  beam_size: 10
+  ctc_weight: 0.30
+  using_eos_threshold: false
+  length_normalization: true
+
+test_search: !new:speechbrain.decoders.S2STransformerBeamSearch
+  modules: [*id005, *id006, *id007]
+  bos_index: 1
+  eos_index: 2
+  blank_index: 0
+  min_decode_ratio: 0.0
+  max_decode_ratio: 1.0
+  beam_size: 10
+  ctc_weight: 0.30
+  lm_weight: 0.9
+  lm_modules: *id003
+  temperature: 1.15
+  temperature_lm: 1.15
+  using_eos_threshold: false
+  length_normalization: true
+
+log_softmax: !new:torch.nn.LogSoftmax
+  dim: -1
+
+ctc_cost: !name:speechbrain.nnet.losses.ctc_loss
+  blank_index: 0
+  reduction: batchmean
+
+seq_cost: !name:speechbrain.nnet.losses.kldiv_loss
+  label_smoothing: 0.1
+  reduction: batchmean
+
+noam_annealing: &id009 !new:speechbrain.nnet.schedulers.NoamScheduler
+  lr_initial: 1.0
+  n_warmup_steps: 25000
+  model_size: 256
+
+checkpointer: !new:speechbrain.utils.checkpoints.Checkpointer
+  checkpoints_dir: results/asr_transformer_seg_char/save
+  recoverables:
+    model: *id008
+    noam_scheduler: *id009
+    normalizer: *id010
+    counter: &id011 !new:speechbrain.utils.epoch_loop.EpochCounter
+
+      limit: 60
+
+epoch_counter: *id011
+normalize: *id010
+augmentation: !new:speechbrain.lobes.augment.SpecAugment
+  time_warp: true
+  time_warp_window: 5
+  time_warp_mode: bicubic
+  freq_mask: true
+  n_freq_mask: 2
+  time_mask: true
+  n_time_mask: 2
+  replace_with_zero: false
+  freq_mask_width: 30
+  time_mask_width: 40
+
+remove_spaces: true
+split_tokens: &id012 !apply:operator.not_ [true]
+
+cer_computer: !name:speechbrain.utils.metric_stats.ErrorRateStats
+  split_tokens: *id012
+acc_computer: !name:speechbrain.utils.Accuracy.AccuracyStats

ASR-model/asr_transformer_seg_char_ctc0.3/results/asr_transformer_seg_char/log.txt

@@ -0,0 +1,120 @@
+2023-07-05 14:57:01,987 - speechbrain.core - INFO - Beginning experiment!
+2023-07-05 14:57:01,987 - speechbrain.core - INFO - Experiment folder: results/asr_transformer_seg_char
+2023-07-05 14:57:02,208 - speechbrain.utils.superpowers - DEBUG - antlr4-python3-runtime==4.8
+anyio==3.7.0
+astroid==2.15.5
+bitarray==2.7.6
+certifi==2023.5.7
+cffi==1.15.1
+cfgv==3.3.1
+charset-normalizer==3.1.0
+click==7.1.2
+cmake==3.26.4
+colorama==0.4.6
+Cython==0.29.36
+dill==0.3.6
+distlib==0.3.6
+exceptiongroup==1.1.2
+fairseq @ git+https://github.com/facebookresearch/fairseq.git@31fba013a070eaff69dec8642e68e7134d60ab0f
+fastapi==0.98.0
+filelock==3.12.2
+flake8==6.0.0
+fsspec==2023.6.0
+h11==0.14.0
+httpcore==0.17.2
+httptools==0.1.2
+httpx==0.24.1
+huggingface-hub==0.15.1
+hydra-core==1.0.7
+HyperPyYAML==1.2.1
+identify==2.5.24
+idna==3.4
+importlib-resources==5.12.0
+iniconfig==2.0.0
+isort==5.12.0
+jieba==0.42.1
+Jinja2==3.1.2
+joblib==1.3.1
+lazy-object-proxy==1.9.0
+lit==16.0.6
+lxml==4.9.2
+MarkupSafe==2.1.3
+mccabe==0.7.0
+mock==5.0.2
+mpmath==1.3.0
+networkx==3.1
+nodeenv==1.8.0
+numpy==1.24.4
+nvidia-cublas-cu11==11.10.3.66
+nvidia-cuda-cupti-cu11==11.7.101
+nvidia-cuda-nvrtc-cu11==11.7.99
+nvidia-cuda-runtime-cu11==11.7.99
+nvidia-cudnn-cu11==8.5.0.96
+nvidia-cufft-cu11==10.9.0.58
+nvidia-curand-cu11==10.2.10.91
+nvidia-cusolver-cu11==11.4.0.1
+nvidia-cusparse-cu11==11.7.4.91
+nvidia-nccl-cu11==2.14.3
+nvidia-nvtx-cu11==11.7.91
+omegaconf==2.0.6
+packaging==23.1
+platformdirs==3.8.0
+pluggy==1.2.0
+portalocker==2.7.0
+pre-commit==3.3.3
+pycodestyle==2.10.0
+pycparser==2.21
+pydantic==1.10.9
+pyflakes==3.0.1
+pylint==2.17.4
+pytest==7.4.0
+python-multipart==0.0.6
+PyYAML==6.0
+regex==2023.6.3
+requests==2.31.0
+ruamel.yaml==0.17.28
+ruamel.yaml.clib==0.2.7
+sacrebleu==2.3.1
+scikit-learn==1.3.0
+scipy==1.10.1
+sentencepiece==0.1.99
+sniffio==1.3.0
+speechbrain==0.5.14
+starlette==0.27.0
+subword-nmt==0.3.8
+sympy==1.12
+tabulate==0.9.0
+threadpoolctl==3.1.0
+tomli==2.0.1
+tomlkit==0.11.8
+torch==2.0.1
+torchaudio==2.0.2
+tqdm==4.65.0
+triton==2.0.0
+typing_extensions==4.7.1
+urllib3==2.0.3
+uvicorn==0.11.3
+uvloop==0.17.0
+virtualenv==20.23.1
+websockets==8.1
+wrapt==1.15.0
+zipp==3.15.0
+
+
+2023-07-05 14:57:02,231 - speechbrain.utils.parameter_transfer - DEBUG - Collecting files (or symlinks) for pretraining in results/asr_transformer_seg_char/save.
+2023-07-05 14:57:02,231 - speechbrain.pretrained.fetching - INFO - Fetch model.ckpt: Delegating to Huggingface hub, source results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-55-37+00.
+2023-07-05 14:57:02,290 - speechbrain.core - ERROR - Exception:
+Traceback (most recent call last):
+  File "test.py", line 308, in <module>
+    run_on_main(hparams["pretrainer"].collect_files)
+  File "/home/sinica_axel/miniconda3/envs/asr_api/lib/python3.8/site-packages/speechbrain/utils/distributed.py", line 61, in run_on_main
+    func(*args, **kwargs)
+  File "/home/sinica_axel/miniconda3/envs/asr_api/lib/python3.8/site-packages/speechbrain/utils/parameter_transfer.py", line 202, in collect_files
+    path = fetch(
+  File "/home/sinica_axel/miniconda3/envs/asr_api/lib/python3.8/site-packages/speechbrain/pretrained/fetching.py", line 120, in fetch
+    fetched_file = huggingface_hub.hf_hub_download(
+  File "/home/sinica_axel/miniconda3/envs/asr_api/lib/python3.8/site-packages/huggingface_hub/utils/_validators.py", line 110, in _inner_fn
+    validate_repo_id(arg_value)
+  File "/home/sinica_axel/miniconda3/envs/asr_api/lib/python3.8/site-packages/huggingface_hub/utils/_validators.py", line 158, in validate_repo_id
+    raise HFValidationError(
+huggingface_hub.utils._validators.HFValidationError: Repo id must be in the form 'repo_name' or 'namespace/repo_name': 'results/TransformerLM_seg_char/save/CKPT+2021-10-05+20-55-37+00'. Use `repo_type` argument if needed.

ASR-model/asr_transformer_seg_char_ctc0.3/results/asr_transformer_seg_char/test.py

@@ -0,0 +1,331 @@
+import sys
+
+import torch
+import speechbrain as sb
+from speechbrain.utils.distributed import run_on_main
+from hyperpyyaml import load_hyperpyyaml
+
+
+class ASR(sb.core.Brain):
+    def compute_forward(self, batch, stage):
+        batch = batch.to(self.device)
+        wavs, wavs_len = batch.sig
+        tokens_bos, _ = batch.tokens_bos
+
+        feats = self.hparams.compute_features(wavs)
+        current_epoch = self.hparams.epoch_counter.current
+        feats = self.hparams.normalize(feats, wavs_len, epoch=current_epoch)
+
+        # if stage == sb.Stage.TRAIN:
+        #     if hasattr(self.modules, "augmentation"):
+        #         feats = self.hparams.augmentation(feats)
+
+        src = self.modules.CNN(feats)
+        enc_out, pred = self.modules.Transformer(
+            src, tokens_bos, wavs_len, pad_idx=self.hparams.pad_index
+        )
+
+        logits = self.modules.ctc_lin(enc_out)
+        p_ctc = self.hparams.log_softmax(logits)
+
+        pred = self.hparams.seq_lin(pred)
+        p_seq = self.hparams.log_softmax(pred)
+
+        hyps = None
+        if stage == sb.Stage.TRAIN:
+            hyps = None
+        elif stage == sb.Stage.VALID:
+            hyps = None
+            current_epoch = self.hparams.epoch_counter.current
+            if current_epoch % self.hparams.valid_search_interval == 0:
+                # for the sake of efficiency, we only perform beamsearch with limited capacity
+                # and no LM to give user some idea of how the AM is doing
+                hyps, _ = self.hparams.valid_search(enc_out.detach(), wavs_len)
+        elif stage == sb.Stage.TEST:
+            hyps, _ = self.hparams.test_search(enc_out.detach(), wavs_len)
+
+        return p_ctc, p_seq, wavs_len, hyps
+
+    def compute_objectives(self, predictions, batch, stage):
+
+        (p_ctc, p_seq, wavs_len, hyps,) = predictions
+
+        ids = batch.id
+        tokens_eos, tokens_eos_len = batch.tokens_eos
+        tokens, tokens_len = batch.tokens
+
+        attention_loss = self.hparams.seq_cost(
+            p_seq, tokens_eos, tokens_eos_len
+        )
+        ctc_loss = self.hparams.ctc_cost(p_ctc, tokens, wavs_len, tokens_len)
+        loss = (
+            self.hparams.ctc_weight * ctc_loss
+            + (1 - self.hparams.ctc_weight) * attention_loss
+        )
+
+        if stage != sb.Stage.TRAIN:
+            current_epoch = self.hparams.epoch_counter.current
+            valid_search_interval = self.hparams.valid_search_interval
+
+            if current_epoch % valid_search_interval == 0 or (
+                stage == sb.Stage.TEST
+            ):
+                predictions = [
+                    self.hparams.tokenizer.decode_ids(utt_seq).split(" ")
+                    for utt_seq in hyps
+                ]
+                targets = [
+                    transcription.split(" ")
+                    for transcription in batch.transcription
+                ]
+                if self.hparams.remove_spaces:
+                    predictions = [
+                        "".join(prediction_words)
+                        for prediction_words in predictions
+                    ]
+                    targets = [
+                        "".join(target_words) for target_words in targets
+                    ]
+                    self.cer_metric.append(ids, predictions, targets)
+
+            self.acc_metric.append(p_seq, tokens_eos, tokens_eos_len)
+
+        return loss
+
+    def fit_batch(self, batch):
+        self.check_and_reset_optimizer()
+
+        predictions = self.compute_forward(batch, sb.Stage.TRAIN)
+        loss = self.compute_objectives(predictions, batch, sb.Stage.TRAIN)
+
+        (loss / self.hparams.gradient_accumulation).backward()
+
+        if self.step % self.hparams.gradient_accumulation == 0:
+            self.check_gradients(loss)
+
+            self.optimizer.step()
+            self.optimizer.zero_grad()
+
+            self.hparams.noam_annealing(self.optimizer)
+
+        return loss.detach()
+
+    def evaluate_batch(self, batch, stage):
+        with torch.no_grad():
+            predictions = self.compute_forward(batch, stage=stage)
+            loss = self.compute_objectives(predictions, batch, stage=stage)
+        # origin function is call loss.detach().cpu()
+        return loss.detach()
+
+    def on_stage_start(self, stage, epoch):
+        if stage != sb.Stage.TRAIN:
+            self.acc_metric = self.hparams.acc_computer()
+            self.cer_metric = self.hparams.cer_computer()
+
+    def on_stage_end(self, stage, stage_loss, epoch):
+        stage_stats = {"loss": stage_loss}
+        if stage == sb.Stage.TRAIN:
+            self.train_stats = stage_stats
+        else:
+            stage_stats["ACC"] = self.acc_metric.summarize()
+            current_epoch = self.hparams.epoch_counter.current
+            valid_search_interval = self.hparams.valid_search_interval
+            if (
+                current_epoch % valid_search_interval == 0
+                or stage == sb.Stage.TEST
+            ):
+                stage_stats["CER"] = self.cer_metric.summarize("error_rate")
+
+        if stage == sb.Stage.VALID and sb.utils.distributed.if_main_process():
+
+            current_epoch = self.hparams.epoch_counter.current
+            if current_epoch <= self.hparams.stage_one_epochs:
+                lr = self.hparams.noam_annealing.current_lr
+                steps = self.hparams.noam_annealing.n_steps
+                optimizer = self.optimizer.__class__.__name__
+            else:
+                lr = self.hparams.lr_sgd
+                steps = -1
+                optimizer = self.optimizer.__class__.__name__
+
+            epoch_stats = {
+                "epoch": epoch,
+                "lr": lr,
+                "steps": steps,
+                "optimizer": optimizer,
+            }
+            self.hparams.train_logger.log_stats(
+                stats_meta=epoch_stats,
+                train_stats=self.train_stats,
+                valid_stats=stage_stats,
+            )
+            self.checkpointer.save_and_keep_only(
+                meta={"ACC": stage_stats["ACC"], "epoch": epoch},
+                max_keys=["ACC"],
+                num_to_keep=10,
+            )
+
+        elif stage == sb.Stage.TEST:
+            self.hparams.train_logger.log_stats(
+                stats_meta={"Epoch loaded": self.hparams.epoch_counter.current},
+                test_stats=stage_stats,
+            )
+            with open(self.hparams.cer_file, "w") as cer_file:
+                self.cer_metric.write_stats(cer_file)
+
+            self.checkpointer.save_and_keep_only(
+                meta={"ACC": 1.1, "epoch": epoch},
+                max_keys=["ACC"],
+                num_to_keep=1,
+            )
+
+    def check_and_reset_optimizer(self):
+        current_epoch = self.hparams.epoch_counter.current
+        if not hasattr(self, "switched"):
+            self.switched = False
+            if isinstance(self.optimizer, torch.optim.SGD):
+                self.switched = True
+
+        if self.switched is True:
+            return
+
+        if current_epoch > self.hparams.stage_one_epochs:
+            self.optimizer = self.hparams.SGD(self.modules.parameters())
+
+            if self.checkpointer is not None:
+                self.checkpointer.add_recoverable("optimizer", self.optimizer)
+
+            self.switched = True
+
+    def on_fit_start(self):
+        """Initialize the right optimizer on the training start"""
+        super().on_fit_start()
+
+        current_epoch = self.hparams.epoch_counter.current
+        current_optimizer = self.optimizer
+        if current_epoch > self.hparams.stage_one_epochs:
+            del self.optimizer
+            self.optimizer = self.hparams.SGD(self.modules.parameters())
+
+            if self.checkpointer is not None:
+                group = current_optimizer.param_groups[0]
+                if "momentum" not in group:
+                    return
+                self.checkpointer.recover_if_possible(
+                    device=torch.device(self.device)
+                )
+
+    def on_evaluate_start(self, max_key=None, min_key=None):
+        super().on_evaluate_start()
+
+        checkpointer = self.checkpointer.find_checkpoints(
+            max_key=max_key, min_key=min_key
+        )
+        checkpointer = sb.utils.checkpoints.average_checkpoints(
+            checkpointer, recoverable_name="model", device=self.device
+        )
+
+        self.hparams.model.load_state_dict(checkpointer, strict=True)
+        self.hparams.model.eval()
+
+
+def dataio_prepare(hparams):
+    @sb.utils.data_pipeline.takes("transcription")
+    @sb.utils.data_pipeline.provides(
+        "transcription", "tokens_bos", "tokens_eos", "tokens"
+    )
+    def transcription_pipline(transcription):
+        yield transcription
+        tokens_list = hparams["tokenizer"].encode_as_ids(transcription)
+        tokens_bos = torch.LongTensor([hparams["bos_index"]] + (tokens_list))
+        yield tokens_bos
+        tokens_eos = torch.LongTensor(tokens_list + [hparams["eos_index"]])
+        yield tokens_eos
+        tokens = torch.LongTensor(tokens_list)
+        yield tokens
+
+    @sb.utils.data_pipeline.takes("wav")
+    @sb.utils.data_pipeline.provides("sig")
+    def audio_pipline(wav):
+        sig = sb.dataio.dataio.read_audio(wav)
+        return sig
+
+    @sb.utils.data_pipeline.takes("wav")
+    @sb.utils.data_pipeline.provides("sig")
+    def sp_audio_pipline(wav):
+        sig = sb.dataio.dataio.read_audio(wav)
+        sig = sig.unsqueeze(0)
+        sig = hparams["speed_perturb"](sig)
+        sig = sig.squeeze(0)
+        return sig
+
+    datasets = {}
+    data_folder = hparams["data_folder"]
+    output_keys = [
+        "transcription",
+        "tokens_bos",
+        "tokens_eos",
+        "tokens",
+        "sig",
+        "id",
+    ]
+    default_dynamic_items = [transcription_pipline, audio_pipline]
+    train_dynamic_item = [transcription_pipline, sp_audio_pipline]
+
+    for dataset_name in ["train", "dev", "test"]:
+        if dataset_name == "train":
+            dynamic_items = train_dynamic_item
+        else:
+            dynamic_items = default_dynamic_items
+
+        json_path = f"{data_folder}/{dataset_name}.json"
+        datasets[dataset_name] = sb.dataio.dataset.DynamicItemDataset.from_json(
+            json_path=json_path,
+            replacements={"data_root": data_folder},
+            dynamic_items=dynamic_items,
+            output_keys=output_keys,
+        )
+
+    return datasets
+
+
+if __name__ == "__main__":
+    #hparams_file_path, run_opts, overrides = sb.parse_arguments(sys.argv[1:])
+    hparams_file_path = "hyperparams.yaml"
+    run_opts = {"device": "cuda", "distributed_launch": False}
+    overrides = None
+    with open(hparams_file_path) as hparams_file:
+        hparams = load_hyperpyyaml(hparams_file, overrides)
+
+    sb.utils.distributed.ddp_init_group(run_opts)
+
+    sb.create_experiment_directory(
+        experiment_directory=hparams["output_folder"],
+        hyperparams_to_save=hparams_file_path,
+        overrides=overrides,
+    )
+
+    run_on_main(hparams["pretrainer"].collect_files)
+    hparams["pretrainer"].load_collected(device=run_opts["device"])
+
+    datasets = dataio_prepare(hparams)
+
+    asr_brain = ASR(
+        modules=hparams["modules"],
+        opt_class=hparams["Adam"],
+        hparams=hparams,
+        run_opts=run_opts,
+        checkpointer=hparams["checkpointer"],
+    )
+
+    # asr_brain.fit(
+    #     asr_brain.hparams.epoch_counter,
+    #     datasets["train"],
+    #     datasets["dev"],
+    #     train_loader_kwargs=hparams["train_dataloader_opts"],
+    #     valid_loader_kwargs=hparams["valid_dataloader_opts"],
+    # )
+
+    asr_brain.evaluate(
+        datasets["test"], max_key="ACC", test_loader_kwargs=hparams["test_dataloader_opts"]
+    )

ASR-model/asr_transformer_seg_char_ctc0.3/save/CKPT+2021-10-11+14-03-15+00/CKPT.yaml

@@ -0,0 +1,5 @@
+# yamllint disable
+ACC: 1.1
+end-of-epoch: true
+epoch: null
+unixtime: 1633932195.1636646

ASR-model/asr_transformer_seg_char_ctc0.3/save/CKPT+2021-10-11+14-03-15+00/brain.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22e854e25aa2ad065885cf277f6017bfa7ab555eb84131dfa2da35605cb8fb14
+size 32

ASR-model/asr_transformer_seg_char_ctc0.3/save/CKPT+2021-10-11+14-03-15+00/counter.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:39fa9ec190eee7b6f4dff1100d6343e10918d044c75eac8f9e9a2596173f80c9
+size 2

ASR-model/asr_transformer_seg_char_ctc0.3/save/CKPT+2021-10-11+14-03-15+00/model.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f325dfc2ecffc1332882e84cac08a9f05e500253413c2346a20f35baaeed705
+size 126713830

ASR-model/asr_transformer_seg_char_ctc0.3/save/CKPT+2021-10-11+14-03-15+00/noam_scheduler.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35ce303ccbe8d1abb2b87a01a29e01ed6a7b345253ab6db9111cbd5eefbf503c
+size 431

ASR-model/asr_transformer_seg_char_ctc0.3/save/CKPT+2021-10-11+14-03-15+00/normalizer.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c4aa56041d2ead93019a7781f786d29cc577aaf24f4766574c47867a2078f80
+size 1782

ASR-model/asr_transformer_seg_char_ctc0.3/test.py

@@ -0,0 +1,181 @@
+import sys
+
+import torch
+import speechbrain as sb
+from speechbrain.utils.distributed import run_on_main
+from hyperpyyaml import load_hyperpyyaml
+
+
+class ASR(sb.core.Brain):
+    def compute_forward(self, batch, stage):
+        batch = batch.to(self.device)
+        wavs, wavs_len = batch.sig
+        tokens_bos, _ = batch.tokens_bos
+
+        feats = self.hparams.compute_features(wavs)
+        current_epoch = self.hparams.epoch_counter.current
+        feats = self.hparams.normalize(feats, wavs_len, epoch=current_epoch)
+
+        # if stage == sb.Stage.TRAIN:
+        #     if hasattr(self.modules, "augmentation"):
+        #         feats = self.hparams.augmentation(feats)
+
+        src = self.modules.CNN(feats)
+        enc_out, pred = self.modules.Transformer(
+            src, tokens_bos, wavs_len, pad_idx=self.hparams.pad_index
+        )
+
+        logits = self.modules.ctc_lin(enc_out)
+        p_ctc = self.hparams.log_softmax(logits)
+
+        pred = self.hparams.seq_lin(pred)
+        p_seq = self.hparams.log_softmax(pred)
+
+        hyps = None
+        if stage == sb.Stage.TRAIN:
+            hyps = None
+        elif stage == sb.Stage.VALID:
+            hyps = None
+            current_epoch = self.hparams.epoch_counter.current
+            if current_epoch % self.hparams.valid_search_interval == 0:
+                # for the sake of efficiency, we only perform beamsearch with limited capacity
+                # and no LM to give user some idea of how the AM is doing
+                hyps, _ = self.hparams.valid_search(enc_out.detach(), wavs_len)
+        elif stage == sb.Stage.TEST:
+            hyps, _ = self.hparams.test_search(enc_out.detach(), wavs_len)
+
+        return p_ctc, p_seq, wavs_len, hyps
+
+    def evaluate_batch(self, batch, stage):
+        with torch.no_grad():
+            predictions = self.compute_forward(batch, stage=stage)
+            loss = self.compute_objectives(predictions, batch, stage=stage)
+        # origin function is call loss.detach().cpu()
+        return loss.detach()
+
+    def on_stage_start(self, stage, epoch):
+        if stage != sb.Stage.TRAIN:
+            self.acc_metric = self.hparams.acc_computer()
+            self.cer_metric = self.hparams.cer_computer()
+
+    def on_stage_end(self, stage, stage_loss, epoch):
+        stage_stats = {"loss": stage_loss}
+        if stage == sb.Stage.TRAIN:
+            self.train_stats = stage_stats
+        else:
+            stage_stats["ACC"] = self.acc_metric.summarize()
+            current_epoch = self.hparams.epoch_counter.current
+            valid_search_interval = self.hparams.valid_search_interval
+            if (
+                current_epoch % valid_search_interval == 0
+                or stage == sb.Stage.TEST
+            ):
+                stage_stats["CER"] = self.cer_metric.summarize("error_rate")
+
+        if stage == sb.Stage.VALID and sb.utils.distributed.if_main_process():
+
+            current_epoch = self.hparams.epoch_counter.current
+            if current_epoch <= self.hparams.stage_one_epochs:
+                lr = self.hparams.noam_annealing.current_lr
+                steps = self.hparams.noam_annealing.n_steps
+                optimizer = self.optimizer.__class__.__name__
+            else:
+                lr = self.hparams.lr_sgd
+                steps = -1
+                optimizer = self.optimizer.__class__.__name__
+
+            epoch_stats = {
+                "epoch": epoch,
+                "lr": lr,
+                "steps": steps,
+                "optimizer": optimizer,
+            }
+            self.hparams.train_logger.log_stats(
+                stats_meta=epoch_stats,
+                train_stats=self.train_stats,
+                valid_stats=stage_stats,
+            )
+            self.checkpointer.save_and_keep_only(
+                meta={"ACC": stage_stats["ACC"], "epoch": epoch},
+                max_keys=["ACC"],
+                num_to_keep=10,
+            )
+
+        elif stage == sb.Stage.TEST:
+            self.hparams.train_logger.log_stats(
+                stats_meta={"Epoch loaded": self.hparams.epoch_counter.current},
+                test_stats=stage_stats,
+            )
+            with open(self.hparams.cer_file, "w") as cer_file:
+                self.cer_metric.write_stats(cer_file)
+
+            self.checkpointer.save_and_keep_only(
+                meta={"ACC": 1.1, "epoch": epoch},
+                max_keys=["ACC"],
+                num_to_keep=1,
+            )
+
+    def check_and_reset_optimizer(self):
+        current_epoch = self.hparams.epoch_counter.current
+        if not hasattr(self, "switched"):
+            self.switched = False
+            if isinstance(self.optimizer, torch.optim.SGD):
+                self.switched = True
+
+        if self.switched is True:
+            return
+
+        if current_epoch > self.hparams.stage_one_epochs:
+            self.optimizer = self.hparams.SGD(self.modules.parameters())
+
+            if self.checkpointer is not None:
+                self.checkpointer.add_recoverable("optimizer", self.optimizer)
+
+            self.switched = True
+
+    def on_fit_start(self):
+        """Initialize the right optimizer on the training start"""
+        super().on_fit_start()
+
+        current_epoch = self.hparams.epoch_counter.current
+        current_optimizer = self.optimizer
+        if current_epoch > self.hparams.stage_one_epochs:
+            del self.optimizer
+            self.optimizer = self.hparams.SGD(self.modules.parameters())
+
+            if self.checkpointer is not None:
+                group = current_optimizer.param_groups[0]
+                if "momentum" not in group:
+                    return
+                self.checkpointer.recover_if_possible(
+                    device=torch.device(self.device)
+                )
+
+    def on_evaluate_start(self, max_key=None, min_key=None):
+        super().on_evaluate_start()
+
+        checkpointer = self.checkpointer.find_checkpoints(
+            max_key=max_key, min_key=min_key
+        )
+        checkpointer = sb.utils.checkpoints.average_checkpoints(
+            checkpointer, recoverable_name="model", device=self.device
+        )
+
+        self.hparams.model.load_state_dict(checkpointer, strict=True)
+        self.hparams.model.eval()
+
+
+if __name__ == "__main__":
+    hparams_file_path = "hyperparams.yaml"
+    run_opts = {"device": "cuda", "distributed_launch": False}
+    with open(hparams_file_path) as hparams_file:
+        hparams = load_hyperpyyaml(hparams_file)
+
+    asr_brain = ASR(
+        modules=hparams["modules"],
+        opt_class=hparams["Adam"],
+        hparams=hparams,
+        run_opts=run_opts,
+        checkpointer=hparams["checkpointer"],
+    )
+

ASR-model/asr_transformer_seg_char_ctc0.3/train.py

@@ -0,0 +1,322 @@
+import sys
+
+import torch
+import speechbrain as sb
+from speechbrain.utils.distributed import run_on_main
+from hyperpyyaml import load_hyperpyyaml
+
+
+class ASR(sb.core.Brain):
+    def compute_forward(self, batch, stage):
+        batch = batch.to(self.device)
+        wavs, wavs_len = batch.sig
+        tokens_bos, _ = batch.tokens_bos
+        feats = self.hparams.compute_features(wavs)
+        current_epoch = self.hparams.epoch_counter.current
+        feats = self.modules.normalize(feats, wavs_len, epoch=current_epoch)
+
+        src = self.modules.CNN(feats)
+        enc_out, pred = self.modules.Transformer(
+            src, tokens_bos, wavs_len, pad_idx=self.hparams.pad_index
+        )
+
+        logits = self.modules.ctc_lin(enc_out)
+        p_ctc = self.hparams.log_softmax(logits)
+
+        pred = self.modules.seq_lin(pred)
+        p_seq = self.hparams.log_softmax(pred)
+
+        hyps = None
+        if stage == sb.Stage.TRAIN:
+            hyps = None
+        elif stage == sb.Stage.VALID:
+            hyps = None
+            current_epoch = self.hparams.epoch_counter.current
+            if current_epoch % self.hparams.valid_search_interval == 0:
+                hyps, _ = self.hparams.valid_search(enc_out.detach(), wavs_len)
+        elif stage == sb.Stage.TEST:
+            hyps, _ = self.hparams.test_search(enc_out.detach(), wavs_len)
+
+        return p_ctc, p_seq, wavs_len, hyps
+
+    def compute_objectives(self, predictions, batch, stage):
+
+        (p_ctc, p_seq, wavs_len, hyps,) = predictions
+
+        ids = batch.id
+        tokens_eos, tokens_eos_len = batch.tokens_eos
+        tokens, tokens_len = batch.tokens
+
+        attention_loss = self.hparams.seq_cost(
+            p_seq, tokens_eos, length=tokens_eos_len
+        )
+        ctc_loss = self.hparams.ctc_cost(p_ctc, tokens, wavs_len, tokens_len)
+        loss = (
+            self.hparams.ctc_weight * ctc_loss
+            + (1 - self.hparams.ctc_weight) * attention_loss
+        )
+
+        if stage != sb.Stage.TRAIN:
+            current_epoch = self.hparams.epoch_counter.current
+            valid_search_interval = self.hparams.valid_search_interval
+
+            if current_epoch % valid_search_interval == 0 or (
+                stage == sb.Stage.TEST
+            ):
+                predictions = [
+                    hparams["tokenizer"].decode_ids(utt_seq).split(" ")
+                    for utt_seq in hyps
+                ]
+                targets = [
+                    transcription.split(" ")
+                    for transcription in batch.transcription
+                ]
+                if self.hparams.remove_spaces:
+                    predictions = [
+                        "".join(prediction_words)
+                        for prediction_words in predictions
+                    ]
+                    targets = [
+                        "".join(target_words) for target_words in targets
+                    ]
+                    self.cer_metric.append(ids, predictions, targets)
+
+            self.acc_metric.append(p_seq, tokens_eos, tokens_eos_len)
+
+        return loss
+
+    def fit_batch(self, batch):
+        self.check_and_reset_optimizer()
+
+        predictions = self.compute_forward(batch, sb.Stage.TRAIN)
+        loss = self.compute_objectives(predictions, batch, sb.Stage.TRAIN)
+
+        (loss / self.hparams.gradient_accumulation).backward()
+
+        if self.step % self.hparams.gradient_accumulation == 0:
+            self.check_gradients(loss)
+
+            self.optimizer.step()
+            self.optimizer.zero_grad()
+
+            self.hparams.noam_annealing(self.optimizer)
+
+        return loss.detach()
+
+    def evaluate_batch(self, batch, stage):
+        with torch.no_grad():
+            predictions = self.compute_forward(batch, stage=stage)
+            loss = self.compute_objectives(predictions, batch, stage=stage)
+        # origin function is call loss.detach().cpu()
+        return loss.detach()
+
+    def on_stage_start(self, stage, epoch):
+        if stage != sb.Stage.TRAIN:
+            self.acc_metric = self.hparams.acc_computer()
+            self.cer_metric = self.hparams.cer_computer()
+
+    def on_stage_end(self, stage, stage_loss, epoch):
+        stage_stats = {"loss": stage_loss}
+        if stage == sb.Stage.TRAIN:
+            self.train_stats = stage_stats
+        else:
+            stage_stats["ACC"] = self.acc_metric.summarize()
+            current_epoch = self.hparams.epoch_counter.current
+            valid_search_interval = self.hparams.valid_search_interval
+            if (
+                current_epoch % valid_search_interval == 0
+                or stage == sb.Stage.TEST
+            ):
+                stage_stats["CER"] = self.cer_metric.summarize("error_rate")
+
+        if stage == sb.Stage.VALID and sb.utils.distributed.if_main_process():
+
+            current_epoch = self.hparams.epoch_counter.current
+            if current_epoch <= self.hparams.stage_one_epochs:
+                lr = self.hparams.noam_annealing.current_lr
+                steps = self.hparams.noam_annealing.n_steps
+                optimizer = self.optimizer.__class__.__name__
+            else:
+                lr = self.hparams.lr_sgd
+                steps = -1
+                optimizer = self.optimizer.__class__.__name__
+
+            epoch_stats = {
+                "epoch": epoch,
+                "lr": lr,
+                "steps": steps,
+                "optimizer": optimizer,
+            }
+            self.hparams.train_logger.log_stats(
+                stats_meta=epoch_stats,
+                train_stats=self.train_stats,
+                valid_stats=stage_stats,
+            )
+            self.checkpointer.save_and_keep_only(
+                meta={"ACC": stage_stats["ACC"], "epoch": epoch},
+                max_keys=["ACC"],
+                num_to_keep=10,
+            )
+
+        elif stage == sb.Stage.TEST:
+            self.hparams.train_logger.log_stats(
+                stats_meta={"Epoch loaded": self.hparams.epoch_counter.current},
+                test_stats=stage_stats,
+            )
+            with open(self.hparams.cer_file, "w") as cer_file:
+                self.cer_metric.write_stats(cer_file)
+
+            self.checkpointer.save_and_keep_only(
+                meta={"ACC": 1.1, "epoch": epoch},
+                max_keys=["ACC"],
+                num_to_keep=1,
+            )
+
+    def check_and_reset_optimizer(self):
+        current_epoch = self.hparams.epoch_counter.current
+        if not hasattr(self, "switched"):
+            self.switched = False
+            if isinstance(self.optimizer, torch.optim.SGD):
+                self.switched = True
+
+        if self.switched is True:
+            return
+
+        if current_epoch > self.hparams.stage_one_epochs:
+            self.optimizer = self.hparams.SGD(self.modules.parameters())
+
+            if self.checkpointer is not None:
+                self.checkpointer.add_recoverable("optimizer", self.optimizer)
+
+            self.switched = True
+
+    def on_fit_start(self):
+        """Initialize the right optimizer on the training start"""
+        super().on_fit_start()
+
+        current_epoch = self.hparams.epoch_counter.current
+        current_optimizer = self.optimizer
+        if current_epoch > self.hparams.stage_one_epochs:
+            del self.optimizer
+            self.optimizer = self.hparams.SGD(self.modules.parameters())
+
+            if self.checkpointer is not None:
+                group = current_optimizer.param_groups[0]
+                if "momentum" not in group:
+                    return
+                self.checkpointer.recover_if_possible(
+                    device=torch.device(self.device)
+                )
+
+    def on_evaluate_start(self, max_key=None, min_key=None):
+        super().on_evaluate_start()
+
+        checkpointers = self.checkpointer.find_checkpoints(
+            max_key=max_key, min_key=min_key
+        )
+        checkpointer = sb.utils.checkpoints.average_checkpoints(
+            checkpointers, recoverable_name="model", device=self.device
+        )
+
+        self.hparams.model.load_state_dict(checkpointer, strict=True)
+        self.hparams.model.eval()
+
+
+def dataio_prepare(hparams):
+    @sb.utils.data_pipeline.takes("transcription")
+    @sb.utils.data_pipeline.provides(
+        "transcription", "tokens_bos", "tokens_eos", "tokens"
+    )
+    def transcription_pipline(transcription):
+        yield transcription
+        tokens_list = hparams["tokenizer"].encode_as_ids(transcription)
+        tokens_bos = torch.LongTensor([hparams["bos_index"]] + (tokens_list))
+        yield tokens_bos
+        tokens_eos = torch.LongTensor(tokens_list + [hparams["eos_index"]])
+        yield tokens_eos
+        tokens = torch.LongTensor(tokens_list)
+        yield tokens
+
+    @sb.utils.data_pipeline.takes("wav")
+    @sb.utils.data_pipeline.provides("sig")
+    def audio_pipline(wav):
+        sig = sb.dataio.dataio.read_audio(wav)
+        return sig
+
+    @sb.utils.data_pipeline.takes("wav")
+    @sb.utils.data_pipeline.provides("sig")
+    def sp_audio_pipline(wav):
+        sig = sb.dataio.dataio.read_audio(wav)
+        sig = sig.unsqueeze(0)
+        sig = hparams["speed_perturb"](sig)
+        sig = sig.squeeze(0)
+        return sig
+
+    datasets = {}
+    data_folder = hparams["data_folder"]
+    output_keys = [
+        "transcription",
+        "tokens_bos",
+        "tokens_eos",
+        "tokens",
+        "sig",
+        "id",
+    ]
+    default_dynamic_items = [transcription_pipline, audio_pipline]
+    train_dynamic_item = [transcription_pipline, sp_audio_pipline]
+
+    for dataset_name in ["train", "dev", "test"]:
+        if dataset_name == "train":
+            dynamic_items = train_dynamic_item
+        else:
+            dynamic_items = default_dynamic_items
+
+        json_path = f"{data_folder}/{dataset_name}.json"
+        datasets[dataset_name] = sb.dataio.dataset.DynamicItemDataset.from_json(
+            json_path=json_path,
+            replacements={"data_root": data_folder},
+            dynamic_items=dynamic_items,
+            output_keys=output_keys,
+        )
+
+    return datasets
+
+
+if __name__ == "__main__":
+    hparams_file_path, run_opts, overrides = sb.parse_arguments(sys.argv[1:])
+
+    sb.utils.distributed.ddp_init_group(run_opts)
+
+    with open(hparams_file_path) as hparams_file:
+        hparams = load_hyperpyyaml(hparams_file, overrides)
+
+    sb.create_experiment_directory(
+        experiment_directory=hparams["output_folder"],
+        hyperparams_to_save=hparams_file_path,
+        overrides=overrides,
+    )
+
+    run_on_main(hparams["pretrainer"].collect_files)
+    hparams["pretrainer"].load_collected(device=run_opts["device"])
+
+    datasets = dataio_prepare(hparams)
+
+    asr_brain = ASR(
+        modules=hparams["modules"],
+        opt_class=hparams["Adam"],
+        hparams=hparams,
+        run_opts=run_opts,
+        checkpointer=hparams["checkpointer"],
+    )
+
+    asr_brain.fit(
+        asr_brain.hparams.epoch_counter,
+        datasets["train"],
+        datasets["dev"],
+        train_loader_kwargs=hparams["train_dataloader_opts"],
+        valid_loader_kwargs=hparams["valid_dataloader_opts"],
+    )
+
+    # asr_brain.evaluate(
+    #     datasets["test"],max_key="ACC", test_loader_kwargs=hparams["test_dataloader_opts"]
+    # )

ASR-model/asr_transformer_seg_char_ctc0.3/train_log.txt

@@ -0,0 +1,70 @@
+epoch: 1, lr: 2.40e-05, steps: 1520, optimizer: Adam - train loss: 3.41e+02 - valid loss: 2.52e+02, valid ACC: 4.07e-01
+epoch: 2, lr: 4.81e-05, steps: 3040, optimizer: Adam - train loss: 2.36e+02 - valid loss: 2.26e+02, valid ACC: 4.36e-01
+epoch: 3, lr: 7.21e-05, steps: 4560, optimizer: Adam - train loss: 2.17e+02 - valid loss: 2.12e+02, valid ACC: 4.66e-01
+epoch: 4, lr: 9.61e-05, steps: 6080, optimizer: Adam - train loss: 2.05e+02 - valid loss: 2.02e+02, valid ACC: 4.86e-01
+epoch: 5, lr: 1.20e-04, steps: 7600, optimizer: Adam - train loss: 1.96e+02 - valid loss: 1.92e+02, valid ACC: 4.98e-01
+epoch: 6, lr: 1.44e-04, steps: 9120, optimizer: Adam - train loss: 1.70e+02 - valid loss: 1.51e+02, valid ACC: 5.30e-01
+epoch: 7, lr: 1.68e-04, steps: 10640, optimizer: Adam - train loss: 1.37e+02 - valid loss: 1.14e+02, valid ACC: 6.31e-01
+epoch: 8, lr: 1.92e-04, steps: 12160, optimizer: Adam - train loss: 1.10e+02 - valid loss: 88.56, valid ACC: 7.21e-01
+epoch: 9, lr: 2.16e-04, steps: 13680, optimizer: Adam - train loss: 86.93 - valid loss: 67.36, valid ACC: 8.03e-01
+epoch: 10, lr: 2.40e-04, steps: 15200, optimizer: Adam - train loss: 64.60 - valid loss: 51.05, valid ACC: 8.61e-01, valid CER: 22.14
+epoch: 11, lr: 2.64e-04, steps: 16720, optimizer: Adam - train loss: 51.94 - valid loss: 42.96, valid ACC: 8.86e-01
+epoch: 12, lr: 2.88e-04, steps: 18240, optimizer: Adam - train loss: 44.10 - valid loss: 37.56, valid ACC: 8.99e-01
+epoch: 13, lr: 3.12e-04, steps: 19760, optimizer: Adam - train loss: 38.66 - valid loss: 33.39, valid ACC: 9.09e-01
+epoch: 14, lr: 3.36e-04, steps: 21280, optimizer: Adam - train loss: 34.66 - valid loss: 31.04, valid ACC: 9.16e-01
+epoch: 15, lr: 3.60e-04, steps: 22800, optimizer: Adam - train loss: 31.69 - valid loss: 29.01, valid ACC: 9.19e-01
+epoch: 16, lr: 3.85e-04, steps: 24320, optimizer: Adam - train loss: 29.34 - valid loss: 27.39, valid ACC: 9.24e-01
+epoch: 17, lr: 3.89e-04, steps: 25840, optimizer: Adam - train loss: 27.37 - valid loss: 25.86, valid ACC: 9.28e-01
+epoch: 18, lr: 3.78e-04, steps: 27360, optimizer: Adam - train loss: 25.12 - valid loss: 24.43, valid ACC: 9.32e-01
+epoch: 19, lr: 3.68e-04, steps: 28880, optimizer: Adam - train loss: 23.16 - valid loss: 23.26, valid ACC: 9.35e-01
+epoch: 20, lr: 3.58e-04, steps: 30400, optimizer: Adam - train loss: 21.53 - valid loss: 22.58, valid ACC: 9.37e-01, valid CER: 10.20
+epoch: 21, lr: 3.50e-04, steps: 31920, optimizer: Adam - train loss: 20.06 - valid loss: 21.66, valid ACC: 9.39e-01
+epoch: 22, lr: 3.42e-04, steps: 33440, optimizer: Adam - train loss: 18.83 - valid loss: 21.18, valid ACC: 9.41e-01
+epoch: 23, lr: 3.34e-04, steps: 34960, optimizer: Adam - train loss: 17.77 - valid loss: 20.92, valid ACC: 9.42e-01
+epoch: 24, lr: 3.27e-04, steps: 36480, optimizer: Adam - train loss: 16.80 - valid loss: 20.30, valid ACC: 9.43e-01
+epoch: 25, lr: 3.21e-04, steps: 38000, optimizer: Adam - train loss: 15.91 - valid loss: 19.68, valid ACC: 9.44e-01
+epoch: 26, lr: 3.14e-04, steps: 39520, optimizer: Adam - train loss: 15.12 - valid loss: 19.73, valid ACC: 9.45e-01
+epoch: 27, lr: 3.09e-04, steps: 41040, optimizer: Adam - train loss: 14.41 - valid loss: 19.30, valid ACC: 9.46e-01
+epoch: 28, lr: 3.03e-04, steps: 42560, optimizer: Adam - train loss: 13.79 - valid loss: 18.82, valid ACC: 9.47e-01
+epoch: 29, lr: 2.98e-04, steps: 44080, optimizer: Adam - train loss: 13.15 - valid loss: 19.04, valid ACC: 9.47e-01
+epoch: 30, lr: 2.93e-04, steps: 45600, optimizer: Adam - train loss: 12.66 - valid loss: 18.89, valid ACC: 9.47e-01, valid CER: 8.39
+epoch: 31, lr: 2.88e-04, steps: 47120, optimizer: Adam - train loss: 12.13 - valid loss: 18.66, valid ACC: 9.48e-01
+epoch: 32, lr: 2.83e-04, steps: 48640, optimizer: Adam - train loss: 11.64 - valid loss: 18.51, valid ACC: 9.48e-01
+epoch: 33, lr: 2.79e-04, steps: 50160, optimizer: Adam - train loss: 11.21 - valid loss: 18.36, valid ACC: 9.48e-01
+epoch: 34, lr: 2.75e-04, steps: 51680, optimizer: Adam - train loss: 10.79 - valid loss: 18.65, valid ACC: 9.48e-01
+epoch: 35, lr: 2.71e-04, steps: 53200, optimizer: Adam - train loss: 10.43 - valid loss: 18.40, valid ACC: 9.49e-01
+epoch: 36, lr: 2.67e-04, steps: 54720, optimizer: Adam - train loss: 10.11 - valid loss: 18.29, valid ACC: 9.49e-01
+epoch: 37, lr: 2.64e-04, steps: 56240, optimizer: Adam - train loss: 9.78 - valid loss: 18.32, valid ACC: 9.50e-01
+epoch: 38, lr: 2.60e-04, steps: 57760, optimizer: Adam - train loss: 9.46 - valid loss: 18.12, valid ACC: 9.50e-01
+epoch: 39, lr: 2.57e-04, steps: 59280, optimizer: Adam - train loss: 9.17 - valid loss: 18.36, valid ACC: 9.49e-01
+epoch: 40, lr: 2.53e-04, steps: 60800, optimizer: Adam - train loss: 8.88 - valid loss: 18.29, valid ACC: 9.49e-01, valid CER: 7.93
+epoch: 41, lr: 2.50e-05, steps: -1, optimizer: SGD - train loss: 7.01 - valid loss: 17.34, valid ACC: 9.54e-01
+epoch: 42, lr: 2.50e-05, steps: -1, optimizer: SGD - train loss: 6.32 - valid loss: 17.22, valid ACC: 9.54e-01
+epoch: 43, lr: 2.50e-05, steps: -1, optimizer: SGD - train loss: 5.97 - valid loss: 17.15, valid ACC: 9.54e-01
+epoch: 44, lr: 2.50e-05, steps: -1, optimizer: SGD - train loss: 5.76 - valid loss: 17.14, valid ACC: 9.55e-01
+epoch: 45, lr: 2.50e-05, steps: -1, optimizer: SGD - train loss: 5.58 - valid loss: 17.19, valid ACC: 9.54e-01
+epoch: 46, lr: 2.50e-05, steps: -1, optimizer: SGD - train loss: 5.45 - valid loss: 17.07, valid ACC: 9.55e-01
+epoch: 47, lr: 2.50e-05, steps: -1, optimizer: SGD - train loss: 5.33 - valid loss: 17.07, valid ACC: 9.55e-01
+epoch: 48, lr: 2.50e-05, steps: -1, optimizer: SGD - train loss: 5.22 - valid loss: 17.17, valid ACC: 9.55e-01
+epoch: 49, lr: 2.50e-05, steps: -1, optimizer: SGD - train loss: 5.13 - valid loss: 17.24, valid ACC: 9.55e-01
+epoch: 50, lr: 2.50e-05, steps: -1, optimizer: SGD - train loss: 5.04 - valid loss: 17.10, valid ACC: 9.55e-01, valid CER: 7.10
+epoch: 51, lr: 2.50e-05, steps: -1, optimizer: SGD - train loss: 4.97 - valid loss: 17.26, valid ACC: 9.55e-01
+epoch: 52, lr: 2.50e-05, steps: -1, optimizer: SGD - train loss: 4.88 - valid loss: 17.23, valid ACC: 9.55e-01
+epoch: 53, lr: 2.50e-05, steps: -1, optimizer: SGD - train loss: 4.82 - valid loss: 17.17, valid ACC: 9.55e-01
+epoch: 54, lr: 2.50e-05, steps: -1, optimizer: SGD - train loss: 4.76 - valid loss: 17.31, valid ACC: 9.55e-01
+epoch: 55, lr: 2.50e-05, steps: -1, optimizer: SGD - train loss: 4.71 - valid loss: 17.37, valid ACC: 9.55e-01
+epoch: 56, lr: 2.50e-05, steps: -1, optimizer: SGD - train loss: 4.66 - valid loss: 17.26, valid ACC: 9.55e-01
+epoch: 57, lr: 2.50e-05, steps: -1, optimizer: SGD - train loss: 4.60 - valid loss: 17.43, valid ACC: 9.55e-01
+epoch: 58, lr: 2.50e-05, steps: -1, optimizer: SGD - train loss: 4.54 - valid loss: 17.52, valid ACC: 9.55e-01
+epoch: 59, lr: 2.50e-05, steps: -1, optimizer: SGD - train loss: 4.49 - valid loss: 17.42, valid ACC: 9.55e-01
+epoch: 60, lr: 2.50e-05, steps: -1, optimizer: SGD - train loss: 4.45 - valid loss: 17.53, valid ACC: 9.55e-01, valid CER: 7.11
+Epoch loaded: 60 - test loss: 30.17, test ACC: 9.27e-01, test CER: 7.50 - TransformerLM 0.2
+Epoch loaded: 60 - test loss: 15.08, test ACC: 9.27e-01, test CER: 7.74 - RNNLM 0.2
+Epoch loaded: 60 - test loss: 10.06, test ACC: 9.27e-01, test CER: 7.97 - w/o LM
+Epoch loaded: 60 - test loss: 7.54, test ACC: 9.27e-01, test CER: 7.62 - TransformerLM 0.1
+Epoch loaded: 60 - test loss: 6.03, test ACC: 9.27e-01, test CER: 7.54 - TransformerLM 0.3
+Epoch loaded: 60 - test loss: 5.03, test ACC: 9.27e-01, test CER: 7.65 - TransformerLM 0.4
+Epoch loaded: 60 - test loss: 4.31, test ACC: 9.27e-01, test CER: 7.87 - TransformerLM 0.5
+Epoch loaded: 60 - test loss: 3.77, test ACC: 9.27e-01, test CER: 8.13 - TransformerLM 0.6
+Epoch loaded: 60 - test loss: 3.35, test ACC: 9.27e-01, test CER: 8.37 - TransformerLM 0.7
+Epoch loaded: 60 - test loss: 3.02, test ACC: 9.27e-01, test CER: 8.74 - TransformerLM 0.8

ASR-model/tokenizer_seg_bpe5k_char/5000_char.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3598620f52b18d0378760db5f5b2443f87ea7e0887303c49a13b936ec66cbb7e
+size 288783

ASR-model/tokenizer_seg_bpe5k_char/5000_char.vocab

@@ -0,0 +1,4257 @@
+<unk>	0
+<s>	0
+</s>	0
+▁	-0.988007
+的	-3.64818
+是	-4.66353
+一	-4.71877
+這	-5.02165
+有	-5.0436
+不	-5.0719
+在	-5.10361
+個	-5.11818
+會	-5.35635
+來	-5.40839
+我	-5.45934
+以	-5.47765
+們	-5.52486
+人	-5.52624
+了	-5.57924
+到	-5.60484
+要	-5.64711
+為	-5.66809
+國	-5.66897
+就	-5.69856
+大	-5.71761
+臺	-5.78127
+也	-5.79987
+他	-5.80268
+年	-5.81117
+過	-5.86746
+十	-5.87511
+中	-5.88534
+天	-5.90132
+上	-5.90745
+公	-5.91384
+新	-5.92983
+可	-5.97377
+民	-6.02273
+出	-6.03295
+能	-6.03346
+生	-6.04135
+法	-6.04712
+政	-6.05771
+那	-6.07501
+經	-6.09248
+對	-6.11149
+多	-6.11851
+說	-6.16196
+後	-6.19182
+學	-6.20966
+家	-6.22306
+行	-6.23075
+成	-6.23928
+還	-6.25927
+時	-6.26309
+部	-6.26869
+都	-6.27577
+現	-6.3008
+今	-6.30895
+所	-6.33091
+之	-6.37008
+地	-6.37097
+發	-6.38745
+導	-6.40385
+報	-6.4079
+分	-6.40974
+方	-6.43329
+視	-6.4367
+三	-6.43784
+看	-6.45143
+因	-6.45181
+長	-6.45761
+前	-6.45974
+立	-6.45994
+而	-6.46383
+沒	-6.46559
+下	-6.46656
+聞	-6.47461
+業	-6.47481
+面	-6.48054
+開	-6.49332
+麼	-6.50747
+得	-6.52806
+很	-6.54124
+工	-6.55567
+外	-6.55674
+灣	-6.56599
+然	-6.56858
+二	-6.57314
+作	-6.5751
+用	-6.58056
+和	-6.58782
+院	-6.5907
+最	-6.59402
+者	-6.60519
+樣	-6.60699
+進	-6.60789
+將	-6.61626
+當	-6.61922
+心	-6.62127
+重	-6.62998
+高	-6.63044
+兩	-6.63367
+於	-6.65139
+啊	-6.65966
+動	-6.66322
+表	-6.66465
+定	-6.66561
+去	-6.66991
+關	-6.67833
+些	-6.69245
+百	-6.70403
+事	-6.7105
+好	-6.7145
+實	-6.71726
+如	-6.72458
+黨	-6.73067
+自	-6.73859
+五	-6.75099
+委	-6.77067
+水	-6.77679
+教	-6.77999
+全	-6.79455
+市	-6.79835
+場	-6.81014
+同	-6.82011
+度	-6.82346
+保	-6.8257
+本	-6.83866
+員	-6.84922
+子	-6.85008
+點	-6.85756
+做	-6.87301
+明	-6.87477
+果	-6.87506
+選	-6.87506
+但	-6.87565
+小	-6.88304
+讓	-6.88363
+已	-6.88631
+體	-6.8869
+採	-6.89257
+月	-6.89407
+內	-6.89527
+主	-6.89587
+比	-6.90009
+目	-6.9025
+間	-6.90281
+北	-6.90857
+訪	-6.91712
+期	-6.91805
+力	-6.92792
+加	-6.93103
+提	-6.93571
+你	-6.93821
+合	-6.95655
+四	-6.95783
+起	-6.95943
+從	-6.96392
+機	-6.96908
+裡	-6.97687
+應	-6.97817
+文	-6.9785
+其	-6.98013
+只	-6.98013
+美	-6.98833
+題	-7.0016
+意	-7.00864
+統	-7.0232
+再	-7.03142
+位	-7.03314
+陳	-7.03729
+府	-7.04493
+問	-7.05123
+金	-7.06074
+種	-7.06678
+情	-7.0675
+案	-7.07
+總	-7.07609
+呢	-7.07825
+跟	-7.08476
+正	-7.08585
+九	-7.09095
+示	-7.09865
+更	-7.10789
+議	-7.11759
+候	-7.11947
+化	-7.12474
+常	-7.1289
+六	-7.13194
+它	-7.13423
+資	-7.14303
+回	-7.14766
+理	-7.15932
+調	-7.15932
+區	-7.1601
+認	-7.16441
+與	-7.1731
+電	-7.17429
+八	-7.17786
+性	-7.17866
+受	-7.19635
+第	-7.1992
+入	-7.20042
+齁	-7.20123
+由	-7.20246
+話	-7.20368
+續	-7.20655
+及	-7.2156
+原	-7.22725
+次	-7.23608
+相	-7.23734
+眾	-7.24243
+利	-7.25396
+望	-7.25957
+想	-7.27174
+氣	-7.27788
+通	-7.28097
+各	-7.28539
+希	-7.29118
+老	-7.30782
+除	-7.311
+建	-7.31784
+把	-7.31922
+團	-7.31968
+接	-7.32336
+解	-7.32983
+夠	-7.3368
+展	-7.33821
+基	-7.34055
+少	-7.34477
+著	-7.34524
+產	-7.34996
+等	-7.35281
+什	-7.35613
+任	-7.35804
+林	-7.3633
+育	-7.3773
+費	-7.3773
+程	-7.37876
+七	-7.38022
+或	-7.38119
+師	-7.38364
+元	-7.38461
+觀	-7.38658
+整	-7.39396
+活	-7.39495
+名	-7.40141
+才	-7.40141
+像	-7.4069
+萬	-7.4084
+帶	-7.4089
+代	-7.41142
+安	-7.41242
+較	-7.41747
+被	-7.42
+先	-7.42356
+權	-7.42611
+東	-7.42713
+計	-7.43174
+改	-7.43225
+特	-7.43328
+嗯	-7.43483
+務	-7.43637
+社	-7.4431
+品	-7.45144
+該	-7.45301
+影	-7.45406
+共	-7.46197
+身	-7.4625
+近	-7.471
+物	-7.47261
+平	-7.47528
+辦	-7.48661
+每	-7.49751
+道	-7.49806
+式	-7.51354
+千	-7.52025
+考	-7.52081
+始	-7.52306
+查	-7.52983
+管	-7.53665
+變	-7.53665
+制	-7.5395
+給	-7.54064
+局	-7.54122
+交	-7.54179
+卻	-7.54351
+書	-7.54409
+決	-7.54523
+並	-7.54696
+日	-7.54985
+您	-7.55274
+放	-7.5539
+持	-7.56322
+界	-7.5685
+商	-7.57145
+造	-7.57678
+喔	-7.57797
+哪	-7.58992
+號	-7.59173
+別	-7.59354
+科	-7.59959
+畫	-7.60629
+數	-7.61058
+醫	-7.61058
+海	-7.6155
+陸	-7.61921
+呃	-7.62293
+非	-7.62542
+術	-7.62792
+預	-7.63167
+此	-7.6323
+助	-7.63293
+頭	-7.63481
+治	-7.64112
+許	-7.64747
+收	-7.64811
+達	-7.64811
+己	-7.65002
+結	-7.6513
+張	-7.65194
+山	-7.65578
+舉	-7.65578
+處	-7.65578
+覺	-7.65836
+件	-7.65964
+量	-7.66029
+無	-7.66287
+感	-7.66741
+至	-7.67001
+節	-7.67721
+病	-7.67983
+信	-7.68842
+強	-7.68842
+南	-7.7065
+必	-7.70718
+車	-7.70785
+消	-7.70853
+完	-7.71057
+演	-7.71328
+單	-7.71464
+則	-7.71533
+路	-7.71533
+求	-7.71874
+縣	-7.72148
+馬	-7.72836
+向	-7.73251
+手	-7.73251
+幾	-7.73599
+檢	-7.73808
+樂	-7.74157
+請	-7.74437
+司	-7.75425
+研	-7.76637
+反	-7.76709
+布	-7.7714
+女	-7.78082
+未	-7.78155
+連	-7.78739
+規	-7.78812
+環	-7.78959
+傳	-7.7918
+使	-7.79401
+繼	-7.79401
+爭	-7.79474
+且	-7.79918
+算	-7.80066
+推	-7.80662
+難	-7.80662
+校	-7.81036
+友	-7.81186
+失	-7.81789
+真	-7.83005
+光	-7.83081
+指	-7.83618
+包	-7.83772
+勞	-7.84081
+世	-7.84158
+專	-7.84235
+財	-7.84313
+另	-7.84856
+濟	-7.85012
+設	-7.85012
+又	-7.85247
+健	-7.85325
+取	-7.85403
+流	-7.85481
+投	-7.85638
+知	-7.8611
+條	-7.86982
+打	-7.88021
+風	-7.88182
+空	-7.88909
+記	-7.88909
+邊	-7.88909
+參	-7.89072
+住	-7.89153
+何	-7.89234
+直	-7.89397
+口	-7.89724
+況	-7.89724
+太	-7.89806
+雖	-7.89806
+西	-7.90051
+照	-7.91373
+告	-7.91539
+首	-7.9204
+運	-7.92796
+際	-7.92965
+晚	-7.93134
+率	-7.93303
+她	-7.93473
+引	-7.93643
+農	-7.93813
+容	-7.94411
+技	-7.94411
+聯	-7.94411
+協	-7.95271
+啦	-7.95444
+根	-7.95444
+項	-7.95444
+策	-7.95878
+底	-7.96839
+走	-7.96839
+越	-7.97368
+清	-7.9781
+藝	-7.97899
+質	-7.97899
+價	-7.98254
+親	-7.98611
+歡	-7.9879
+需	-7.99149
+官	-7.99238
+施	-7.99328
+響	-7.99418
+色	-8.00323
+見	-8.00505
+透	-8.01053
+往	-8.0142
+據	-8.0142
+論	-8.02158
+門	-8.02622
+組	-8.02809
+講	-8.03463
+義	-8.03557
+軍	-8.03557
+例	-8.03651
+園	-8.03745
+營	-8.03933
+境	-8.04216
+戰	-8.04311
+態	-8.045
+轉	-8.05453
+星	-8.0574
+象	-8.05836
+須	-8.06029
+創	-8.06125
+源	-8.06899
+障	-8.06899
+花	-8.07483
+朋	-8.07874
+驗	-8.07972
+防	-8.08267
+怎	-8.08563
+億	-8.0876
+票	-8.08958
+署	-8.09157
+確	-8.10557
+英	-8.10759
+擔	-8.11467
+步	-8.11467
+李	-8.11671
+職	-8.12284
+低	-8.13107
+嚴	-8.13314
+零	-8.13521
+修	-8.13729
+究	-8.1425
+落	-8.1425
+息	-8.14564
+音	-8.15195
+領	-8.15195
+孩	-8.153
+錢	-8.153
+服	-8.16575
+客	-8.17543
+標	-8.17651
+福	-8.18628
+察	-8.18738
+支	-8.18847
+護	-8.18847
+雄	-8.18847
+兒	-8.19837
+華	-8.20058
+藥	-8.20168
+黃	-8.20724
+療	-8.20948
+土	-8.22071
+增	-8.22071
+限	-8.22297
+蘭	-8.22411
+約	-8.22524
+系	-8.23093
+岸	-8.23551
+廠	-8.24126
+形	-8.24126
+快	-8.24473
+熱	-8.24821
+證	-8.24821
+景	-8.25286
+具	-8.25403
+半	-8.25403
+供	-8.25989
+效	-8.2646
+深	-8.26933
+興	-8.26933
+衛	-8.26933
+士	-8.27528
+類	-8.27528
+早	-8.27887
+準	-8.28007
+格	-8.28127
+警	-8.28367
+顯	-8.28367
+即	-8.28487
+訴	-8.28487
+球	-8.28608
+隊	-8.28608
+談	-8.28729
+滿	-8.29092
+討	-8.29578
+訊	-8.297
+備	-8.30557
+愛	-8.30803
+曾	-8.31298
+習	-8.31546
+款	-8.31919
+野	-8.32294
+負	-8.32545
+遊	-8.3267
+剛	-8.32796
+審	-8.33553
+料	-8.3368
+幫	-8.33807
+嘛	-8.34062
+配	-8.34062
+迎	-8.34317
+補	-8.34445
+線	-8.34701
+列	-8.35215
+假	-8.35344
+段	-8.35344
+注	-8.35473
+竹	-8.35473
+狀	-8.35732
+片	-8.35861
+宣	-8.36121
+降	-8.36381
+勢	-8.36512
+鄉	-8.36773
+排	-8.37167
+課	-8.3743
+減	-8.37694
+律	-8.37826
+養	-8.38223
+居	-8.38356
+害	-8.38756
+維	-8.38756
+識	-8.38756
+易	-8.3889
+午	-8.3983
+試	-8.39965
+積	-8.40779
+米	-8.40779
+免	-8.40915
+紀	-8.40915
+型	-8.41052
+欸	-8.41052
+針	-8.41052
+歷	-8.41189
+謝	-8.416
+聲	-8.42151
+獎	-8.42289
+站	-8.42289
+歲	-8.42427
+評	-8.42566
+找	-8.42983
+臨	-8.42983
+族	-8.43542
+功	-8.43822
+副	-8.44386
+括	-8.44386
+拿	-8.44527
+集	-8.44527
+留	-8.4481
+念	-8.45094
+油	-8.45094
+登	-8.45237
+超	-8.45237
+止	-8.45522
+願	-8.45522
+執	-8.45808
+群	-8.45951
+溫	-8.46239
+份	-8.46383
+拉	-8.46383
+網	-8.46383
+責	-8.46671
+酒	-8.46671
+語	-8.46961
+編	-8.47251
+依	-8.47688
+銀	-8.48421
+劇	-8.48715
+神	-8.48715
+測	-8.48863
+待	-8.49011
+概	-8.49159
+般	-8.49753
+航	-8.50501
+週	-8.50501
+細	-8.5171
+吃	-8.51862
+製	-8.52014
+買	-8.52014
+雨	-8.52472
+命	-8.52626
+聽	-8.52626
+獲	-8.52779
+差	-8.53087
+彩	-8.53241
+媽	-8.5355
+舞	-8.5355
+係	-8.5386
+存	-8.54016
+離	-8.54171
+益	-8.54327
+便	-8.5464
+媒	-8.54953
+層	-8.54953
+言	-8.54953
+班	-8.5511
+疑	-8.5511
+精	-8.5511
+甚	-8.55583
+席	-8.55899
+死	-8.55899
+斷	-8.56217
+河	-8.56535
+升	-8.56695
+角	-8.56855
+屬	-8.57015
+染	-8.57015
+游	-8.57015
+斯	-8.57658
+核	-8.57981
+革	-8.58305
+朝	-8.58793
+史	-8.59448
+值	-8.59612
+級	-8.59942
+停	-8.61104
+葉	-8.61104
+腦	-8.62279
+白	-8.62448
+印	-8.62618
+塊	-8.62618
+抗	-8.62618
+島	-8.62787
+館	-8.62957
+版	-8.63128
+短	-8.63128
+蘇	-8.63298
+輕	-8.64327
+顧	-8.64673
+申	-8.64846
+足	-8.65019
+險	-8.65019
+判	-8.65193
+企	-8.65367
+擊	-8.65541
+善	-8.65716
+付	-8.66066
+賣	-8.66066
+啟	-8.66242
+復	-8.66242
+王	-8.66242
+食	-8.66242
+休	-8.66418
+極	-8.67125
+壓	-8.6748
+央	-8.67658
+退	-8.68194
+送	-8.68194
+困	-8.68373
+競	-8.68553
+阿	-8.68553
+字	-8.68913
+錯	-8.68913
+毒	-8.69094
+監	-8.69456
+讀	-8.6982
+隆	-8.70368
+救	-8.70551
+廣	-8.70735
+錄	-8.70735
+否	-8.70919
+呼	-8.70919
+綜	-8.71288
+卡	-8.71473
+峰	-8.71473
+換	-8.71473
+洲	-8.71658
+庭	-8.7203
+遭	-8.72778
+石	-8.72966
+稍	-8.72966
+羅	-8.73531
+戲	-8.7391
+眼	-8.741
+遠	-8.741
+額	-8.74863
+優	-8.75055
+破	-8.75055
+礙	-8.75632
+雙	-8.75632
+郭	-8.76213
+溪	-8.76602
+稅	-8.76602
+青	-8.76993
+構	-8.77189
+榮	-8.77385
+隨	-8.77385
+患	-8.77582
+批	-8.77779
+童	-8.77779
+亞	-8.77976
+里	-8.77976
+器	-8.78174
+承	-8.78372
+銷	-8.78372
+拜	-8.78571
+介	-8.7877
+禮	-8.78969
+港	-8.79169
+火	-8.79169
+喜	-8.79369
+庫	-8.79369
+絕	-8.79369
+刑	-8.79771
+吸	-8.79771
+志	-8.79771
+初	-8.80376
+缺	-8.80579
+德	-8.80782
+波	-8.80782
+切	-8.82215
+古	-8.82215
+宗	-8.82215
+適	-8.82215
+災	-8.82628
+估	-8.83251
+速	-8.8346
+慧	-8.83669
+終	-8.83669
+克	-8.8472
+旅	-8.8472
+扁	-8.84932
+戶	-8.85144
+謂	-8.85144
+宜	-8.8557
+傷	-8.86211
+座	-8.86426
+籍	-8.86642
+劃	-8.87074
+召	-8.8729
+範	-8.8729
+恐	-8.87507
+飛	-8.87943
+康	-8.88162
+獨	-8.88381
+令	-8.89041
+富	-8.89041
+血	-8.89041
+密	-8.89928
+裝	-8.90823
+織	-8.91047
+致	-8.91047
+乎	-8.91499
+村	-8.91953
+充	-8.9218
+掉	-8.92408
+暫	-8.92408
+慢	-8.93095
+置	-8.93326
+房	-8.93787
+互	-8.94019
+擇	-8.94019
+昨	-8.94019
+故	-8.94484
+母	-8.94484
+刻	-8.94952
+父	-8.94952
+廢	-8.95421
+延	-8.95421
+歌	-8.95421
+俊	-8.95657
+汙	-8.95657
+釋	-8.95657
+材	-8.95893
+症	-8.96367
+街	-8.96367
+緊	-8.96605
+黑	-8.96605
+練	-8.96843
+洪	-8.97321
+珍	-8.97321
+融	-8.98285
+素	-8.98528
+思	-8.98771
+派	-8.99014
+避	-8.99014
+危	-8.99749
+股	-8.99995
+似	-9.00488
+移	-9.00488
+楚	-9.00736
+船	-9.00736
+雲	-9.00736
+努	-9.00985
+店	-9.00985
+祖	-9.01234
+模	-9.01484
+控	-9.01734
+突	-9.01734
+漲	-9.01985
+域	-9.02995
+江	-9.02995
+煌	-9.03249
+購	-9.03249
+冷	-9.03504
+婦	-9.03504
+賽	-9.03504
+嘉	-9.03759
+巴	-9.03759
+寫	-9.04272
+蕙	-9.04788
+久	-9.05046
+籲	-9.05306
+醒	-9.05306
+仍	-9.06088
+博	-9.06088
+吳	-9.0635
+圖	-9.0635
+訓	-9.0635
+播	-9.06613
+順	-9.06613
+劉	-9.06876
+盟	-9.07405
+牠	-9.07671
+紅	-9.07671
+娟	-9.07937
+券	-9.08472
+簡	-9.08741
+玉	-9.0901
+草	-9.0901
+守	-9.0955
+慶	-9.0955
+叫	-9.09822
+急	-9.09822
+盡	-9.1064
+違	-9.1064
+逐	-9.1119
+竟	-9.11466
+譯	-9.11466
+昆	-9.11742
+靜	-9.11742
+男	-9.1202
+訂	-9.1202
+鐵	-9.1202
+銘	-9.12298
+閣	-9.12577
+夫	-9.13137
+罪	-9.13137
+貨	-9.13137
+爾	-9.137
+錫	-9.13983
+湖	-9.14267
+衝	-9.14267
+怕	-9.14551
+套	-9.14837
+怡	-9.15123
+輔	-9.15123
+犯	-9.1541
+漸	-9.15697
+貸	-9.16276
+慮	-9.16566
+城	-9.16857
+嗎	-9.17442
+龍	-9.17735
+智	-9.1803
+室	-9.18325
+擴	-9.18919
+玩	-9.18919
+稱	-9.18919
+屆	-9.19516
+省	-9.19516
+烈	-9.19816
+罰	-9.19816
+香	-9.19816
+藉	-9.20117
+魚	-9.20721
+幅	-9.2194
+永	-9.2194
+措	-9.22248
+木	-9.22556
+紹	-9.22556
+複	-9.22556
+漁	-9.22865
+貿	-9.22865
+階	-9.22865
+奇	-9.23175
+微	-9.23486
+苦	-9.23486
+附	-9.23486
+右	-9.24111
+威	-9.24111
+爆	-9.24425
+左	-9.2474
+彈	-9.2474
+趕	-9.2474
+陽	-9.2474
+鼓	-9.2474
+夜	-9.25056
+追	-9.25056
+靠	-9.25056
+圍	-9.25691
+韓	-9.25691
+畢	-9.26651
+背	-9.26651
+尊	-9.26973
+幣	-9.26973
+歐	-9.26973
+紛	-9.26973
+賴	-9.26973
+典	-9.27296
+納	-9.27296
+哈	-9.2762
+隻	-9.27945
+餐	-9.27945
+勵	-9.28271
+堂	-9.28271
+拍	-9.28271
+揚	-9.28271
+季	-9.29257
+損	-9.29257
+良	-9.29257
+輪	-9.29257
+徐	-9.30586
+樓	-9.30586
+舊	-9.31257
+鏡	-9.31257
+植	-9.31594
+鎮	-9.31594
+筆	-9.31932
+簽	-9.31932
+豐	-9.31932
+倒	-9.32272
+蓮	-9.32272
+蠻	-9.32272
+春	-9.32613
+棄	-9.32613
+替	-9.33298
+秀	-9.33298
+誤	-9.33298
+普	-9.33642
+綠	-9.33987
+跑	-9.33987
+亡	-9.34334
+周	-9.34334
+答	-9.34334
+探	-9.34682
+貼	-9.34682
+京	-9.35031
+焦	-9.35031
+趣	-9.35031
+曉	-9.35381
+儘	-9.35733
+溝	-9.35733
+癌	-9.35733
+私	-9.35733
+腳	-9.35733
+柏	-9.36085
+欣	-9.36085
+痛	-9.36085
+圓	-9.36439
+樹	-9.36439
+糖	-9.36439
+僱	-9.37151
+售	-9.37509
+幕	-9.37509
+貴	-9.38228
+寶	-9.3859
+潮	-9.3859
+激	-9.3859
+董	-9.3859
+均	-9.38953
+桃	-9.38953
+束	-9.39317
+弱	-9.39683
+涉	-9.39683
+陣	-9.39683
+尼	-9.4005
+穩	-9.4005
+禁	-9.40418
+芳	-9.40788
+遺	-9.41531
+堅	-9.41905
+邀	-9.41905
+亮	-9.4228
+攻	-9.42657
+諾	-9.42657
+余	-9.43035
+唱	-9.43415
+序	-9.44178
+懷	-9.44178
+餘	-9.44178
+射	-9.44562
+藏	-9.44562
+償	-9.44947
+雜	-9.45334
+挑	-9.45722
+曲	-9.45722
+暴	-9.46112
+碼	-9.46112
+聰	-9.46112
+尋	-9.46897
+泰	-9.46897
+符	-9.46897
+賞	-9.46897
+享	-9.47291
+偏	-9.47687
+忙	-9.48085
+奈	-9.48484
+昌	-9.48484
+疾	-9.48484
+築	-9.48885
+味	-9.49287
+徵	-9.49287
+緩	-9.49287
+洋	-9.49691
+疫	-9.49691
+皮	-9.49691
+援	-9.50504
+浪	-9.50913
+宏	-9.51324
+遇	-9.51324
+呈	-9.51736
+憲	-9.51736
+秘	-9.51736
+薪	-9.51736
+培	-9.5215
+索	-9.5215
+軟	-9.5215
+幹	-9.52566
+攝	-9.52566
+爸	-9.52566
+績	-9.52984
+肯	-9.52984
+壞	-9.53403
+掌	-9.53403
+揮	-9.53403
+抽	-9.53824
+耀	-9.53824
+賠	-9.53824
+按	-9.54247
+武	-9.54247
+診	-9.54247
+麗	-9.55098
+仁	-9.55526
+尤	-9.55526
+屋	-9.55526
+捷	-9.55526
+燈	-9.55526
+誰	-9.55526
+玲	-9.55956
+盛	-9.56388
+藍	-9.56388
+繳	-9.56822
+尺	-9.57258
+惠	-9.57258
+捐	-9.57258
+裁	-9.57695
+擁	-9.58576
+籌	-9.58576
+胞	-9.58576
+晶	-9.5902
+異	-9.5902
+振	-9.59913
+齡	-9.59913
+慈	-9.60362
+彭	-9.60814
+架	-9.60814
+澎	-9.60814
+蓋	-9.60814
+距	-9.60814
+麟	-9.61267
+川	-9.61723
+末	-9.61723
+聚	-9.61723
+賄	-9.62181
+鄭	-9.62181
+屏	-9.6264
+牌	-9.6264
+詳	-9.6264
+宮	-9.63102
+觸	-9.63102
+震	-9.63102
+乾	-9.63566
+勒	-9.63566
+殘	-9.63566
+邱	-9.64032
+固	-9.64501
+雅	-9.64501
+夏	-9.65444
+雕	-9.65919
+露	-9.65919
+伊	-9.66397
+諭	-9.66397
+鐘	-9.66397
+唯	-9.67842
+促	-9.68329
+戒	-9.68329
+蔡	-9.68329
+坐	-9.68818
+婚	-9.68818
+授	-9.68818
+略	-9.68818
+紙	-9.68818
+覽	-9.68818
+辛	-9.68818
+汝	-9.69309
+颱	-9.69309
+靈	-9.69803
+佳	-9.703
+搶	-9.713
+洗	-9.713
+炸	-9.713
+穿	-9.713
+雪	-9.713
+招	-9.71803
+殺	-9.71803
+偵	-9.7231
+掃	-9.7231
+板	-9.7231
+滋	-9.7231
+僅	-9.72819
+握	-9.72819
+祝	-9.72819
+章	-9.72819
+累	-9.72819
+覆	-9.72819
+詢	-9.72819
+恆	-9.7333
+津	-9.73844
+輸	-9.73844
+旦	-9.74361
+森	-9.74361
+橋	-9.74361
+縮	-9.7488
+迫	-9.7488
+鬆	-9.7488
+句	-9.75403
+塑	-9.75403
+沙	-9.75403
+跨	-9.75403
+迷	-9.75403
+吧	-9.75928
+頻	-9.75928
+侵	-9.76455
+瑤	-9.76455
+亂	-9.76986
+坤	-9.76986
+敢	-9.76986
+殊	-9.76986
+泉	-9.76986
+怖	-9.77519
+漫	-9.78055
+鍵	-9.78055
+哥	-9.78594
+飲	-9.78594
+坦	-9.79136
+某	-9.79681
+炎	-9.79681
+散	-9.80229
+途	-9.80229
+勝	-9.8078
+托	-9.8078
+擾	-9.8078
+菸	-9.8078
+鬥	-9.8078
+折	-9.81334
+搭	-9.81334
+盲	-9.81334
+歸	-9.82452
+純	-9.82452
+肉	-9.82452
+遷	-9.83015
+雷	-9.83015
+剩	-9.83582
+獻	-9.83582
+佔	-9.84151
+債	-9.84151
+煙	-9.84151
+蹈	-9.84151
+恢	-9.84724
+既	-9.85301
+熟	-9.85301
+袁	-9.85301
+圾	-9.85881
+垃	-9.85881
+慣	-9.85881
+蔣	-9.85881
+蟲	-9.85881
+譬	-9.85881
+趨	-9.85881
+寬	-9.86464
+泛	-9.86464
+菜	-9.8705
+訟	-9.8705
+遍	-9.8705
+扣	-9.8764
+抵	-9.8764
+澤	-9.8764
+盤	-9.8764
+龔	-9.8764
+隔	-9.88234
+榜	-9.88831
+瑞	-9.88831
+含	-9.89431
+奏	-9.89431
+擬	-9.89431
+田	-9.89431
+乏	-9.90035
+倍	-9.90035
+聖	-9.90035
+跳	-9.90035
+州	-9.90643
+淡	-9.90643
+飯	-9.90643
+串	-9.91255
+寒	-9.91255
+幸	-9.91255
+幼	-9.91255
+惡	-9.91255
+潛	-9.91255
+析	-9.9187
+狗	-9.9187
+硬	-9.9187
+端	-9.9187
+苗	-9.9187
+徒	-9.9249
+託	-9.9249
+貫	-9.9249
+擺	-9.93113
+閒	-9.93113
+宋	-9.9374
+脫	-9.9374
+賓	-9.9374
+儀	-9.9437
+珊	-9.9437
+督	-9.9437
+肥	-9.9437
+憂	-9.95005
+賺	-9.95005
+予	-9.95644
+巡	-9.95644
+昇	-9.95644
+罹	-9.96288
+駐	-9.96288
+敗	-9.96935
+顏	-9.96935
+併	-9.97586
+冬	-9.97586
+誠	-9.97586
+閱	-9.97586
+拆	-9.98242
+燒	-9.98242
+茶	-9.98242
+袋	-9.98242
+骨	-9.98242
+麻	-9.98242
+旁	-9.98902
+獄	-9.98902
+呂	-9.99566
+哲	-9.99566
+描	-9.99566
+操	-10.0024
+敏	-10.0024
+朗	-10.0024
+牙	-10.0024
+床	-10.0091
+君	-10.0159
+呀	-10.0159
+憶	-10.0159
+衣	-10.0159
+誌	-10.0227
+隱	-10.0227
+勇	-10.0296
+嫌	-10.0296
+拒	-10.0296
+汽	-10.0296
+偷	-10.0365
+誕	-10.0365
+輝	-10.0365
+鮮	-10.0365
+淨	-10.0435
+疏	-10.0435
+菁	-10.0435
+阻	-10.0435
+匯	-10.0505
+廷	-10.0505
+欠	-10.0505
+瓶	-10.0505
+衡	-10.0505
+倫	-10.0575
+封	-10.0575
+翻	-10.0575
+耶	-10.0575
+默	-10.0575
+丁	-10.0647
+契	-10.0647
+帳	-10.0647
+楊	-10.0647
+盜	-10.0647
+繁	-10.0647
+胡	-10.0647
+塞	-10.0718
+忽	-10.0718
+貝	-10.0718
+夢	-10.079
+搜	-10.079
+梅	-10.079
+沈	-10.079
+辭	-10.0863
+壇	-10.0936
+摩	-10.0936
+晴	-10.0936
+汗	-10.0936
+贊	-10.0936
+邦	-10.0936
+驚	-10.0936
+廳	-10.101
+抓	-10.101
+捕	-10.101
+祭	-10.101
+忘	-10.116
+笑	-10.116
+蛋	-10.116
+逃	-10.116
+刺	-10.1235
+喝	-10.1235
+扮	-10.1235
+池	-10.1235
+烏	-10.1235
+牛	-10.1235
+稻	-10.1235
+贈	-10.1235
+蹤	-10.1235
+吋	-10.1311
+暑	-10.1311
+駕	-10.1311
+俄	-10.1388
+兼	-10.1388
+爐	-10.1388
+抱	-10.1465
+棒	-10.1465
+鬧	-10.1465
+擠	-10.1543
+忠	-10.1621
+撤	-10.1621
+濃	-10.1621
+碰	-10.1621
+辯	-10.1621
+兆	-10.17
+搖	-10.17
+沿	-10.17
+瑚	-10.17
+裂	-10.17
+膠	-10.178
+菌	-10.178
+遲	-10.178
+乘	-10.186
+劑	-10.186
+晨	-10.186
+浮	-10.186
+礎	-10.186
+毅	-10.1941
+混	-10.1941
+莉	-10.1941
+鈔	-10.1941
+濫	-10.2023
+兵	-10.2105
+懲	-10.2105
+朱	-10.2105
+瑪	-10.2105
+莊	-10.2105
+陪	-10.2105
+鳥	-10.2105
+嬰	-10.2188
+鋒	-10.2188
+鎖	-10.2188
+頓	-10.2188
+冒	-10.2272
+脅	-10.2272
+顆	-10.2272
+伯	-10.2356
+怪	-10.2356
+脈	-10.2356
+蹟	-10.2356
+吉	-10.2441
+填	-10.2441
+映	-10.2441
+貌	-10.2441
+販	-10.2441
+魯	-10.2441
+厚	-10.2527
+嘿	-10.2527
+巨	-10.2527
+撥	-10.2527
+牽	-10.2527
+租	-10.2527
+艦	-10.2527
+輩	-10.2527
+佈	-10.2614
+嶼	-10.2614
+毛	-10.2614
+琴	-10.2614
+蚊	-10.2614
+鋼	-10.2614
+盈	-10.2701
+聘	-10.2701
+臉	-10.2701
+返	-10.2701
+坡	-10.2789
+堆	-10.2789
+奶	-10.2789
+娘	-10.2789
+役	-10.2789
+荷	-10.2789
+袖	-10.2789
+載	-10.2789
+宿	-10.2878
+巧	-10.2878
+拼	-10.2878
+旗	-10.2878
+洞	-10.2878
+胎	-10.2878
+臟	-10.2878
+錦	-10.2878
+姓	-10.2968
+懂	-10.2968
+掛	-10.2968
+氛	-10.2968
+澳	-10.2968
+瞭	-10.2968
+紐	-10.3058
+腸	-10.3058
+膽	-10.3058
+贏	-10.3058
+禎	-10.315
+糾	-10.315
+肝	-10.315
+虧	-10.315
+詞	-10.315
+豪	-10.315
+渡	-10.3242
+踏	-10.3242
+佛	-10.3335
+占	-10.3335
+弊	-10.3335
+跡	-10.3335
+攤	-10.3429
+漢	-10.3429
+輻	-10.3429
+刪	-10.3523
+敬	-10.3523
+湯	-10.3523
+澄	-10.3523
+秋	-10.3523
+耕	-10.3523
+鄰	-10.3523
+塔	-10.3619
+漏	-10.3619
+粉	-10.3619
+伴	-10.3716
+冰	-10.3716
+姐	-10.3716
+戴	-10.3716
+斑	-10.3716
+耳	-10.3716
+跌	-10.3716
+龐	-10.3716
+弟	-10.3813
+徹	-10.3813
+殖	-10.3813
+洛	-10.3813
+窮	-10.3813
+賀	-10.3813
+偶	-10.3912
+募	-10.3912
+斤	-10.3912
+松	-10.3912
+氏	-10.3912
+頒	-10.3912
+咖	-10.4011
+滑	-10.4011
+甲	-10.4011
+鯨	-10.4011
+鳳	-10.4011
+歉	-10.4112
+泥	-10.4112
+瑜	-10.4112
+閉	-10.4112
+齣	-10.4112
+准	-10.4213
+喊	-10.4213
+夕	-10.4213
+暖	-10.4213
+詩	-10.4213
+乳	-10.4316
+噴	-10.4316
+壽	-10.4316
+彼	-10.4316
+槍	-10.4316
+謀	-10.4316
+郵	-10.4316
+雞	-10.4316
+瓦	-10.442
+述	-10.442
+鑑	-10.442
+頂	-10.442
+仔	-10.463
+孟	-10.463
+桌	-10.463
+漂	-10.463
+谷	-10.463
+啡	-10.4737
+墾	-10.4737
+宅	-10.4737
+拖	-10.4737
+歧	-10.4737
+絲	-10.4737
+蒙	-10.4737
+軌	-10.4737
+迴	-10.4737
+頃	-10.4737
+干	-10.4845
+搬	-10.4845
+栗	-10.4845
+豬	-10.4845
+駛	-10.4845
+齊	-10.4845
+俗	-10.4954
+忍	-10.4954
+押	-10.4954
+髮	-10.4954
+梁	-10.5065
+涵	-10.5065
+睛	-10.5065
+礁	-10.5065
+荒	-10.5065
+借	-10.5177
+凍	-10.5177
+慎	-10.5177
+截	-10.5177
+揭	-10.5177
+旱	-10.5177
+瓜	-10.5177
+肢	-10.5177
+蝶	-10.5177
+圈	-10.529
+廚	-10.529
+泡	-10.529
+滅	-10.529
+若	-10.529
+鼠	-10.529
+丟	-10.5404
+妹	-10.5404
+寧	-10.5404
+潔	-10.5404
+繪	-10.5404
+菲	-10.5404
+陷	-10.5404
+添	-10.552
+罷	-10.552
+遣	-10.552
+仟	-10.5636
+尿	-10.5636
+炮	-10.5636
+翠	-10.5636
+膚	-10.5636
+襲	-10.5636
+貧	-10.5636
+墨	-10.5755
+奉	-10.5755
+椅	-10.5755
+榔	-10.5755
+櫃	-10.5755
+燃	-10.5755
+牆	-10.5755
+牧	-10.5755
+絡	-10.5755
+妨	-10.5875
+孫	-10.5875
+彰	-10.5875
+扶	-10.5875
+挖	-10.5875
+曆	-10.5875
+淹	-10.5875
+緒	-10.5875
+輛	-10.5875
+闆	-10.5875
+噸	-10.5996
+毀	-10.5996
+耐	-10.5996
+伐	-10.6119
+搞	-10.6119
+檳	-10.6119
+液	-10.6119
+綁	-10.6119
+龜	-10.6119
+刷	-10.6243
+珠	-10.6243
+甘	-10.6243
+莫	-10.6243
+伸	-10.6369
+勁	-10.6369
+喪	-10.6369
+妻	-10.6369
+循	-10.6369
+惜	-10.6369
+賭	-10.6369
+儲	-10.6496
+窗	-10.6496
+肌	-10.6496
+虛	-10.6496
+鍾	-10.6496
+餅	-10.6496
+傾	-10.6625
+坑	-10.6625
+廟	-10.6625
+弄	-10.6625
+梓	-10.6625
+溼	-10.6625
+甦	-10.6625
+陰	-10.6625
+霧	-10.6625
+鴨	-10.6625
+唸	-10.6756
+悉	-10.6756
+撈	-10.6756
+暗	-10.6756
+棟	-10.6756
+淺	-10.6756
+燕	-10.6756
+紋	-10.6756
+舒	-10.6756
+刀	-10.6888
+堤	-10.6888
+尖	-10.6888
+悲	-10.6888
+憑	-10.6888
+摸	-10.6888
+洩	-10.6888
+胖	-10.6888
+兌	-10.7022
+暢	-10.7022
+檔	-10.7022
+犬	-10.7022
+箱	-10.7022
+遵	-10.7022
+仰	-10.7158
+婆	-10.7158
+沖	-10.7158
+卸	-10.7296
+寄	-10.7296
+徽	-10.7296
+敵	-10.7296
+牲	-10.7296
+菊	-10.7296
+黎	-10.7296
+凌	-10.7436
+奮	-10.7436
+泳	-10.7436
+瘤	-10.7436
+膜	-10.7436
+騎	-10.7436
+鼻	-10.7436
+割	-10.7578
+灰	-10.7578
+煩	-10.7578
+矚	-10.7578
+締	-10.7578
+賦	-10.7578
+酸	-10.7578
+冠	-10.7722
+刊	-10.7722
+勤	-10.7722
+嚇	-10.7722
+址	-10.7722
+妳	-10.7722
+孔	-10.7722
+孝	-10.7722
+尾	-10.7722
+拔	-10.7722
+昭	-10.7722
+蒐	-10.7722
+薦	-10.7722
+衰	-10.7722
+諮	-10.7722
+趁	-10.7722
+陶	-10.7722
+怨	-10.7868
+悠	-10.7868
+柯	-10.7868
+毫	-10.7868
+睡	-10.7868
+筋	-10.7868
+貢	-10.7868
+躍	-10.7868
+邁	-10.7868
+釐	-10.7868
+飽	-10.7868
+催	-10.8016
+嘗	-10.8016
+奧	-10.8016
+娃	-10.8016
+寮	-10.8016
+皇	-10.8016
+盧	-10.8016
+礦	-10.8016
+壁	-10.8166
+杜	-10.8166
+橫	-10.8166
+氧	-10.8166
+犧	-10.8166
+伍	-10.8319
+傍	-10.8319
+庸	-10.8319
+爺	-10.8319
+番	-10.8319
+舍	-10.8319
+譽	-10.8319
+遞	-10.8319
+郎	-10.8319
+鹽	-10.8319
+濤	-10.8474
+縱	-10.8474
+誓	-10.8474
+赴	-10.8474
+丸	-10.8632
+喘	-10.8632
+妥	-10.8632
+婕	-10.8632
+愈	-10.8632
+插	-10.8632
+杯	-10.8632
+桶	-10.8632
+滾	-10.8632
+紫	-10.8632
+綾	-10.8632
+脆	-10.8632
+螢	-10.8632
+貓	-10.8632
+飾	-10.8632
+麥	-10.8632
+岩	-10.8792
+枝	-10.8792
+甸	-10.8792
+聊	-10.8792
+峽	-10.8954
+斥	-10.8954
+棲	-10.8954
+滴	-10.8954
+玫	-10.8954
+玻	-10.8954
+芝	-10.8954
+芬	-10.8954
+虎	-10.8954
+辨	-10.8954
+徑	-10.912
+敦	-10.912
+篩	-10.912
+註	-10.912
+輟	-10.912
+騙	-10.912
+埔	-10.9288
+彎	-10.9288
+慰	-10.9288
+拓	-10.9288
+焚	-10.9288
+罕	-10.9288
+耗	-10.9288
+蓄	-10.9288
+釀	-10.9288
+駁	-10.9288
+冊	-10.9459
+塗	-10.9459
+姚	-10.9459
+孕	-10.9459
+崇	-10.9459
+撐	-10.9459
+璃	-10.9459
+癮	-10.9459
+砍	-10.9459
+繫	-10.9459
+脂	-10.9459
+豚	-10.9459
+趟	-10.9459
+鞋	-10.9459
+偉	-10.9632
+卑	-10.9632
+卷	-10.9632
+姝	-10.9632
+憾	-10.9632
+沉	-10.9632
+祕	-10.9632
+翡	-10.9632
+葛	-10.9632
+蕭	-10.9632
+酵	-10.9632
+丙	-10.9809
+傑	-10.9809
+姻	-10.9809
+孤	-10.9809
+擋	-10.9809
+碟	-10.9809
+緝	-10.9809
+緣	-10.9809
+緻	-10.9809
+茂	-10.9809
+僑	-10.999
+壯	-10.999
+撞	-10.999
+旺	-10.999
+凱	-11.0173
+叛	-11.0173
+慘	-11.0173
+敲	-11.0173
+椰	-11.0173
+狂	-11.0173
+腔	-11.0173
+艱	-11.0173
+仲	-11.036
+刮	-11.036
+剝	-11.036
+哼	-11.036
+淑	-11.036
+潤	-11.036
+盃	-11.036
+粽	-11.036
+逼	-11.036
+頸	-11.036
+騷	-11.036
+丹	-11.0551
+偽	-11.0551
+吐	-11.0551
+恩	-11.0551
+涼	-11.0551
+煮	-11.0551
+熊	-11.0551
+盼	-11.0551
+眷	-11.0551
+秉	-11.0551
+綱	-11.0551
+罵	-11.0551
+趙	-11.0551
+迪	-11.0551
+遙	-11.0551
+骸	-11.0551
+黏	-11.0551
+吹	-11.0745
+呆	-11.0745
+撫	-11.0745
+攸	-11.0745
+獵	-11.0745
+盾	-11.0745
+繞	-11.0745
+芭	-11.0745
+菇	-11.0745
+萊	-11.0745
+藻	-11.0745
+誘	-11.0745
+醇	-11.0745
+勸	-11.0943
+唐	-11.0943
+喚	-11.0943
+奔	-11.0943
+奪	-11.0943
+姦	-11.0943
+擱	-11.0943
+甄	-11.0943
+痕	-11.0943
+砂	-11.0943
+脊	-11.0943
+勳	-11.1145
+匪	-11.1145
+妙	-11.1145
+尚	-11.1145
+廁	-11.1145
+慕	-11.1145
+抬	-11.1145
+棵	-11.1145
+牢	-11.1145
+盒	-11.1145
+稚	-11.1145
+腐	-11.1145
+腰	-11.1145
+踢	-11.1145
+麵	-11.1145
+傅	-11.1351
+凡	-11.1351
+凸	-11.1351
+圳	-11.1351
+坪	-11.1351
+墓	-11.1351
+壤	-11.1351
+崩	-11.1351
+攀	-11.1351
+旬	-11.1351
+甜	-11.1351
+稽	-11.1351
+肺	-11.1351
+謹	-11.1351
+逮	-11.1351
+魔	-11.1351
+冤	-11.1562
+堡	-11.1562
+巷	-11.1562
+廖	-11.1562
+挫	-11.1562
+祥	-11.1562
+籠	-11.1562
+纖	-11.1562
+翁	-11.1562
+葬	-11.1562
+貶	-11.1562
+迅	-11.1562
+鬼	-11.1562
+側	-11.1777
+宵	-11.1777
+帝	-11.1777
+幻	-11.1777
+彌	-11.1777
+椎	-11.1777
+疼	-11.1777
+碑	-11.1777
+艘	-11.1777
+蕾	-11.1777
+鯊	-11.1777
+鴻	-11.1777
+削	-11.1996
+吊	-11.1996
+塵	-11.1996
+幽	-11.1996
+柴	-11.1996
+瘦	-11.1996
+腫	-11.1996
+蝠	-11.1996
+裔	-11.1996
+采	-11.1996
+銅	-11.1996
+馨	-11.1996
+剪	-11.2221
+唉	-11.2221
+埋	-11.2221
+妝	-11.2221
+姆	-11.2221
+掩	-11.2221
+揆	-11.2221
+曼	-11.2221
+梨	-11.2221
+湧	-11.2221
+灌	-11.2221
+爬	-11.2221
+磨	-11.2221
+籤	-11.2221
+蔬	-11.2221
+允	-11.2451
+凝	-11.2451
+厲	-11.2451
+呵	-11.2451
+挺	-11.2451
+檻	-11.2451
+殼	-11.2451
+璩	-11.2451
+羊	-11.2451
+鑽	-11.2451
+僵	-11.2686
+吵	-11.2686
+哭	-11.2686
+娛	-11.2686
+婉	-11.2686
+廉	-11.2686
+琪	-11.2686
+痺	-11.2686
+稀	-11.2686
+篇	-11.2686
+薩	-11.2686
+蝴	-11.2686
+詮	-11.2686
+謠	-11.2686
+赫	-11.2686
+躲	-11.2686
+輯	-11.2686
+銜	-11.2686
+陌	-11.2686
+函	-11.2927
+埃	-11.2927
+姿	-11.2927
+峇	-11.2927
+攜	-11.2927
+濁	-11.2927
+琳	-11.2927
+瑯	-11.2927
+稿	-11.2927
+竊	-11.2927
+糕	-11.2927
+蔓	-11.2927
+薄	-11.2927
+謨	-11.2927
+霞	-11.2927
+飄	-11.2927
+騰	-11.2927
+卓	-11.3174
+抨	-11.3174
+旋	-11.3174
+盆	-11.3174
+秩	-11.3174
+糊	-11.3174
+腹	-11.3174
+莎	-11.3174
+萍	-11.3174
+蛙	-11.3174
+詐	-11.3174
+賢	-11.3174
+醉	-11.3174
+鵝	-11.3174
+寵	-11.3427
+抹	-11.3427
+挪	-11.3427
+擦	-11.3427
+曬	-11.3427
+栽	-11.3427
+梯	-11.3427
+濱	-11.3427
+灑	-11.3427
+町	-11.3427
+盪	-11.3427
+粹	-11.3427
+紓	-11.3427
+翔	-11.3427
+茜	-11.3427
+萎	-11.3427
+蝙	-11.3427
+餵	-11.3427
+僚	-11.3687
+勘	-11.3687
+咬	-11.3687
+喀	-11.3687
+嘴	-11.3687
+夥	-11.3687
+恰	-11.3687
+懼	-11.3687
+拋	-11.3687
+挽	-11.3687
+枯	-11.3687
+汰	-11.3687
+滯	-11.3687
+爛	-11.3687
+猩	-11.3687
+琛	-11.3687
+禍	-11.3687
+秒	-11.3687
+芽	-11.3687
+赤	-11.3687
+辜	-11.3687
+鷺	-11.3687
+夾	-11.3954
+嫁	-11.3954
+戀	-11.3954
+杉	-11.3954
+欺	-11.3954
+涯	-11.3954
+潑	-11.3954
+碧	-11.3954
+祈	-11.3954
+翰	-11.3954
+肚	-11.3954
+胃	-11.3954
+臍	-11.3954
+蒂	-11.3954
+蘋	-11.3954
+雁	-11.3954
+乙	-11.4228
+井	-11.4228
+勾	-11.4228
+姑	-11.4228
+宴	-11.4228
+屍	-11.4228
+憤	-11.4228
+撒	-11.4228
+浩	-11.4228
+淚	-11.4228
+疲	-11.4228
+緬	-11.4228
+萄	-11.4228
+葡	-11.4228
+譜	-11.4228
+郁	-11.4228
+饋	-11.4228
+魂	-11.4228
+堵	-11.451
+怒	-11.451
+挾	-11.451
+曜	-11.451
+朵	-11.451
+檜	-11.451
+浴	-11.451
+眉	-11.451
+禪	-11.451
+粒	-11.451
+粗	-11.451
+翼	-11.451
+肖	-11.451
+蘊	-11.451
+轄	-11.451
+逝	-11.451
+酷	-11.451
+霸	-11.451
+鞭	-11.451
+驅	-11.451
+倉	-11.4799
+兄	-11.4799
+咦	-11.4799
+坵	-11.4799
+塌	-11.4799
+拚	-11.4799
+橘	-11.4799
+灘	-11.4799
+炭	-11.4799
+砲	-11.4799
+笛	-11.4799
+羈	-11.4799
+胚	-11.4799
+艾	-11.4799
+蜜	-11.4799
+裕	-11.4799
+銳	-11.4799
+闖	-11.4799
+頗	-11.4799
+囚	-11.5098
+孳	-11.5098
+慾	-11.5098
+捉	-11.5098
+捨	-11.5098
+撿	-11.5098
+昂	-11.5098
+汁	-11.5098
+矛	-11.5098
+碎	-11.5098
+緹	-11.5098
+胸	-11.5098
+腎	-11.5098
+衍	-11.5098
+辰	-11.5098
+閩	-11.5098
+飆	-11.5098
+鬱	-11.5098
+魅	-11.5098
+俠	-11.5406
+勉	-11.5406
+坊	-11.5406
+墮	-11.5406
+岳	-11.5406
+曝	-11.5406
+槽	-11.5406
+欲	-11.5406
+溯	-11.5406
+煉	-11.5406
+疊	-11.5406
+穎	-11.5406
+肩	-11.5406
+蟹	-11.5406
+踩	-11.5406
+轟	-11.5406
+醜	-11.5406
+闊	-11.5406
+鷹	-11.5406
+嬌	-11.5723
+崗	-11.5723
+廈	-11.5723
+彥	-11.5723
+愉	-11.5723
+慌	-11.5723
+框	-11.5723
+棚	-11.5723
+漿	-11.5723
+炳	-11.5723
+爵	-11.5723
+瑰	-11.5723
+肅	-11.5723
+脹	-11.5723
+膩	-11.5723
+艙	-11.5723
+茲	-11.5723
+虐	-11.5723
+虹	-11.5723
+蚵	-11.5723
+誼	-11.5723
+遜	-11.5723
+鄧	-11.5723
+鍰	-11.5723
+闢	-11.5723
+陵	-11.5723
+兔	-11.6051
+劫	-11.6051
+卵	-11.6051
+嘆	-11.6051
+囊	-11.6051
+堪	-11.6051
+奕	-11.6051
+屈	-11.6051
+撲	-11.6051
+柔	-11.6051
+樸	-11.6051
+檯	-11.6051
+氯	-11.6051
+汐	-11.6051
+淘	-11.6051
+淤	-11.6051
+猜	-11.6051
+矮	-11.6051
+碗	-11.6051
+碳	-11.6051
+腿	-11.6051
+蒜	-11.6051
+蜂	-11.6051
+蝦	-11.6051
+諒	-11.6051
+貪	-11.6051
+逕	-11.6051
+酬	-11.6051
+鈴	-11.6051
+靳	-11.6051
+乃	-11.639
+划	-11.639
+宇	-11.639
+寸	-11.639
+巒	-11.639
+廊	-11.639
+悅	-11.639
+敘	-11.639
+斗	-11.639
+梭	-11.639
+械	-11.639
+歇	-11.639
+猛	-11.639
+睹	-11.639
+穴	-11.639
+讚	-11.639
+邏	-11.639
+鉉	-11.639
+頁	-11.639
+鹿	-11.639
+齒	-11.639
+仙	-11.6741
+俱	-11.6741
+傘	-11.6741
+勃	-11.6741
+匆	-11.6741
+垮	-11.6741
+姊	-11.6741
+宕	-11.6741
+巫	-11.6741
+懸	-11.6741
+暨	-11.6741
+泊	-11.6741
+涂	-11.6741
+濾	-11.6741
+瑕	-11.6741
+縫	-11.6741
+胰	-11.6741
+蛛	-11.6741
+軸	-11.6741
+逢	-11.6741
+鍊	-11.6741
+隧	-11.6741
+丈	-11.7105
+抑	-11.7105
+捲	-11.7105
+掀	-11.7105
+擅	-11.7105
+攔	-11.7105
+攬	-11.7105
+潭	-11.7105
+澱	-11.7105
+猶	-11.7105
+碩	-11.7105
+磁	-11.7105
+糞	-11.7105
+蜘	-11.7105
+蝕	-11.7105
+諸	-11.7105
+豆	-11.7105
+鴉	-11.7105
+仿	-11.7482
+偕	-11.7482
+后	-11.7482
+哀	-11.7482
+哇	-11.7482
+娜	-11.7482
+娼	-11.7482
+崎	-11.7482
+扇	-11.7482
+扯	-11.7482
+拾	-11.7482
+棗	-11.7482
+氾	-11.7482
+洽	-11.7482
+琵	-11.7482
+瑄	-11.7482
+疵	-11.7482
+睜	-11.7482
+窘	-11.7482
+籃	-11.7482
+纏	-11.7482
+罩	-11.7482
+范	-11.7482
+贖	-11.7482
+閃	-11.7482
+髒	-11.7482
+乖	-11.7874
+埤	-11.7874
+壘	-11.7874
+夷	-11.7874
+寓	-11.7874
+帽	-11.7874
+悄	-11.7874
+掘	-11.7874
+斌	-11.7874
+桿	-11.7874
+歹	-11.7874
+焊	-11.7874
+綿	-11.7874
+艷	-11.7874
+逾	-11.7874
+佑	-11.8282
+佩	-11.8282
+兇	-11.8282
+卹	-11.8282
+噪	-11.8282
+囤	-11.8282
+寺	-11.8282
+嶺	-11.8282
+悟	-11.8282
+旭	-11.8282
+曹	-11.8282
+淵	-11.8282
+穆	-11.8282
+肪	-11.8282
+臭	-11.8282
+茫	-11.8282
+蒸	-11.8282
+詹	-11.8282
+諧	-11.8282
+譴	-11.8282
+郝	-11.8282
+佰	-11.8708
+傲	-11.8708
+噶	-11.8708
+尬	-11.8708
+尷	-11.8708
+巢	-11.8708
+弘	-11.8708
+忌	-11.8708
+惱	-11.8708
+撼	-11.8708
+擄	-11.8708
+攏	-11.8708
+昏	-11.8708
+沛	-11.8708
+漆	-11.8708
+狹	-11.8708
+磯	-11.8708
+穫	-11.8708
+糟	-11.8708
+紮	-11.8708
+臣	-11.8708
+裸	-11.8708
+覓	-11.8708
+赦	-11.8708
+軒	-11.8708
+鼎	-11.8708
+于	-11.9153
+喬	-11.9153
+屢	-11.9153
+庚	-11.9153
+恭	-11.9153
+惑	-11.9153
+摘	-11.9153
+樺	-11.9153
+汞	-11.9153
+淪	-11.9153
+窄	-11.9153
+紗	-11.9153
+艇	-11.9153
+蒲	-11.9153
+蓬	-11.9153
+衷	-11.9153
+貯	-11.9153
+遮	-11.9153
+鉛	-11.9153
+靡	-11.9153
+駿	-11.9153
+仗	-11.9618
+卿	-11.9618
+厘	-11.9618
+塢	-11.9618
+扭	-11.9618
+斐	-11.9618
+晃	-11.9618
+枉	-11.9618
+浸	-11.9618
+渴	-11.9618
+澡	-11.9618
+磚	-11.9618
+糧	-11.9618
+臥	-11.9618
+舖	-11.9618
+萃	-11.9618
+諷	-11.9618
+謎	-11.9618
+辱	-11.9618
+逛	-11.9618
+鳴	-11.9618
+亭	-12.0106
+侯	-12.0106
+劣	-12.0106
+卜	-12.0106
+喇	-12.0106
+喻	-12.0106
+垂	-12.0106
+塘	-12.0106
+墊	-12.0106
+墜	-12.0106
+壟	-12.0106
+廂	-12.0106
+慨	-12.0106
+扎	-12.0106
+抒	-12.0106
+掏	-12.0106
+柱	-12.0106
+榕	-12.0106
+樁	-12.0106
+泌	-12.0106
+湊	-12.0106
+溉	-12.0106
+潘	-12.0106
+燙	-12.0106
+燥	-12.0106
+獸	-12.0106
+瘋	-12.0106
+碌	-12.0106
+禦	-12.0106
+聾	-12.0106
+蔚	-12.0106
+証	-12.0106
+豹	-12.0106
+貞	-12.0106
+鉅	-12.0106
+鑼	-12.0106
+鑿	-12.0106
+魄	-12.0106
+侶	-12.0619
+倡	-12.0619
+坍	-12.0619
+屯	-12.0619
+幟	-12.0619
+弦	-12.0619
+汪	-12.0619
+淋	-12.0619
+瀉	-12.0619
+煞	-12.0619
+獅	-12.0619
+癒	-12.0619
+眠	-12.0619
+箭	-12.0619
+臂	-12.0619
+虞	-12.0619
+蠟	-12.0619
+醞	-12.0619
+髓	-12.0619
+鮪	-12.0619
+鷲	-12.0619
+侷	-12.1159
+俘	-12.1159
+倪	-12.1159
+卉	-12.1159
+卯	-12.1159
+嚐	-12.1159
+墅	-12.1159
+壩	-12.1159
+姨	-12.1159
+姬	-12.1159
+婷	-12.1159
+寂	-12.1159
+履	-12.1159
+嶄	-12.1159
+帕	-12.1159
+庇	-12.1159
+拘	-12.1159
+搓	-12.1159
+晉	-12.1159
+柑	-12.1159
+殷	-12.1159
+淇	-12.1159
+煤	-12.1159
+燦	-12.1159
+爽	-12.1159
+玄	-12.1159
+瑋	-12.1159
+瓊	-12.1159
+疆	-12.1159
+睞	-12.1159
+竭	-12.1159
+筒	-12.1159
+紡	-12.1159
+耘	-12.1159
+蕩	-12.1159
+蛾	-12.1159
+袱	-12.1159
+諱	-12.1159
+逆	-12.1159
+遏	-12.1159
+醬	-12.1159
+驟	-12.1159
+魏	-12.1159
+劾	-12.1731
+吞	-12.1731
+墩	-12.1731
+妓	-12.1731
+妮	-12.1731
+寞	-12.1731
+寡	-12.1731
+屁	-12.1731
+彗	-12.1731
+抄	-12.1731
+拳	-12.1731
+斜	-12.1731
+枚	-12.1731
+欽	-12.1731
+殿	-12.1731
+炒	-12.1731
+烹	-12.1731
+畜	-12.1731
+皂	-12.1731
+禽	-12.1731
+竄	-12.1731
+竿	-12.1731
+糙	-12.1731
+罐	-12.1731
+臘	-12.1731
+舟	-12.1731
+蔗	-12.1731
+訝	-12.1731
+賜	-12.1731
+踐	-12.1731
+踴	-12.1731
+躬	-12.1731
+輒	-12.1731
+邪	-12.1731
+銓	-12.1731
+鑰	-12.1731
+隸	-12.1731
+靖	-12.1731
+鞠	-12.1731
+飼	-12.1731
+鹼	-12.1731
+叢	-12.2337
+哨	-12.2337
+帥	-12.2337
+庄	-12.2337
+彙	-12.2337
+忱	-12.2337
+愁	-12.2337
+桂	-12.2337
+桐	-12.2337
+桑	-12.2337
+棘	-12.2337
+沾	-12.2337
+浦	-12.2337
+潰	-12.2337
+瀕	-12.2337
+烘	-12.2337
+畏	-12.2337
+瞻	-12.2337
+砸	-12.2337
+穀	-12.2337
+窒	-12.2337
+窩	-12.2337
+簿	-12.2337
+耨	-12.2337
+芒	-12.2337
+蘆	-12.2337
+蛇	-12.2337
+蠹	-12.2337
+褲	-12.2337
+襄	-12.2337
+訣	-12.2337
+謊	-12.2337
+躺	-12.2337
+酗	-12.2337
+酪	-12.2337
+霍	-12.2337
+頑	-12.2337
+顱	-12.2337
+颼	-12.2337
+飢	-12.2337
+駭	-12.2337
+丘	-12.2983
+咯	-12.2983
+哺	-12.2983
+嚼	-12.2983
+墟	-12.2983
+壺	-12.2983
+孑	-12.2983
+徙	-12.2983
+悶	-12.2983
+捍	-12.2983
+朽	-12.2983
+毯	-12.2983
+漠	-12.2983
+甫	-12.2983
+瘡	-12.2983
+睿	-12.2983
+繃	-12.2983
+蚓	-12.2983
+蚯	-12.2983
+謙	-12.2983
+豔	-12.2983
+輿	-12.2983
+鄒	-12.2983
+釣	-12.2983
+闡	-12.2983
+仇	-12.3672
+俸	-12.3672
+啤	-12.3672
+壢	-12.3672
+孓	-12.3672
+巖	-12.3672
+征	-12.3672
+怠	-12.3672
+恨	-12.3672
+憐	-12.3672
+捏	-12.3672
+斃	-12.3672
+斬	-12.3672
+曙	-12.3672
+欒	-12.3672
+氓	-12.3672
+沃	-12.3672
+洶	-12.3672
+溜	-12.3672
+澈	-12.3672
+熬	-12.3672
+燭	-12.3672
+琦	-12.3672
+瑛	-12.3672
+繽	-12.3672
+脾	-12.3672
+腺	-12.3672
+葆	-12.3672
+螃	-12.3672
+螺	-12.3672
+辣	-12.3672
+鋪	-12.3672
+閘	-12.3672
+陀	-12.3672
+霖	-12.3672
+霜	-12.3672
+靚	-12.3672
+餓	-12.3672
+伏	-12.4414
+伙	-12.4414
+卦	-12.4414
+叮	-12.4414
+咎	-12.4414
+咳	-12.4414
+圭	-12.4414
+姜	-12.4414
+崙	-12.4414
+弓	-12.4414
+悼	-12.4414
+憨	-12.4414
+拯	-12.4414
+挨	-12.4414
+摔	-12.4414
+撓	-12.4414
+擎	-12.4414
+敞	-12.4414
+旨	-12.4414
+晤	-12.4414
+曠	-12.4414
+朴	-12.4414
+柳	-12.4414
+歪	-12.4414
+沫	-12.4414
+滄	-12.4414
+滲	-12.4414
+澆	-12.4414
+炫	-12.4414
+烯	-12.4414
+璋	-12.4414
+癬	-12.4414
+癱	-12.4414
+盯	-12.4414
+瞧	-12.4414
+禱	-12.4414
+窯	-12.4414
+肇	-12.4414
+膀	-12.4414
+蟻	-12.4414
+詠	-12.4414
+誇	-12.4414
+賈	-12.4414
+邸	-12.4414
+釘	-12.4414
+鈞	-12.4414
+鉤	-12.4414
+鋸	-12.4414
+鍋	-12.4414
+顛	-12.4414
+饒	-12.4414
+馳	-12.4414
+鱗	-12.4414
+黛	-12.4414
+亨	-12.5214
+倖	-12.5214
+儼	-12.5214
+匿	-12.5214
+孽	-12.5214
+弗	-12.5214
+悔	-12.5214
+惶	-12.5214
+惹	-12.5214
+憩	-12.5214
+披	-12.5214
+昔	-12.5214
+昧	-12.5214
+昱	-12.5214
+柵	-12.5214
+栩	-12.5214
+棣	-12.5214
+棺	-12.5214
+橡	-12.5214
+沐	-12.5214
+淆	-12.5214
+渺	-12.5214
+澀	-12.5214
+牡	-12.5214
+狼	-12.5214
+瑣	-12.5214
+皆	-12.5214
+瞄	-12.5214
+硫	-12.5214
+礫	-12.5214
+籬	-12.5214
+耍	-12.5214
+腕	-12.5214
+臻	-12.5214
+舌	-12.5214
+苛	-12.5214
+荼	-12.5214
+蒼	-12.5214
+薑	-12.5214
+袍	-12.5214
+裹	-12.5214
+迢	-12.5214
+鈷	-12.5214
+闈	-12.5214
+陋	-12.5214
+雀	-12.5214
+驕	-12.5214
+亦	-12.6084
+佣	-12.6084
+厭	-12.6084
+咽	-12.6084
+喧	-12.6084
+嗓	-12.6084
+嗽	-12.6084
+奠	-12.6084
+宰	-12.6084
+峻	-12.6084
+懶	-12.6084
+抖	-12.6084
+拭	-12.6084
+搏	-12.6084
+暈	-12.6084
+札	-12.6084
+株	-12.6084
+椿	-12.6084
+濬	-12.6084
+瀰	-12.6084
+猴	-12.6084
+瓷	-12.6084
+疽	-12.6084
+瘓	-12.6084
+瞬	-12.6084
+砌	-12.6084
+禿	-12.6084
+篤	-12.6084
+粟	-12.6084
+耽	-12.6084
+胺	-12.6084
+脖	-12.6084
+膨	-12.6084
+茹	-12.6084
+藤	-12.6084
+蝗	-12.6084
+螞	-12.6084
+蠵	-12.6084
+謬	-12.6084
+遴	-12.6084
+醣	-12.6084
+銲	-12.6084
+鎔	-12.6084
+阱	-12.6084
+霾	-12.6084
+伶	-12.7037
+俐	-12.7037
+傻	-12.7037
+儉	-12.7037
+凹	-12.7037
+嚮	-12.7037
+埕	-12.7037
+宙	-12.7037
+屠	-12.7037
+岡	-12.7037
+忡	-12.7037
+恤	-12.7037
+惕	-12.7037
+戈	-12.7037
+扛	-12.7037
+扼	-12.7037
+揉	-12.7037
+摧	-12.7037
+撕	-12.7037
+晒	-12.7037
+沸	-12.7037
+渾	-12.7037
+溶	-12.7037
+熄	-12.7037
+爪	-12.7037
+痢	-12.7037
+眨	-12.7037
+矽	-12.7037
+磋	-12.7037
+祐	-12.7037
+禧	-12.7037
+穗	-12.7037
+窟	-12.7037
+箏	-12.7037
+綢	-12.7037
+繩	-12.7037
+缸	-12.7037
+罌	-12.7037
+羞	-12.7037
+翅	-12.7037
+膳	-12.7037
+薇	-12.7037
+誹	-12.7037
+諦	-12.7037
+謗	-12.7037
+趾	-12.7037
+逗	-12.7037
+郊	-12.7037
+酌	-12.7037
+酮	-12.7037
+釁	-12.7037
+韌	-12.7037
+鮑	-12.7037
+鷗	-12.7037
+倆	-12.8091
+匹	-12.8091
+唇	-12.8091
+唷	-12.8091
+喉	-12.8091
+嘔	-12.8091
+垢	-12.8091
+墳	-12.8091
+峙	-12.8091
+巾	-12.8091
+帆	-12.8091
+弭	-12.8091
+彷	-12.8091
+愕	-12.8091
+懈	-12.8091
+捧	-12.8091
+撇	-12.8091
+撰	-12.8091
+斧	-12.8091
+晰	-12.8091
+梵	-12.8091
+棋	-12.8091
+樞	-12.8091
+殃	-12.8091
+汛	-12.8091
+渠	-12.8091
+瀾	-12.8091
+烤	-12.8091
+璿	-12.8091
+癥	-12.8091
+秦	-12.8091
+稜	-12.8091
+竅	-12.8091
+綻	-12.8091
+羨	-12.8091
+翊	-12.8091
+聆	-12.8091
+膏	-12.8091
+舵	-12.8091
+芮	-12.8091
+芹	-12.8091
+茵	-12.8091
+葵	-12.8091
+蔘	-12.8091
+薯	-12.8091
+藩	-12.8091
+襟	-12.8091
+詭	-12.8091
+誦	-12.8091
+蹲	-12.8091
+鈉	-12.8091
+鈕	-12.8091
+鈣	-12.8091
+鑒	-12.8091
+霉	-12.8091
+鶯	-12.8091
+侍	-12.9269
+兢	-12.9269
+凋	-12.9269
+凶	-12.9269
+剷	-12.9269
+匙	-12.9269
+厝	-12.9269
+吻	-12.9269
+咕	-12.9269
+嚥	-12.9269
+囂	-12.9269
+堰	-12.9269
+嫖	-12.9269
+帖	-12.9269
+廝	-12.9269
+彿	-12.9269
+戚	-12.9269
+拮	-12.9269
+摹	-12.9269
+敷	-12.9269
+殯	-12.9269
+氫	-12.9269
+汶	-12.9269
+沮	-12.9269
+湛	-12.9269
+溢	-12.9269
+瀆	-12.9269
+灼	-12.9269
+焰	-12.9269
+熔	-12.9269
+牯	-12.9269
+畔	-12.9269
+癢	-12.9269
+瞞	-12.9269
+矯	-12.9269
+笨	-12.9269
+簾	-12.9269
+綵	-12.9269
+繹	-12.9269
+羽	-12.9269
+翩	-12.9269
+聳	-12.9269
+肆	-12.9269
+肛	-12.9269
+茅	-12.9269
+萱	-12.9269
+薰	-12.9269
+褪	-12.9269
+襪	-12.9269
+襯	-12.9269
+詬	-12.9269
+躁	-12.9269
+輾	-12.9269
+轎	-12.9269
+逸	-12.9269
+銫	-12.9269
+錶	-12.9269
+鍛	-12.9269
+鏽	-12.9269
+鑫	-12.9269
+阮	-12.9269
+雋	-12.9269
+雯	-12.9269
+鞏	-12.9269
+頌	-12.9269
+餾	-12.9269
+骼	-12.9269
+齋	-12.9269
+侃	-13.0604
+倚	-13.0604
+傢	-13.0604
+僧	-13.0604
+儕	-13.0604
+冥	-13.0604
+剎	-13.0604
+叉	-13.0604
+哎	-13.0604
+噓	-13.0604
+圃	-13.0604
+坻	-13.0604
+嫻	-13.0604
+寨	-13.0604
+岱	-13.0604
+崖	-13.0604
+徊	-13.0604
+徘	-13.0604
+慷	-13.0604
+憫	-13.0604
+戎	-13.0604
+戳	-13.0604
+扒	-13.0604
+扳	-13.0604
+挹	-13.0604
+摻	-13.0604
+斂	-13.0604
+杰	-13.0604
+椒	-13.0604
+楣	-13.0604
+櫥	-13.0604
+淞	-13.0604
+渣	-13.0604
+漓	-13.0604
+狄	-13.0604
+猖	-13.0604
+瑟	-13.0604
+疝	-13.0604
+癡	-13.0604
+眺	-13.0604
+瞎	-13.0604
+矩	-13.0604
+祀	-13.0604
+祉	-13.0604
+祿	-13.0604
+紘	-13.0604
+絮	-13.0604
+綴	-13.0604
+繆	-13.0604
+耆	-13.0604
+聶	-13.0604
+舜	-13.0604
+芋	-13.0604
+茄	-13.0604
+茱	-13.0604
+莖	-13.0604
+菱	-13.0604
+藐	-13.0604
+蘿	-13.0604
+虔	-13.0604
+蛀	-13.0604
+裙	-13.0604
+裳	-13.0604
+詰	-13.0604
+贓	-13.0604
+趴	-13.0604
+遐	-13.0604
+邵	-13.0604
+郡	-13.0604
+鏈	-13.0604
+隙	-13.0604
+雛	-13.0604
+顫	-13.0604
+餚	-13.0604
+馮	-13.0604
+魟	-13.0604
+鵬	-13.0604
+鶴	-13.0604
+鹹	-13.0604
+黯	-13.0604
+么	-13.2145
+伺	-13.2145
+侮	-13.2145
+傭	-13.2145
+僻	-13.2145
+冗	-13.2145
+剔	-13.2145
+劍	-13.2145
+勿	-13.2145
+叔	-13.2145
+叩	-13.2145
+吼	-13.2145
+啞	-13.2145
+喂	-13.2145
+嘟	-13.2145
+嘲	-13.2145
+嚕	-13.2145
+奢	-13.2145
+妃	-13.2145
+娓	-13.2145
+婪	-13.2145
+婿	-13.2145
+媳	-13.2145
+嬋	-13.2145
+屑	-13.2145
+崑	-13.2145
+庶	-13.2145
+廿	-13.2145
+弔	-13.2145
+惰	-13.2145
+懇	-13.2145
+拷	-13.2145
+掠	-13.2145
+揣	-13.2145
+摒	-13.2145
+攪	-13.2145
+暇	-13.2145
+曦	-13.2145
+杖	-13.2145
+枋	-13.2145
+枕	-13.2145
+棉	-13.2145
+楓	-13.2145
+榴	-13.2145
+櫻	-13.2145
+欄	-13.2145
+泄	-13.2145
+渥	-13.2145
+滌	-13.2145
+熙	-13.2145
+燬	-13.2145
+狒	-13.2145
+狩	-13.2145
+獗	-13.2145
+琉	-13.2145
+疹	-13.2145
+痊	-13.2145
+痰	-13.2145
+瘍	-13.2145
+矗	-13.2145
+祂	-13.2145
+穌	-13.2145
+筱	-13.2145
+絞	-13.2145
+綽	-13.2145
+腥	-13.2145
+臆	-13.2145
+茁	-13.2145
+莞	-13.2145
+蔔	-13.2145
+蜓	-13.2145
+蜥	-13.2145
+蜻	-13.2145
+衫	-13.2145
+褐	-13.2145
+諜	-13.2145
+諺	-13.2145
+讞	-13.2145
+賂	-13.2145
+賑	-13.2145
+轍	-13.2145
+迦	-13.2145
+遼	-13.2145
+遽	-13.2145
+鋁	-13.2145
+鎂	-13.2145
+閥	-13.2145
+陡	-13.2145
+頤	-13.2145
+餃	-13.2145
+餡	-13.2145
+饕	-13.2145
+鬣	-13.2145
+丑	-13.3969
+丰	-13.3969
+亢	-13.3969
+体	-13.3969
+佤	-13.3969
+俯	-13.3969
+兜	-13.3969
+冉	-13.3969
+刁	-13.3969
+剖	-13.3969
+匠	-13.3969
+叭	-13.3969
+吒	-13.3969
+吟	-13.3969
+哄	-13.3969
+哩	-13.3969
+啼	-13.3969
+嗅	-13.3969
+嗜	-13.3969
+噁	-13.3969
+嚨	-13.3969
+囉	-13.3969
+夭	-13.3969
+娥	-13.3969
+嫩	-13.3969
+嬉	-13.3969
+嬸	-13.3969
+宛	-13.3969
+寢	-13.3969
+弧	-13.3969
+彬	-13.3969
+恫	-13.3969
+愧	-13.3969
+拱	-13.3969
+掙	-13.3969
+搐	-13.3969
+摟	-13.3969
+擲	-13.3969
+斡	-13.3969
+曖	-13.3969
+杭	-13.3969
+栓	-13.3969
+梢	-13.3969
+棍	-13.3969
+毆	-13.3969
+泣	-13.3969
+湘	-13.3969
+潢	-13.3969
+瀏	-13.3969
+瀧	-13.3969
+燉	-13.3969
+狙	-13.3969
+珮	-13.3969
+璉	-13.3969
+畸	-13.3969
+疸	-13.3969
+痙	-13.3969
+痴	-13.3969
+皺	-13.3969
+睫	-13.3969
+磺	-13.3969
+祁	-13.3969
+秤	-13.3969
+竣	-13.3969
+筍	-13.3969
+紜	-13.3969
+羌	-13.3969
+肘	-13.3969
+舅	-13.3969
+芥	-13.3969
+荔	-13.3969
+荻	-13.3969
+萌	-13.3969
+蒞	-13.3969
+蔀	-13.3969
+蕃	-13.3969
+蕉	-13.3969
+薛	-13.3969
+蛻	-13.3969
+蜴	-13.3969
+蟬	-13.3969
+蠶	-13.3969
+誣	-13.3969
+謾	-13.3969
+譎	-13.3969
+蹄	-13.3969
+鑣	-13.3969
+陲	-13.3969
+鞍	-13.3969
+韋	-13.3969
+韻	-13.3969
+飪	-13.3969
+饗	-13.3969
+馴	-13.3969
+鬚	-13.3969
+魁	-13.3969
+鮭	-13.3969
+鵑	-13.3969
+乍	-13.62
+亟	-13.62
+佐	-13.62
+侖	-13.62
+俏	-13.62
+倩	-13.62
+僥	-13.62
+凰	-13.62
+刨	-13.62
+吩	-13.62
+呎	-13.62
+咀	-13.62
+咚	-13.62
+咸	-13.62
+噱	-13.62
+囪	-13.62
+圜	-13.62
+奄	-13.62
+嬴	-13.62
+孵	-13.62
+孺	-13.62
+屹	-13.62
+岌	-13.62
+峨	-13.62
+峭	-13.62
+巽	-13.62
+幌	-13.62
+怯	-13.62
+怵	-13.62
+懿	-13.62
+扔	-13.62
+拇	-13.62
+拌	-13.62
+拙	-13.62
+据	-13.62
+掐	-13.62
+搗	-13.62
+摳	-13.62
+擂	-13.62
+攣	-13.62
+斟	-13.62
+柬	-13.62
+柺	-13.62
+柿	-13.62
+梆	-13.62
+棰	-13.62
+榨	-13.62
+榻	-13.62
+槤	-13.62
+歎	-13.62
+氨	-13.62
+氮	-13.62
+汀	-13.62
+汲	-13.62
+沼	-13.62
+浙	-13.62
+浚	-13.62
+涮	-13.62
+湄	-13.62
+湮	-13.62
+溺	-13.62
+滔	-13.62
+潟	-13.62
+澹	-13.62
+瀑	-13.62
+炙	-13.62
+烙	-13.62
+烽	-13.62
+煎	-13.62
+煒	-13.62
+煥	-13.62
+珣	-13.62
+琢	-13.62
+璧	-13.62
+畝	-13.62
+疤	-13.62
+痘	-13.62
+盔	-13.62
+眩	-13.62
+睦	-13.62
+睽	-13.62
+瞰	-13.62
+矇	-13.62
+磅	-13.62
+窺	-13.62
+筠	-13.62
+筷	-13.62
+粱	-13.62
+糯	-13.62
+絨	-13.62
+緋	-13.62
+縛	-13.62
+繡	-13.62
+芸	-13.62
+苑	-13.62
+苓	-13.62
+荳	-13.62
+莘	-13.62
+蓉	-13.62
+蕈	-13.62
+蕨	-13.62
+蕪	-13.62
+虱	-13.62
+蠢	-13.62
+裴	-13.62
+覈	-13.62
+訶	-13.62
+譚	-13.62
+豁	-13.62
+豎	-13.62
+賊	-13.62
+賤	-13.62
+贅	-13.62
+跋	-13.62
+蹋	-13.62
+蹦	-13.62
+蹺	-13.62
+躂	-13.62
+躪	-13.62
+軀	-13.62
+遂	-13.62
+錨	-13.62
+鍚	-13.62
+閡	-13.62
+頰	-13.62
+馭	-13.62
+駱	-13.62
+驥	-13.62
+魘	-13.62
+鯉	-13.62
+鶩	-13.62
+仕	-13.9077
+倘	-13.9077
+倦	-13.9077
+儂	-13.9077
+冀	-13.9077
+剃	-13.9077
+剌	-13.9077
+匍	-13.9077
+匐	-13.9077
+匣	-13.9077
+匱	-13.9077
+厄	-13.9077
+叱	-13.9077
+吝	-13.9077
+咱	-13.9077
+哮	-13.9077
+唏	-13.9077
+唾	-13.9077
+喋	-13.9077
+嗑	-13.9077
+嗦	-13.9077
+嗨	-13.9077
+嘎	-13.9077
+噢	-13.9077
+嚷	-13.9077
+囑	-13.9077
+塭	-13.9077
+壅	-13.9077
+奚	-13.9077
+姪	-13.9077
+媚	-13.9077
+寥	-13.9077
+尉	-13.9077
+尹	-13.9077
+岐	-13.9077
+崔	-13.9077
+崛	-13.9077
+嶇	-13.9077
+嶽	-13.9077
+巔	-13.9077
+弩	-13.9077
+彪	-13.9077
+御	-13.9077
+徨	-13.9077
+恕	-13.9077
+恥	-13.9077
+悖	-13.9077
+慚	-13.9077
+憧	-13.9077
+憬	-13.9077
+戕	-13.9077
+扥	-13.9077
+拐	-13.9077
+摺	-13.9077
+攘	-13.9077
+晞	-13.9077
+晟	-13.9077
+晾	-13.9077
+曄	-13.9077
+曳	-13.9077
+梗	-13.9077
+梳	-13.9077
+棕	-13.9077
+棠	-13.9077
+棧	-13.9077
+楠	-13.9077
+榭	-13.9077
+榷	-13.9077
+樟	-13.9077
+橢	-13.9077
+殉	-13.9077
+浬	-13.9077
+涕	-13.9077
+淒	-13.9077
+淫	-13.9077
+淮	-13.9077
+渲	-13.9077
+湍	-13.9077
+溥	-13.9077
+漥	-13.9077
+濂	-13.9077
+濛	-13.9077
+灸	-13.9077
+燮	-13.9077
+爍	-13.9077
+牟	-13.9077
+狠	-13.9077
+璀	-13.9077
+璇	-13.9077
+瓣	-13.9077
+疚	-13.9077
+痞	-13.9077
+痧	-13.9077
+瘟	-13.9077
+皎	-13.9077
+皿	-13.9077
+睬	-13.9077
+礡	-13.9077
+祺	-13.9077
+秧	-13.9077
+笆	-13.9077
+筊	-13.9077
+筏	-13.9077
+箋	-13.9077
+箴	-13.9077
+簣	-13.9077
+糜	-13.9077
+紳	-13.9077
+絃	-13.9077
+緯	-13.9077
+縝	-13.9077
+繭	-13.9077
+纜	-13.9077
+羶	-13.9077
+羸	-13.9077
+耿	-13.9077
+芙	-13.9077
+苞	-13.9077
+苟	-13.9077
+茍	-13.9077
+茸	-13.9077
+莠	-13.9077
+葫	-13.9077
+蓓	-13.9077
+蔑	-13.9077
+蔥	-13.9077
+蚩	-13.9077
+蜀	-13.9077
+螂	-13.9077
+衪	-13.9077
+袒	-13.9077
+褒	-13.9077
+褘	-13.9077
+褥	-13.9077
+覦	-13.9077
+覬	-13.9077
+誨	-13.9077
+諉	-13.9077
+謁	-13.9077
+譁	-13.9077
+豫	-13.9077
+貳	-13.9077
+跪	-13.9077
+踰	-13.9077
+蹂	-13.9077
+蹊	-13.9077
+輓	-13.9077
+迺	-13.9077
+逍	-13.9077
+遨	-13.9077
+酋	-13.9077
+醮	-13.9077
+釧	-13.9077
+銬	-13.9077
+錡	-13.9077
+錮	-13.9077
+頹	-13.9077
+餌	-13.9077
+駝	-13.9077
+骷	-13.9077
+髏	-13.9077
+鱷	-13.9077
+鱺	-13.9077
+鴛	-13.9077
+鴿	-13.9077
+鷥	-13.9077
+黜	-13.9077
+乞	-14.3132
+仆	-14.3132
+佬	-14.3132
+佷	-14.3132
+佼	-14.3132
+俞	-14.3132
+儒	-14.3132
+冕	-14.3132
+凳	-14.3132
+刃	-14.3132
+剋	-14.3132
+剿	-14.3132
+勻	-14.3132
+匈	-14.3132
+匝	-14.3132
+卒	-14.3132
+吁	-14.3132
+吆	-14.3132
+吠	-14.3132
+吶	-14.3132
+吾	-14.3132
+呱	-14.3132
+咄	-14.3132
+咆	-14.3132
+咒	-14.3132
+咫	-14.3132
+唆	-14.3132
+喲	-14.3132
+喳	-14.3132
+嗆	-14.3132
+嗇	-14.3132
+嗚	-14.3132
+嘖	-14.3132
+嘩	-14.3132
+嘰	-14.3132
+嚀	-14.3132
+坎	-14.3132
+坨	-14.3132
+垣	-14.3132
+埠	-14.3132
+堯	-14.3132
+塹	-14.3132
+壹	-14.3132
+奴	-14.3132
+妄	-14.3132
+姣	-14.3132
+姮	-14.3132
+娣	-14.3132
+婁	-14.3132
+媲	-14.3132
+嫂	-14.3132
+嫡	-14.3132
+孰	-14.3132
+屎	-14.3132
+峴	-14.3132
+帷	-14.3132
+廓	-14.3132
+徬	-14.3132
+忪	-14.3132
+忿	-14.3132
+恃	-14.3132
+恙	-14.3132
+恣	-14.3132
+恿	-14.3132
+惺	-14.3132
+愿	-14.3132
+慫	-14.3132
+戊	-14.3132
+戮	-14.3132
+扈	-14.3132
+拈	-14.3132
+拎	-14.3132
+挶	-14.3132
+捅	-14.3132
+捩	-14.3132
+揀	-14.3132
+揪	-14.3132
+搤	-14.3132
+摀	-14.3132
+摜	-14.3132
+摯	-14.3132
+撩	-14.3132
+撻	-14.3132
+擒	-14.3132
+擷	-14.3132
+擻	-14.3132
+晝	-14.3132
+暄	-14.3132
+曇	-14.3132
+朔	-14.3132
+枷	-14.3132
+柄	-14.3132
+柚	-14.3132
+榆	-14.3132
+槳	-14.3132
+殞	-14.3132
+沌	-14.3132
+洄	-14.3132
+浹	-14.3132
+涅	-14.3132
+涸	-14.3132
+淦	-14.3132
+淳	-14.3132
+滬	-14.3132
+漣	-14.3132
+漧	-14.3132
+漪	-14.3132
+漱	-14.3132
+漳	-14.3132
+漾	-14.3132
+澐	-14.3132
+瀚	-14.3132
+瀟	-14.3132
+灶	-14.3132
+炊	-14.3132
+炯	-14.3132
+烊	-14.3132
+焉	-14.3132
+焙	-14.3132
+牴	-14.3132
+犁	-14.3132
+猥	-14.3132
+玨	-14.3132
+珈	-14.3132
+琅	-14.3132
+琨	-14.3132
+瑩	-14.3132
+瑾	-14.3132
+璨	-14.3132
+璽	-14.3132
+瓢	-14.3132
+甥	-14.3132
+甯	-14.3132
+疇	-14.3132
+疣	-14.3132
+痍	-14.3132
+皚	-14.3132
+盬	-14.3132
+眈	-14.3132
+瞌	-14.3132
+瞥	-14.3132
+瞪	-14.3132
+瞳	-14.3132
+矓	-14.3132
+祟	-14.3132
+祠	-14.3132
+稠	-14.3132
+笠	-14.3132
+箔	-14.3132
+篙	-14.3132
+篷	-14.3132
+簷	-14.3132
+簸	-14.3132
+粥	-14.3132
+粼	-14.3132
+糰	-14.3132
+綺	-14.3132
+纍	-14.3132
+羹	-14.3132
+翦	-14.3132
+翱	-14.3132
+翹	-14.3132
+胯	-14.3132
+腴	-14.3132
+膝	-14.3132
+膾	-14.3132
+臀	-14.3132
+舢	-14.3132
+芃	-14.3132
+芎	-14.3132
+芯	-14.3132
+芻	-14.3132
+苔	-14.3132
+苯	-14.3132
+苳	-14.3132
+莆	-14.3132
+莓	-14.3132
+菩	-14.3132
+蒴	-14.3132
+蓍	-14.3132
+蚤	-14.3132
+蜒	-14.3132
+蜿	-14.3132
+螳	-14.3132
+袂	-14.3132
+褫	-14.3132
+褻	-14.3132
+訛	-14.3132
+詔	-14.3132
+詛	-14.3132
+詼	-14.3132
+誡	-14.3132
+謐	-14.3132
+謢	-14.3132
+譏	-14.3132
+跤	-14.3132
+蹶	-14.3132
+轅	-14.3132
+轆	-14.3132
+迥	-14.3132
+迭	-14.3132
+遁	-14.3132
+遑	-14.3132
+邃	-14.3132
+鄙	-14.3132
+酢	-14.3132
+醃	-14.3132
+醋	-14.3132
+醯	-14.3132
+釗	-14.3132
+釜	-14.3132
+釦	-14.3132
+錕	-14.3132
+鍘	-14.3132
+鍥	-14.3132
+鍬	-14.3132
+鎊	-14.3132
+鎩	-14.3132
+鏟	-14.3132
+鏤	-14.3132
+鐮	-14.3132
+鑲	-14.3132
+閻	-14.3132
+闌	-14.3132
+闕	-14.3132
+阪	-14.3132
+陜	-14.3132
+霄	-14.3132
+霆	-14.3132
+霹	-14.3132
+靶	-14.3132
+餞	-14.3132
+饑	-14.3132
+騁	-14.3132
+髦	-14.3132
+鬍	-14.3132
+魷	-14.3132
+鯖	-14.3132
+鯛	-14.3132
+鴦	-14.3132
+麒	-14.3132
+齦	-14.3132
+齲	-14.3132
+丐	-15.0063
+丕	-15.0063
+乒	-15.0063
+乓	-15.0063
+仄	-15.0063
+伎	-15.0063
+伕	-15.0063
+佃	-15.0063
+佇	-15.0063
+侈	-15.0063
+侑	-15.0063
+俑	-15.0063
+倔	-15.0063
+倣	-15.0063
+傀	-15.0063
+僈	-15.0063
+兀	-15.0063
+冑	-15.0063
+剁	-15.0063
+剉	-15.0063
+匕	-15.0063
+厥	-15.0063
+叟	-15.0063
+叼	-15.0063
+吏	-15.0063
+呦	-15.0063
+咋	-15.0063
+咐	-15.0063
+咨	-15.0063
+咪	-15.0063
+哦	-15.0063
+哽	-15.0063
+唄	-15.0063
+唬	-15.0063
+啃	-15.0063
+啣	-15.0063
+啥	-15.0063
+啪	-15.0063
+喵	-15.0063
+嗄	-15.0063
+嘯	-15.0063
+嘶	-15.0063
+噗	-15.0063
+噩	-15.0063
+噬	-15.0063
+嚓	-15.0063
+囍	-15.0063
+圴	-15.0063
+埸	-15.0063
+壑	-15.0063
+奎	-15.0063
+奐	-15.0063
+奘	-15.0063
+妊	-15.0063
+妖	-15.0063
+妾	-15.0063
+娠	-15.0063
+婀	-15.0063
+婓	-15.0063
+嫚	-15.0063
+嫣	-15.0063
+嬤	-15.0063
+孿	-15.0063
+寇	-15.0063
+寐	-15.0063
+屜	-15.0063
+岔	-15.0063
+崁	-15.0063
+崢	-15.0063
+嵋	-15.0063
+嵩	-15.0063
+嶸	-15.0063
+巍	-15.0063
+帑	-15.0063
+帘	-15.0063
+帚	-15.0063
+帛	-15.0063
+幔	-15.0063
+幗	-15.0063
+幛	-15.0063
+幡	-15.0063
+廬	-15.0063
+弈	-15.0063
+弛	-15.0063
+彊	-15.0063
+彤	-15.0063
+徇	-15.0063
+徉	-15.0063
+徜	-15.0063
+忐	-15.0063
+忑	-15.0063
+悍	-15.0063
+悚	-15.0063
+悱	-15.0063
+悴	-15.0063
+悸	-15.0063
+惇	-15.0063
+惘	-15.0063
+惟	-15.0063
+惻	-15.0063
+愜	-15.0063
+愫	-15.0063
+愷	-15.0063
+憊	-15.0063
+憔	-15.0063
+懦	-15.0063
+懵	-15.0063
+懾	-15.0063
+戡	-15.0063
+戩	-15.0063
+扉	-15.0063
+扺	-15.0063
+抉	-15.0063
+拄	-15.0063
+拴	-15.0063
+捆	-15.0063
+捺	-15.0063
+捻	-15.0063
+掬	-15.0063
+掰	-15.0063
+揍	-15.0063
+揹	-15.0063
+搆	-15.0063
+搥	-15.0063
+攢	-15.0063
+攫	-15.0063
+昀	-15.0063
+晦	-15.0063
+暉	-15.0063
+暱	-15.0063
+曩	-15.0063
+朮	-15.0063
+杆	-15.0063
+杏	-15.0063
+杓	-15.0063
+杞	-15.0063
+杵	-15.0063
+枸	-15.0063
+桔	-15.0063
+椪	-15.0063
+楞	-15.0063
+楨	-15.0063
+楷	-15.0063
+楸	-15.0063
+楹	-15.0063
+榖	-15.0063
+槃	-15.0063
+槌	-15.0063
+槲	-15.0063
+樵	-15.0063
+橇	-15.0063
+檀	-15.0063
+檬	-15.0063
+檸	-15.0063
+欷	-15.0063
+殆	-15.0063
+殭	-15.0063
+殲	-15.0063
+毋	-15.0063
+毌	-15.0063
+毓	-15.0063
+氐	-15.0063
+氦	-15.0063
+汎	-15.0063
+沚	-15.0063
+泓	-15.0063
+洒	-15.0063
+浥	-15.0063
+涇	-15.0063
+涎	-15.0063
+涷	-15.0063
+淀	-15.0063
+淅	-15.0063
+渙	-15.0063
+渝	-15.0063
+渭	-15.0063
+湃	-15.0063
+湎	-15.0063
+滸	-15.0063
+漕	-15.0063
+濘	-15.0063
+濺	-15.0063
+瀝	-15.0063
+炘	-15.0063
+炬	-15.0063
+炷	-15.0063
+焗	-15.0063
+煽	-15.0063
+熨	-15.0063
+燁	-15.0063
+燜	-15.0063
+燻	-15.0063
+燼	-15.0063
+爹	-15.0063
+爻	-15.0063
+牒	-15.0063
+犀	-15.0063
+犢	-15.0063
+狐	-15.0063
+狡	-15.0063
+狸	-15.0063
+猝	-15.0063
+猷	-15.0063
+猾	-15.0063
+獼	-15.0063
+玥	-15.0063
+玼	-15.0063
+珀	-15.0063
+琥	-15.0063
+琺	-15.0063
+瑙	-15.0063
+璟	-15.0063
+瓅	-15.0063
+瓔	-15.0063
+甪	-15.0063
+痠	-15.0063
+痲	-15.0063
+痹	-15.0063
+瘀	-15.0063
+瘠	-15.0063
+癇	-15.0063
+癲	-15.0063
+皓	-15.0063
+盎	-15.0063
+盥	-15.0063
+矣	-15.0063
+硝	-15.0063
+硨	-15.0063
+硼	-15.0063
+碇	-15.0063
+碘	-15.0063
+碾	-15.0063
+���	-15.0063
+磲	-15.0063
+磷	-15.0063
+祇	-15.0063
+秏	-15.0063
+穢	-15.0063
+穹	-15.0063
+窈	-15.0063
+窕	-15.0063
+窖	-15.0063
+窪	-15.0063
+窿	-15.0063
+竽	-15.0063
+笈	-15.0063
+笙	-15.0063
+筐	-15.0063
+筑	-15.0063
+箇	-15.0063
+箕	-15.0063
+簧	-15.0063
+簪	-15.0063
+籟	-15.0063
+籮	-15.0063
+籽	-15.0063
+粲	-15.0063
+糠	-15.0063
+糢	-15.0063
+紆	-15.0063
+紉	-15.0063
+紕	-15.0063
+絀	-15.0063
+絆	-15.0063
+絢	-15.0063
+綏	-15.0063
+緘	-15.0063
+緞	-15.0063
+繕	-15.0063
+繚	-15.0063
+缽	-15.0063
+罔	-15.0063
+罝	-15.0063
+羔	-15.0063
+羲	-15.0063
+翎	-15.0063
+肋	-15.0063
+肴	-15.0063
+胱	-15.0063
+脘	-15.0063
+脣	-15.0063
+脯	-15.0063
+腆	-15.0063
+腋	-15.0063
+腑	-15.0063
+腱	-15.0063
+臃	-15.0063
+舀	-15.0063
+舔	-15.0063
+舺	-15.0063
+艋	-15.0063
+茉	-15.0063
+茗	-15.0063
+茯	-15.0063
+荊	-15.0063
+荖	-15.0063
+莒	-15.0063
+萁	-15.0063
+蓀	-15.0063
+蔭	-15.0063
+蔽	-15.0063
+薏	-15.0063
+藪	-15.0063
+蘑	-15.0063
+蚪	-15.0063
+蛉	-15.0063
+蛹	-15.0063
+蜍	-15.0063
+蜷	-15.0063
+蝌	-15.0063
+蝸	-15.0063
+螯	-15.0063
+螽	-15.0063
+蟑	-15.0063
+蟾	-15.0063
+蠅	-15.0063
+蠆	-15.0063
+蠋	-15.0063
+蠕	-15.0063
+蠣	-15.0063
+衊	-15.0063
+裘	-15.0063
+褟	-15.0063
+褵	-15.0063
+褶	-15.0063
+襌	-15.0063
+襖	-15.0063
+覲	-15.0063
+訐	-15.0063
+訖	-15.0063
+訥	-15.0063
+誅	-15.0063
+諳	-15.0063
+謔	-15.0063
+譟	-15.0063
+豌	-15.0063
+豺	-15.0063
+貽	-15.0063
+賁	-15.0063
+贍	-15.0063
+跚	-15.0063
+蹣	-15.0063
+蹭	-15.0063
+軋	-15.0063
+迂	-15.0063
+迄	-15.0063
+遶	-15.0063
+酚	-15.0063
+酥	-15.0063
+醱	-15.0063
+釉	-15.0063
+鈍	-15.0063
+鈐	-15.0063
+鈙	-15.0063
+鈦	-15.0063
+鈺	-15.0063
+鉀	-15.0063
+鉗	-15.0063
+鋅	-15.0063
+鋌	-15.0063
+鋤	-15.0063
+錏	-15.0063
+錐	-15.0063
+錘	-15.0063
+錠	-15.0063
+錳	-15.0063
+鍍	-15.0063
+鎘	-15.0063
+鎳	-15.0063
+鏘	-15.0063
+鐓	-15.0063
+鑄	-15.0063
+鑾	-15.0063
+闔	-15.0063
+陂	-15.0063
+陛	-15.0063
+隄	-15.0063
+隍	-15.0063
+隕	-15.0063
+隘	-15.0063
+隴	-15.0063
+隼	-15.0063
+雌	-15.0063
+靂	-15.0063
+靛	-15.0063
+靦	-15.0063
+靨	-15.0063
+鞦	-15.0063
+韁	-15.0063
+韆	-15.0063
+韜	-15.0063
+頡	-15.0063
+顎	-15.0063
+顴	-15.0063
+颳	-15.0063
+飴	-15.0063
+餒	-15.0063
+餔	-15.0063
+餽	-15.0063
+饌	-15.0063
+駢	-15.0063
+騖	-15.0063
+驍	-15.0063
+驛	-15.0063
+驢	-15.0063
+驪	-15.0063
+骯	-15.0063
+髖	-15.0063
+鬢	-15.0063
+鬩	-15.0063
+鰭	-15.0063
+鰲	-15.0063
+鱒	-15.0063
+鳧	-15.0063
+鳩	-15.0063
+鴆	-15.0063
+鴯	-15.0063
+鵜	-15.0063
+鵡	-15.0063
+鷿	-15.0063
+鸚	-15.0063
+鸞	-15.0063
+黝	-15.0063
+齪	-15.0063
+齷	-15.0063
+龕	-15.0063

ASR-model/tokenizer_seg_bpe5k_char/env.log

@@ -0,0 +1,195 @@
+SpeechBrain system description
+==============================
+Python version:
+3.8.10 (default, Jun  2 2021, 10:49:15) 
+[GCC 9.4.0]
+==============================
+Installed Python packages:
+appdirs==1.4.4
+argon2-cffi==20.1.0
+async-generator==1.10
+attrs==19.3.0
+Automat==0.8.0
+autopep8==1.5.7
+backcall==0.2.0
+backports.entry-points-selectable==1.1.0
+black==19.10b0
+bleach==3.3.1
+blessings==1.7
+blinker==1.4
+bottle==0.12.19
+certifi==2019.11.28
+cffi==1.14.6
+cfgv==3.3.0
+chardet==3.0.4
+Click==7.0
+cloud-init==21.2
+colorama==0.4.3
+command-not-found==0.3
+configobj==5.0.6
+constantly==15.1.0
+cryptography==2.8
+cupshelpers==1.0
+cycler==0.10.0
+d2l==0.16.6
+datasets==1.11.0
+dbus-python==1.2.16
+debugpy==1.3.0
+decorator==5.0.9
+defer==1.0.6
+defusedxml==0.7.1
+dill==0.3.4
+distlib==0.3.2
+distro==1.4.0
+distro-info===0.23ubuntu1
+entrypoints==0.3
+filelock==3.0.12
+flake8==3.7.9
+fsspec==2021.7.0
+gpustat==0.6.0
+gpuview==0.4.0
+httplib2==0.14.0
+huggingface-hub==0.0.16
+hyperlink==19.0.0
+HyperPyYAML==1.0.0
+identify==2.2.11
+idna==2.8
+importlib-metadata==1.5.0
+incremental==16.10.1
+ipykernel==6.0.2
+ipython==7.25.0
+ipython-genutils==0.2.0
+ipywidgets==7.6.3
+jedi==0.18.0
+Jinja2==2.10.1
+joblib==1.0.1
+jsonpatch==1.22
+jsonpointer==2.0
+jsonschema==3.2.0
+jupyter==1.0.0
+jupyter-client==6.1.12
+jupyter-console==6.4.0
+jupyter-core==4.7.1
+jupyterlab-pygments==0.1.2
+jupyterlab-widgets==1.0.0
+keyring==18.0.1
+kiwisolver==1.3.1
+language-selector==0.1
+launchpadlib==1.10.13
+lazr.restfulclient==0.14.2
+lazr.uri==1.0.3
+macaroonbakery==1.3.1
+MarkupSafe==1.1.0
+matplotlib==3.4.2
+matplotlib-inline==0.1.2
+mccabe==0.6.1
+mistune==0.8.4
+more-itertools==4.2.0
+multiprocess==0.70.12.2
+nbclient==0.5.3
+nbconvert==6.1.0
+nbformat==5.1.3
+nest-asyncio==1.5.1
+netifaces==0.10.4
+nodeenv==1.6.0
+notebook==6.4.0
+numpy==1.21.2
+nvidia-ml-py3==7.352.0
+oauthlib==3.1.0
+packaging==21.0
+pandas==1.3.0
+pandocfilters==1.4.3
+parso==0.8.2
+pathspec==0.9.0
+pexpect==4.6.0
+pickleshare==0.7.5
+Pillow==8.3.1
+platformdirs==2.0.2
+pluggy==0.13.1
+pre-commit==2.15.0
+prometheus-client==0.11.0
+prompt-toolkit==3.0.19
+protobuf==3.6.1
+psutil==5.8.0
+ptyprocess==0.7.0
+py==1.10.0
+pyarrow==5.0.0
+pyasn1==0.4.2
+pyasn1-modules==0.2.1
+pycairo==1.16.2
+pycodestyle==2.5.0
+pycparser==2.20
+pycups==1.9.73
+pyflakes==2.1.1
+Pygments==2.9.0
+PyGObject==3.36.0
+PyHamcrest==1.9.0
+PyJWT==1.7.1
+pymacaroons==0.13.0
+PyMySQL==1.0.2
+PyNaCl==1.3.0
+pyOpenSSL==19.0.0
+pyparsing==2.4.7
+pyRFC3339==1.1
+pyrsistent==0.15.5
+pyserial==3.4
+pytest==5.4.1
+python-apt==2.0.0+ubuntu0.20.4.5
+python-dateutil==2.8.2
+python-debian===0.1.36ubuntu1
+pytube==10.9.3
+pytz==2019.3
+PyYAML==5.3.1
+pyzmq==22.1.0
+qtconsole==5.1.1
+QtPy==1.9.0
+regex==2021.7.6
+requests==2.22.0
+requests-unixsocket==0.2.0
+ruamel.yaml==0.17.10
+ruamel.yaml.clib==0.2.6
+scipy==1.7.1
+screen-resolution-extra==0.0.0
+SecretStorage==2.3.1
+Send2Trash==1.7.1
+sentencepiece==0.1.96
+service-identity==18.1.0
+simplejson==3.16.0
+six==1.14.0
+sos==4.1
+-e git+https://github.com/speechbrain/speechbrain.git@1d194bfc51ae20b9e38596d220cdf0f4977e69de#egg=speechbrain
+ssh-import-id==5.10
+supervisor==4.1.0
+systemd-python==234
+terminado==0.10.1
+testpath==0.5.0
+toml==0.10.2
+torch==1.8.1
+torchaudio==0.8.1
+torchvision==0.10.0
+tornado==6.1
+tqdm==4.62.2
+traitlets==5.0.5
+Twisted==18.9.0
+typed-ast==1.4.3
+typing-extensions==3.10.0.0
+ubuntu-advantage-tools==27.2
+ufw==0.36
+unattended-upgrades==0.1
+urllib3==1.25.8
+virtualenv==20.6.0
+wadllib==1.3.3
+wcwidth==0.2.5
+webencodings==0.5.1
+widgetsnbextension==3.5.1
+xkit==0.0.0
+xxhash==2.0.2
+yamllint==1.23.0
+zipp==1.0.0
+zope.interface==4.7.1
+==============================
+Git revision:
+e3e51338
+==============================
+Cuda version:
+10.2

ASR-model/tokenizer_seg_bpe5k_char/hyperparams.yaml

@@ -0,0 +1,31 @@
+# Generated 2021-10-04 from:
+# /mnt/md0/user_wayne/speechbrain/recipes/MATBN/Tokenizer/hparams/tokenizer_seg_bpe5k_char.yaml
+# yamllint disable
+dataset_folder: /home/wayne/CORPUS/MATBN_SEG
+prepare_folder: results/prepare_seg
+output_folder: results/tokenizer_seg_bpe5k_char
+keep_unk: false
+
+token_type: char # ["unigram", "bpe", "char"]
+token_output: 5000 # index(blank/eos/bos/unk) = 0
+character_coverage: 1.0
+annotation_read: transcription
+
+train_json: results/prepare_seg/train.json
+dev_json: results/prepare_seg/dev.json
+eval_json: results/prepare_seg/eval.json
+test_json: results/prepare_seg/test.json
+
+
+tokenizer: !name:speechbrain.tokenizers.SentencePiece.SentencePiece
+  model_dir: results/tokenizer_seg_bpe5k_char
+  vocab_size: 5000
+  annotation_train: results/prepare_seg/train.json
+  annotation_read: transcription
+  model_type: char              # ["unigram", "bpe", "char"]
+  character_coverage: 1.0
+  annotation_list_to_check: [results/prepare_seg/dev.json, results/prepare_seg/eval.json,
+    results/prepare_seg/test.json]
+  annotation_format: json
+  bos_id: 1
+  eos_id: 2

ASR-model/tokenizer_seg_bpe5k_char/log.txt

@@ -0,0 +1,1037 @@
+2021-09-16 18:20:17,572 - speechbrain.core - INFO - Beginning experiment!
+2021-09-16 18:20:17,572 - speechbrain.core - INFO - Experiment folder: results/tokenizer_seg_bpe5k_char
+2021-09-16 18:20:17,773 - speechbrain.utils.superpowers - DEBUG - appdirs==1.4.4
+argon2-cffi==20.1.0
+async-generator==1.10
+attrs==21.2.0
+autopep8==1.5.7
+backcall==0.2.0
+backports.entry-points-selectable==1.1.0
+black==19.10b0
+bleach==3.3.1
+certifi==2021.5.30
+cffi==1.14.6
+cfgv==3.3.0
+charset-normalizer==2.0.4
+click==8.0.1
+cycler==0.10.0
+d2l==0.16.6
+datasets==1.11.0
+debugpy==1.3.0
+decorator==5.0.9
+defusedxml==0.7.1
+dill==0.3.4
+distlib==0.3.2
+entrypoints==0.3
+filelock==3.0.12
+flake8==3.7.9
+fsspec==2021.7.0
+huggingface-hub==0.0.16
+HyperPyYAML==1.0.0
+identify==2.2.11
+idna==3.2
+ipykernel==6.0.2
+ipython==7.25.0
+ipython-genutils==0.2.0
+ipywidgets==7.6.3
+jedi==0.18.0
+joblib==1.0.1
+jupyter==1.0.0
+jupyter-client==6.1.12
+jupyter-console==6.4.0
+jupyter-core==4.7.1
+jupyterlab-pygments==0.1.2
+jupyterlab-widgets==1.0.0
+kiwisolver==1.3.1
+matplotlib==3.4.2
+matplotlib-inline==0.1.2
+mccabe==0.6.1
+mistune==0.8.4
+mkl-fft==1.3.0
+mkl-random @ file:///tmp/build/80754af9/mkl_random_1626186064646/work
+mkl-service==2.4.0
+more-itertools==8.9.0
+multiprocess==0.70.12.2
+nbclient==0.5.3
+nbconvert==6.1.0
+nbformat==5.1.3
+nest-asyncio==1.5.1
+nodeenv==1.6.0
+notebook==6.4.0
+numpy==1.21.2
+olefile @ file:///Users/ktietz/demo/mc3/conda-bld/olefile_1629805411829/work
+packaging==21.0
+pandas==1.3.0
+pandocfilters==1.4.3
+parso==0.8.2
+pathspec==0.9.0
+pickleshare==0.7.5
+Pillow==8.3.1
+platformdirs==2.0.2
+pluggy==0.13.1
+pre-commit==2.15.0
+prometheus-client==0.11.0
+prompt-toolkit==3.0.19
+ptyprocess==0.7.0
+py==1.10.0
+pyarrow==5.0.0
+pycodestyle==2.5.0
+pycparser==2.20
+pydub @ file:///home/conda/feedstock_root/build_artifacts/pydub_1615612442567/work
+pyflakes==2.1.1
+Pygments==2.9.0
+pyparsing==2.4.7
+pytest==5.4.1
+python-dateutil==2.8.2
+pytube==10.9.3
+PyYAML==5.4.1
+pyzmq==22.1.0
+qtconsole==5.1.1
+QtPy==1.9.0
+regex==2021.7.6
+requests==2.26.0
+ruamel.yaml==0.17.10
+ruamel.yaml.clib==0.2.6
+scikit-learn @ file:///tmp/build/80754af9/scikit-learn_1621370412049/work
+scipy==1.7.1
+Send2Trash==1.7.1
+sentencepiece==0.1.96
+six @ file:///tmp/build/80754af9/six_1623709665295/work
+-e git+https://github.com/speechbrain/speechbrain.git@2ec4839746970875fc763aa354c44a3356685ef6#egg=speechbrain
+terminado==0.10.1
+testpath==0.5.0
+threadpoolctl @ file:///Users/ktietz/demo/mc3/conda-bld/threadpoolctl_1629802263681/work
+toml==0.10.2
+torch==1.8.1
+torchaudio==0.8.1
+torchvision==0.10.0
+tornado==6.1
+tqdm==4.62.2
+traitlets==5.0.5
+typed-ast==1.4.3
+typing-extensions==3.10.0.0
+urllib3==1.26.6
+virtualenv==20.6.0
+wcwidth==0.2.5
+webencodings==0.5.1
+widgetsnbextension==3.5.1
+xxhash==2.0.2
+yamllint==1.23.0
+
+
+2021-09-16 18:20:17,777 - speechbrain.utils.superpowers - DEBUG - e3e51338
+
+
+2021-09-16 18:20:19,582 - speechbrain.tokenizers.SentencePiece - INFO - Train tokenizer with type:char
+2021-09-16 18:20:19,582 - speechbrain.tokenizers.SentencePiece - INFO - Extract transcription sequences from:results/prepare_seg/train.json
+2021-09-16 18:20:19,788 - speechbrain.tokenizers.SentencePiece - INFO - Text file created at: results/prepare_seg/train.txt
+2021-09-16 18:20:20,080 - speechbrain.tokenizers.SentencePiece - INFO - ==== Loading Tokenizer ===
+2021-09-16 18:20:20,080 - speechbrain.tokenizers.SentencePiece - INFO - Tokenizer path: results/tokenizer_seg_bpe5k_char/5000_char.model
+2021-09-16 18:20:20,080 - speechbrain.tokenizers.SentencePiece - INFO - Tokenizer vocab_size: 5000
+2021-09-16 18:20:20,080 - speechbrain.tokenizers.SentencePiece - INFO - Tokenizer type: char
+2021-09-16 18:20:20,082 - speechbrain.tokenizers.SentencePiece - INFO - ==== Accuracy checking for recovering text from tokenizer ===
+2021-09-16 18:20:32,793 - speechbrain.tokenizers.SentencePiece - INFO - recover words from: results/prepare_seg/dev.json
+2021-09-16 18:20:32,793 - speechbrain.tokenizers.SentencePiece - INFO - Wrong recover words: 0
+2021-09-16 18:20:32,793 - speechbrain.tokenizers.SentencePiece - WARNING - accuracy recovering words: 1.0
+2021-09-16 18:20:32,793 - speechbrain.tokenizers.SentencePiece - INFO - ==== Accuracy checking for recovering text from tokenizer ===
+2021-09-16 18:20:45,713 - speechbrain.tokenizers.SentencePiece - INFO - recover words from: results/prepare_seg/eval.json
+2021-09-16 18:20:45,713 - speechbrain.tokenizers.SentencePiece - INFO - Wrong recover words: 0
+2021-09-16 18:20:45,714 - speechbrain.tokenizers.SentencePiece - WARNING - accuracy recovering words: 1.0
+2021-09-16 18:20:45,714 - speechbrain.tokenizers.SentencePiece - INFO - ==== Accuracy checking for recovering text from tokenizer ===
+2021-09-16 18:20:58,798 - speechbrain.tokenizers.SentencePiece - INFO - recover words from: results/prepare_seg/test.json
+2021-09-16 18:20:58,798 - speechbrain.tokenizers.SentencePiece - INFO - Wrong recover words: 0
+2021-09-16 18:20:58,798 - speechbrain.tokenizers.SentencePiece - WARNING - accuracy recovering words: 1.0
+2021-09-16 18:57:44,327 - speechbrain.core - INFO - Beginning experiment!
+2021-09-16 18:57:44,327 - speechbrain.core - INFO - Experiment folder: results/tokenizer_seg_bpe5k_char
+2021-09-16 18:57:44,526 - speechbrain.utils.superpowers - DEBUG - appdirs==1.4.4
+argon2-cffi==20.1.0
+async-generator==1.10
+attrs==21.2.0
+autopep8==1.5.7
+backcall==0.2.0
+backports.entry-points-selectable==1.1.0
+black==19.10b0
+bleach==3.3.1
+certifi==2021.5.30
+cffi==1.14.6
+cfgv==3.3.0
+charset-normalizer==2.0.4
+click==8.0.1
+cycler==0.10.0
+d2l==0.16.6
+datasets==1.11.0
+debugpy==1.3.0
+decorator==5.0.9
+defusedxml==0.7.1
+dill==0.3.4
+distlib==0.3.2
+entrypoints==0.3
+filelock==3.0.12
+flake8==3.7.9
+fsspec==2021.7.0
+huggingface-hub==0.0.16
+HyperPyYAML==1.0.0
+identify==2.2.11
+idna==3.2
+ipykernel==6.0.2
+ipython==7.25.0
+ipython-genutils==0.2.0
+ipywidgets==7.6.3
+jedi==0.18.0
+joblib==1.0.1
+jupyter==1.0.0
+jupyter-client==6.1.12
+jupyter-console==6.4.0
+jupyter-core==4.7.1
+jupyterlab-pygments==0.1.2
+jupyterlab-widgets==1.0.0
+kiwisolver==1.3.1
+matplotlib==3.4.2
+matplotlib-inline==0.1.2
+mccabe==0.6.1
+mistune==0.8.4
+mkl-fft==1.3.0
+mkl-random @ file:///tmp/build/80754af9/mkl_random_1626186064646/work
+mkl-service==2.4.0
+more-itertools==8.9.0
+multiprocess==0.70.12.2
+nbclient==0.5.3
+nbconvert==6.1.0
+nbformat==5.1.3
+nest-asyncio==1.5.1
+nodeenv==1.6.0
+notebook==6.4.0
+numpy==1.21.2
+olefile @ file:///Users/ktietz/demo/mc3/conda-bld/olefile_1629805411829/work
+packaging==21.0
+pandas==1.3.0
+pandocfilters==1.4.3
+parso==0.8.2
+pathspec==0.9.0
+pickleshare==0.7.5
+Pillow==8.3.1
+platformdirs==2.0.2
+pluggy==0.13.1
+pre-commit==2.15.0
+prometheus-client==0.11.0
+prompt-toolkit==3.0.19
+ptyprocess==0.7.0
+py==1.10.0
+pyarrow==5.0.0
+pycodestyle==2.5.0
+pycparser==2.20
+pydub @ file:///home/conda/feedstock_root/build_artifacts/pydub_1615612442567/work
+pyflakes==2.1.1
+Pygments==2.9.0
+pyparsing==2.4.7
+pytest==5.4.1
+python-dateutil==2.8.2
+pytube==10.9.3
+PyYAML==5.4.1
+pyzmq==22.1.0
+qtconsole==5.1.1
+QtPy==1.9.0
+regex==2021.7.6
+requests==2.26.0
+ruamel.yaml==0.17.10
+ruamel.yaml.clib==0.2.6
+scikit-learn @ file:///tmp/build/80754af9/scikit-learn_1621370412049/work
+scipy==1.7.1
+Send2Trash==1.7.1
+sentencepiece==0.1.96
+six @ file:///tmp/build/80754af9/six_1623709665295/work
+-e git+https://github.com/speechbrain/speechbrain.git@2ec4839746970875fc763aa354c44a3356685ef6#egg=speechbrain
+terminado==0.10.1
+testpath==0.5.0
+threadpoolctl @ file:///Users/ktietz/demo/mc3/conda-bld/threadpoolctl_1629802263681/work
+toml==0.10.2
+torch==1.8.1
+torchaudio==0.8.1
+torchvision==0.10.0
+tornado==6.1
+tqdm==4.62.2
+traitlets==5.0.5
+typed-ast==1.4.3
+typing-extensions==3.10.0.0
+urllib3==1.26.6
+virtualenv==20.6.0
+wcwidth==0.2.5
+webencodings==0.5.1
+widgetsnbextension==3.5.1
+xxhash==2.0.2
+yamllint==1.23.0
+
+
+2021-09-16 18:57:44,530 - speechbrain.utils.superpowers - DEBUG - e3e51338
+
+
+2021-09-16 18:57:46,443 - speechbrain.tokenizers.SentencePiece - INFO - Tokenizer is already trained.
+2021-09-16 18:57:46,443 - speechbrain.tokenizers.SentencePiece - INFO - ==== Loading Tokenizer ===
+2021-09-16 18:57:46,443 - speechbrain.tokenizers.SentencePiece - INFO - Tokenizer path: results/tokenizer_seg_bpe5k_char/5000_char.model
+2021-09-16 18:57:46,443 - speechbrain.tokenizers.SentencePiece - INFO - Tokenizer vocab_size: 5000
+2021-09-16 18:57:46,444 - speechbrain.tokenizers.SentencePiece - INFO - Tokenizer type: char
+2021-09-16 18:57:46,447 - speechbrain.tokenizers.SentencePiece - INFO - ==== Accuracy checking for recovering text from tokenizer ===
+2021-09-16 18:57:59,747 - speechbrain.tokenizers.SentencePiece - INFO - recover words from: results/prepare_seg/dev.json
+2021-09-16 18:57:59,747 - speechbrain.tokenizers.SentencePiece - INFO - Wrong recover words: 0
+2021-09-16 18:57:59,747 - speechbrain.tokenizers.SentencePiece - WARNING - accuracy recovering words: 1.0
+2021-09-16 18:57:59,747 - speechbrain.tokenizers.SentencePiece - INFO - ==== Accuracy checking for recovering text from tokenizer ===
+2021-09-16 18:58:12,647 - speechbrain.tokenizers.SentencePiece - INFO - recover words from: results/prepare_seg/eval.json
+2021-09-16 18:58:12,647 - speechbrain.tokenizers.SentencePiece - INFO - Wrong recover words: 0
+2021-09-16 18:58:12,647 - speechbrain.tokenizers.SentencePiece - WARNING - accuracy recovering words: 1.0
+2021-09-16 18:58:12,647 - speechbrain.tokenizers.SentencePiece - INFO - ==== Accuracy checking for recovering text from tokenizer ===
+2021-09-16 18:58:25,444 - speechbrain.tokenizers.SentencePiece - INFO - recover words from: results/prepare_seg/test.json
+2021-09-16 18:58:25,444 - speechbrain.tokenizers.SentencePiece - INFO - Wrong recover words: 0
+2021-09-16 18:58:25,444 - speechbrain.tokenizers.SentencePiece - WARNING - accuracy recovering words: 1.0
+2021-09-18 01:09:55,556 - speechbrain.core - INFO - Beginning experiment!
+2021-09-18 01:09:55,576 - speechbrain.core - INFO - Experiment folder: results/tokenizer_seg_bpe5k_char
+2021-09-18 01:09:55,858 - speechbrain.utils.superpowers - DEBUG - appdirs==1.4.4
+argon2-cffi==20.1.0
+async-generator==1.10
+attrs==21.2.0
+autopep8==1.5.7
+backcall==0.2.0
+backports.entry-points-selectable==1.1.0
+black==19.10b0
+bleach==3.3.1
+certifi==2021.5.30
+cffi==1.14.6
+cfgv==3.3.0
+charset-normalizer==2.0.4
+click==8.0.1
+cycler==0.10.0
+d2l==0.16.6
+datasets==1.11.0
+debugpy==1.3.0
+decorator==5.0.9
+defusedxml==0.7.1
+dill==0.3.4
+distlib==0.3.2
+entrypoints==0.3
+filelock==3.0.12
+flake8==3.7.9
+fsspec==2021.7.0
+huggingface-hub==0.0.16
+HyperPyYAML==1.0.0
+identify==2.2.11
+idna==3.2
+ipykernel==6.0.2
+ipython==7.25.0
+ipython-genutils==0.2.0
+ipywidgets==7.6.3
+jedi==0.18.0
+joblib==1.0.1
+jupyter==1.0.0
+jupyter-client==6.1.12
+jupyter-console==6.4.0
+jupyter-core==4.7.1
+jupyterlab-pygments==0.1.2
+jupyterlab-widgets==1.0.0
+kiwisolver==1.3.1
+matplotlib==3.4.2
+matplotlib-inline==0.1.2
+mccabe==0.6.1
+mistune==0.8.4
+mkl-fft==1.3.0
+mkl-random @ file:///tmp/build/80754af9/mkl_random_1626186064646/work
+mkl-service==2.4.0
+more-itertools==8.9.0
+multiprocess==0.70.12.2
+nbclient==0.5.3
+nbconvert==6.1.0
+nbformat==5.1.3
+nest-asyncio==1.5.1
+nodeenv==1.6.0
+notebook==6.4.0
+numpy==1.21.2
+olefile @ file:///Users/ktietz/demo/mc3/conda-bld/olefile_1629805411829/work
+packaging==21.0
+pandas==1.3.0
+pandocfilters==1.4.3
+parso==0.8.2
+pathspec==0.9.0
+pickleshare==0.7.5
+Pillow==8.3.1
+platformdirs==2.0.2
+pluggy==0.13.1
+pre-commit==2.15.0
+prometheus-client==0.11.0
+prompt-toolkit==3.0.19
+ptyprocess==0.7.0
+py==1.10.0
+pyarrow==5.0.0
+pycodestyle==2.5.0
+pycparser==2.20
+pydub @ file:///home/conda/feedstock_root/build_artifacts/pydub_1615612442567/work
+pyflakes==2.1.1
+Pygments==2.9.0
+pyparsing==2.4.7
+pytest==5.4.1
+python-dateutil==2.8.2
+pytube==10.9.3
+PyYAML==5.4.1
+pyzmq==22.1.0
+qtconsole==5.1.1
+QtPy==1.9.0
+regex==2021.7.6
+requests==2.26.0
+ruamel.yaml==0.17.10
+ruamel.yaml.clib==0.2.6
+scikit-learn @ file:///tmp/build/80754af9/scikit-learn_1621370412049/work
+scipy==1.7.1
+Send2Trash==1.7.1
+sentencepiece==0.1.96
+six @ file:///tmp/build/80754af9/six_1623709665295/work
+-e git+https://github.com/speechbrain/speechbrain.git@2ec4839746970875fc763aa354c44a3356685ef6#egg=speechbrain
+terminado==0.10.1
+testpath==0.5.0
+threadpoolctl @ file:///Users/ktietz/demo/mc3/conda-bld/threadpoolctl_1629802263681/work
+toml==0.10.2
+torch==1.8.1
+torchaudio==0.8.1
+torchvision==0.10.0
+tornado==6.1
+tqdm==4.62.2
+traitlets==5.0.5
+typed-ast==1.4.3
+typing-extensions==3.10.0.0
+urllib3==1.26.6
+virtualenv==20.6.0
+wcwidth==0.2.5
+webencodings==0.5.1
+widgetsnbextension==3.5.1
+xxhash==2.0.2
+yamllint==1.23.0
+
+
+2021-09-18 01:09:55,862 - speechbrain.utils.superpowers - DEBUG - e3e51338
+
+
+2021-09-18 01:09:57,778 - speechbrain.tokenizers.SentencePiece - INFO - Tokenizer is already trained.
+2021-09-18 01:09:57,778 - speechbrain.tokenizers.SentencePiece - INFO - ==== Loading Tokenizer ===
+2021-09-18 01:09:57,778 - speechbrain.tokenizers.SentencePiece - INFO - Tokenizer path: results/tokenizer_seg_bpe5k_char/5000_char.model
+2021-09-18 01:09:57,778 - speechbrain.tokenizers.SentencePiece - INFO - Tokenizer vocab_size: 5000
+2021-09-18 01:09:57,778 - speechbrain.tokenizers.SentencePiece - INFO - Tokenizer type: char
+2021-09-18 01:09:57,780 - speechbrain.tokenizers.SentencePiece - INFO - ==== Accuracy checking for recovering text from tokenizer ===
+2021-09-18 01:10:15,083 - speechbrain.tokenizers.SentencePiece - INFO - recover words from: results/prepare_seg/dev.json
+2021-09-18 01:10:15,083 - speechbrain.tokenizers.SentencePiece - INFO - Wrong recover words: 0
+2021-09-18 01:10:15,083 - speechbrain.tokenizers.SentencePiece - WARNING - accuracy recovering words: 1.0
+2021-09-18 01:10:15,083 - speechbrain.tokenizers.SentencePiece - INFO - ==== Accuracy checking for recovering text from tokenizer ===
+2021-09-18 01:10:32,413 - speechbrain.tokenizers.SentencePiece - INFO - recover words from: results/prepare_seg/eval.json
+2021-09-18 01:10:32,413 - speechbrain.tokenizers.SentencePiece - INFO - Wrong recover words: 0
+2021-09-18 01:10:32,413 - speechbrain.tokenizers.SentencePiece - WARNING - accuracy recovering words: 1.0
+2021-09-18 01:10:32,413 - speechbrain.tokenizers.SentencePiece - INFO - ==== Accuracy checking for recovering text from tokenizer ===
+2021-09-18 01:10:49,667 - speechbrain.tokenizers.SentencePiece - INFO - recover words from: results/prepare_seg/test.json
+2021-09-18 01:10:49,667 - speechbrain.tokenizers.SentencePiece - INFO - Wrong recover words: 0
+2021-09-18 01:10:49,667 - speechbrain.tokenizers.SentencePiece - WARNING - accuracy recovering words: 1.0
+2021-10-04 03:04:09,349 - speechbrain.core - INFO - Beginning experiment!
+2021-10-04 03:04:09,349 - speechbrain.core - INFO - Experiment folder: results/tokenizer_seg_bpe5k_char
+2021-10-04 03:04:09,986 - speechbrain.utils.superpowers - DEBUG - appdirs==1.4.4
+argon2-cffi==20.1.0
+async-generator==1.10
+attrs==19.3.0
+Automat==0.8.0
+autopep8==1.5.7
+backcall==0.2.0
+backports.entry-points-selectable==1.1.0
+black==19.10b0
+bleach==3.3.1
+blessings==1.7
+blinker==1.4
+bottle==0.12.19
+certifi==2019.11.28
+cffi==1.14.6
+cfgv==3.3.0
+chardet==3.0.4
+Click==7.0
+cloud-init==21.2
+colorama==0.4.3
+command-not-found==0.3
+configobj==5.0.6
+constantly==15.1.0
+cryptography==2.8
+cupshelpers==1.0
+cycler==0.10.0
+d2l==0.16.6
+datasets==1.11.0
+dbus-python==1.2.16
+debugpy==1.3.0
+decorator==5.0.9
+defer==1.0.6
+defusedxml==0.7.1
+dill==0.3.4
+distlib==0.3.2
+distro==1.4.0
+distro-info===0.23ubuntu1
+entrypoints==0.3
+filelock==3.0.12
+flake8==3.7.9
+fsspec==2021.7.0
+gpustat==0.6.0
+gpuview==0.4.0
+httplib2==0.14.0
+huggingface-hub==0.0.16
+hyperlink==19.0.0
+HyperPyYAML==1.0.0
+identify==2.2.11
+idna==2.8
+importlib-metadata==1.5.0
+incremental==16.10.1
+ipykernel==6.0.2
+ipython==7.25.0
+ipython-genutils==0.2.0
+ipywidgets==7.6.3
+jedi==0.18.0
+Jinja2==2.10.1
+joblib==1.0.1
+jsonpatch==1.22
+jsonpointer==2.0
+jsonschema==3.2.0
+jupyter==1.0.0
+jupyter-client==6.1.12
+jupyter-console==6.4.0
+jupyter-core==4.7.1
+jupyterlab-pygments==0.1.2
+jupyterlab-widgets==1.0.0
+keyring==18.0.1
+kiwisolver==1.3.1
+language-selector==0.1
+launchpadlib==1.10.13
+lazr.restfulclient==0.14.2
+lazr.uri==1.0.3
+macaroonbakery==1.3.1
+MarkupSafe==1.1.0
+matplotlib==3.4.2
+matplotlib-inline==0.1.2
+mccabe==0.6.1
+mistune==0.8.4
+more-itertools==4.2.0
+multiprocess==0.70.12.2
+nbclient==0.5.3
+nbconvert==6.1.0
+nbformat==5.1.3
+nest-asyncio==1.5.1
+netifaces==0.10.4
+nodeenv==1.6.0
+notebook==6.4.0
+numpy==1.21.2
+nvidia-ml-py3==7.352.0
+oauthlib==3.1.0
+packaging==21.0
+pandas==1.3.0
+pandocfilters==1.4.3
+parso==0.8.2
+pathspec==0.9.0
+pexpect==4.6.0
+pickleshare==0.7.5
+Pillow==8.3.1
+platformdirs==2.0.2
+pluggy==0.13.1
+pre-commit==2.15.0
+prometheus-client==0.11.0
+prompt-toolkit==3.0.19
+protobuf==3.6.1
+psutil==5.8.0
+ptyprocess==0.7.0
+py==1.10.0
+pyarrow==5.0.0
+pyasn1==0.4.2
+pyasn1-modules==0.2.1
+pycairo==1.16.2
+pycodestyle==2.5.0
+pycparser==2.20
+pycups==1.9.73
+pyflakes==2.1.1
+Pygments==2.9.0
+PyGObject==3.36.0
+PyHamcrest==1.9.0
+PyJWT==1.7.1
+pymacaroons==0.13.0
+PyMySQL==1.0.2
+PyNaCl==1.3.0
+pyOpenSSL==19.0.0
+pyparsing==2.4.7
+pyRFC3339==1.1
+pyrsistent==0.15.5
+pyserial==3.4
+pytest==5.4.1
+python-apt==2.0.0+ubuntu0.20.4.5
+python-dateutil==2.8.2
+python-debian===0.1.36ubuntu1
+pytube==10.9.3
+pytz==2019.3
+PyYAML==5.3.1
+pyzmq==22.1.0
+qtconsole==5.1.1
+QtPy==1.9.0
+regex==2021.7.6
+requests==2.22.0
+requests-unixsocket==0.2.0
+ruamel.yaml==0.17.10
+ruamel.yaml.clib==0.2.6
+scipy==1.7.1
+screen-resolution-extra==0.0.0
+SecretStorage==2.3.1
+Send2Trash==1.7.1
+sentencepiece==0.1.96
+service-identity==18.1.0
+simplejson==3.16.0
+six==1.14.0
+sos==4.1
+-e git+https://github.com/speechbrain/speechbrain.git@1d194bfc51ae20b9e38596d220cdf0f4977e69de#egg=speechbrain
+ssh-import-id==5.10
+supervisor==4.1.0
+systemd-python==234
+terminado==0.10.1
+testpath==0.5.0
+toml==0.10.2
+torch==1.8.1
+torchaudio==0.8.1
+torchvision==0.10.0
+tornado==6.1
+tqdm==4.62.2
+traitlets==5.0.5
+Twisted==18.9.0
+typed-ast==1.4.3
+typing-extensions==3.10.0.0
+ubuntu-advantage-tools==27.2
+ufw==0.36
+unattended-upgrades==0.1
+urllib3==1.25.8
+virtualenv==20.6.0
+wadllib==1.3.3
+wcwidth==0.2.5
+webencodings==0.5.1
+widgetsnbextension==3.5.1
+xkit==0.0.0
+xxhash==2.0.2
+yamllint==1.23.0
+zipp==1.0.0
+zope.interface==4.7.1
+
+
+2021-10-04 03:04:09,989 - speechbrain.utils.superpowers - DEBUG - e3e51338
+
+
+2021-10-04 03:04:10,195 - speechbrain.core - ERROR - Exception:
+Traceback (most recent call last):
+  File "Tokenizer/train.py", line 21, in <module>
+    run_on_main(
+  File "/home/wayne/speechbrain/speechbrain/utils/distributed.py", line 61, in run_on_main
+    func(*args, **kwargs)
+  File "/mnt/md0/user_wayne/speechbrain/recipes/MATBN/Tokenizer/matbn_prepare.py", line 67, in prepare_matbn
+    segments_info = extract_segments_info(segments_path)
+  File "/mnt/md0/user_wayne/speechbrain/recipes/MATBN/Tokenizer/matbn_prepare.py", line 117, in extract_segments_info
+    with open(segments_path, "r", encoding="utf-8") as segments_file:
+FileNotFoundError: [Errno 2] No such file or directory: '/home/wayne/CORPUS/MATBN_SEG/data/eval/segments'
+2021-10-04 03:04:50,119 - speechbrain.core - INFO - Beginning experiment!
+2021-10-04 03:04:50,119 - speechbrain.core - INFO - Experiment folder: results/tokenizer_seg_bpe5k_char
+2021-10-04 03:04:50,727 - speechbrain.utils.superpowers - DEBUG - appdirs==1.4.4
+argon2-cffi==20.1.0
+async-generator==1.10
+attrs==19.3.0
+Automat==0.8.0
+autopep8==1.5.7
+backcall==0.2.0
+backports.entry-points-selectable==1.1.0
+black==19.10b0
+bleach==3.3.1
+blessings==1.7
+blinker==1.4
+bottle==0.12.19
+certifi==2019.11.28
+cffi==1.14.6
+cfgv==3.3.0
+chardet==3.0.4
+Click==7.0
+cloud-init==21.2
+colorama==0.4.3
+command-not-found==0.3
+configobj==5.0.6
+constantly==15.1.0
+cryptography==2.8
+cupshelpers==1.0
+cycler==0.10.0
+d2l==0.16.6
+datasets==1.11.0
+dbus-python==1.2.16
+debugpy==1.3.0
+decorator==5.0.9
+defer==1.0.6
+defusedxml==0.7.1
+dill==0.3.4
+distlib==0.3.2
+distro==1.4.0
+distro-info===0.23ubuntu1
+entrypoints==0.3
+filelock==3.0.12
+flake8==3.7.9
+fsspec==2021.7.0
+gpustat==0.6.0
+gpuview==0.4.0
+httplib2==0.14.0
+huggingface-hub==0.0.16
+hyperlink==19.0.0
+HyperPyYAML==1.0.0
+identify==2.2.11
+idna==2.8
+importlib-metadata==1.5.0
+incremental==16.10.1
+ipykernel==6.0.2
+ipython==7.25.0
+ipython-genutils==0.2.0
+ipywidgets==7.6.3
+jedi==0.18.0
+Jinja2==2.10.1
+joblib==1.0.1
+jsonpatch==1.22
+jsonpointer==2.0
+jsonschema==3.2.0
+jupyter==1.0.0
+jupyter-client==6.1.12
+jupyter-console==6.4.0
+jupyter-core==4.7.1
+jupyterlab-pygments==0.1.2
+jupyterlab-widgets==1.0.0
+keyring==18.0.1
+kiwisolver==1.3.1
+language-selector==0.1
+launchpadlib==1.10.13
+lazr.restfulclient==0.14.2
+lazr.uri==1.0.3
+macaroonbakery==1.3.1
+MarkupSafe==1.1.0
+matplotlib==3.4.2
+matplotlib-inline==0.1.2
+mccabe==0.6.1
+mistune==0.8.4
+more-itertools==4.2.0
+multiprocess==0.70.12.2
+nbclient==0.5.3
+nbconvert==6.1.0
+nbformat==5.1.3
+nest-asyncio==1.5.1
+netifaces==0.10.4
+nodeenv==1.6.0
+notebook==6.4.0
+numpy==1.21.2
+nvidia-ml-py3==7.352.0
+oauthlib==3.1.0
+packaging==21.0
+pandas==1.3.0
+pandocfilters==1.4.3
+parso==0.8.2
+pathspec==0.9.0
+pexpect==4.6.0
+pickleshare==0.7.5
+Pillow==8.3.1
+platformdirs==2.0.2
+pluggy==0.13.1
+pre-commit==2.15.0
+prometheus-client==0.11.0
+prompt-toolkit==3.0.19
+protobuf==3.6.1
+psutil==5.8.0
+ptyprocess==0.7.0
+py==1.10.0
+pyarrow==5.0.0
+pyasn1==0.4.2
+pyasn1-modules==0.2.1
+pycairo==1.16.2
+pycodestyle==2.5.0
+pycparser==2.20
+pycups==1.9.73
+pyflakes==2.1.1
+Pygments==2.9.0
+PyGObject==3.36.0
+PyHamcrest==1.9.0
+PyJWT==1.7.1
+pymacaroons==0.13.0
+PyMySQL==1.0.2
+PyNaCl==1.3.0
+pyOpenSSL==19.0.0
+pyparsing==2.4.7
+pyRFC3339==1.1
+pyrsistent==0.15.5
+pyserial==3.4
+pytest==5.4.1
+python-apt==2.0.0+ubuntu0.20.4.5
+python-dateutil==2.8.2
+python-debian===0.1.36ubuntu1
+pytube==10.9.3
+pytz==2019.3
+PyYAML==5.3.1
+pyzmq==22.1.0
+qtconsole==5.1.1
+QtPy==1.9.0
+regex==2021.7.6
+requests==2.22.0
+requests-unixsocket==0.2.0
+ruamel.yaml==0.17.10
+ruamel.yaml.clib==0.2.6
+scipy==1.7.1
+screen-resolution-extra==0.0.0
+SecretStorage==2.3.1
+Send2Trash==1.7.1
+sentencepiece==0.1.96
+service-identity==18.1.0
+simplejson==3.16.0
+six==1.14.0
+sos==4.1
+-e git+https://github.com/speechbrain/speechbrain.git@1d194bfc51ae20b9e38596d220cdf0f4977e69de#egg=speechbrain
+ssh-import-id==5.10
+supervisor==4.1.0
+systemd-python==234
+terminado==0.10.1
+testpath==0.5.0
+toml==0.10.2
+torch==1.8.1
+torchaudio==0.8.1
+torchvision==0.10.0
+tornado==6.1
+tqdm==4.62.2
+traitlets==5.0.5
+Twisted==18.9.0
+typed-ast==1.4.3
+typing-extensions==3.10.0.0
+ubuntu-advantage-tools==27.2
+ufw==0.36
+unattended-upgrades==0.1
+urllib3==1.25.8
+virtualenv==20.6.0
+wadllib==1.3.3
+wcwidth==0.2.5
+webencodings==0.5.1
+widgetsnbextension==3.5.1
+xkit==0.0.0
+xxhash==2.0.2
+yamllint==1.23.0
+zipp==1.0.0
+zope.interface==4.7.1
+
+
+2021-10-04 03:04:50,730 - speechbrain.utils.superpowers - DEBUG - e3e51338
+
+
+2021-10-04 03:04:52,523 - speechbrain.tokenizers.SentencePiece - INFO - Tokenizer is already trained.
+2021-10-04 03:04:52,523 - speechbrain.tokenizers.SentencePiece - INFO - ==== Loading Tokenizer ===
+2021-10-04 03:04:52,523 - speechbrain.tokenizers.SentencePiece - INFO - Tokenizer path: results/tokenizer_seg_bpe5k_char/5000_char.model
+2021-10-04 03:04:52,523 - speechbrain.tokenizers.SentencePiece - INFO - Tokenizer vocab_size: 5000
+2021-10-04 03:04:52,523 - speechbrain.tokenizers.SentencePiece - INFO - Tokenizer type: char
+2021-10-04 03:04:52,558 - speechbrain.tokenizers.SentencePiece - INFO - ==== Accuracy checking for recovering text from tokenizer ===
+2021-10-04 03:05:04,514 - speechbrain.tokenizers.SentencePiece - INFO - recover words from: results/prepare_seg/dev.json
+2021-10-04 03:05:04,514 - speechbrain.tokenizers.SentencePiece - INFO - Wrong recover words: 0
+2021-10-04 03:05:04,514 - speechbrain.tokenizers.SentencePiece - WARNING - accuracy recovering words: 1.0
+2021-10-04 03:05:04,514 - speechbrain.tokenizers.SentencePiece - INFO - ==== Accuracy checking for recovering text from tokenizer ===
+2021-10-04 03:05:16,440 - speechbrain.tokenizers.SentencePiece - INFO - recover words from: results/prepare_seg/eval.json
+2021-10-04 03:05:16,440 - speechbrain.tokenizers.SentencePiece - INFO - Wrong recover words: 0
+2021-10-04 03:05:16,440 - speechbrain.tokenizers.SentencePiece - WARNING - accuracy recovering words: 1.0
+2021-10-04 03:05:16,440 - speechbrain.tokenizers.SentencePiece - INFO - ==== Accuracy checking for recovering text from tokenizer ===
+2021-10-04 03:05:28,162 - speechbrain.tokenizers.SentencePiece - INFO - recover words from: results/prepare_seg/test.json
+2021-10-04 03:05:28,162 - speechbrain.tokenizers.SentencePiece - INFO - Wrong recover words: 0
+2021-10-04 03:05:28,162 - speechbrain.tokenizers.SentencePiece - WARNING - accuracy recovering words: 1.0
+2021-10-04 19:38:40,556 - speechbrain.core - INFO - Beginning experiment!
+2021-10-04 19:38:40,556 - speechbrain.core - INFO - Experiment folder: results/tokenizer_seg_bpe5k_char
+2021-10-04 19:38:41,182 - speechbrain.utils.superpowers - DEBUG - appdirs==1.4.4
+argon2-cffi==20.1.0
+async-generator==1.10
+attrs==19.3.0
+Automat==0.8.0
+autopep8==1.5.7
+backcall==0.2.0
+backports.entry-points-selectable==1.1.0
+black==19.10b0
+bleach==3.3.1
+blessings==1.7
+blinker==1.4
+bottle==0.12.19
+certifi==2019.11.28
+cffi==1.14.6
+cfgv==3.3.0
+chardet==3.0.4
+Click==7.0
+cloud-init==21.2
+colorama==0.4.3
+command-not-found==0.3
+configobj==5.0.6
+constantly==15.1.0
+cryptography==2.8
+cupshelpers==1.0
+cycler==0.10.0
+d2l==0.16.6
+datasets==1.11.0
+dbus-python==1.2.16
+debugpy==1.3.0
+decorator==5.0.9
+defer==1.0.6
+defusedxml==0.7.1
+dill==0.3.4
+distlib==0.3.2
+distro==1.4.0
+distro-info===0.23ubuntu1
+entrypoints==0.3
+filelock==3.0.12
+flake8==3.7.9
+fsspec==2021.7.0
+gpustat==0.6.0
+gpuview==0.4.0
+httplib2==0.14.0
+huggingface-hub==0.0.16
+hyperlink==19.0.0
+HyperPyYAML==1.0.0
+identify==2.2.11
+idna==2.8
+importlib-metadata==1.5.0
+incremental==16.10.1
+ipykernel==6.0.2
+ipython==7.25.0
+ipython-genutils==0.2.0
+ipywidgets==7.6.3
+jedi==0.18.0
+Jinja2==2.10.1
+joblib==1.0.1
+jsonpatch==1.22
+jsonpointer==2.0
+jsonschema==3.2.0
+jupyter==1.0.0
+jupyter-client==6.1.12
+jupyter-console==6.4.0
+jupyter-core==4.7.1
+jupyterlab-pygments==0.1.2
+jupyterlab-widgets==1.0.0
+keyring==18.0.1
+kiwisolver==1.3.1
+language-selector==0.1
+launchpadlib==1.10.13
+lazr.restfulclient==0.14.2
+lazr.uri==1.0.3
+macaroonbakery==1.3.1
+MarkupSafe==1.1.0
+matplotlib==3.4.2
+matplotlib-inline==0.1.2
+mccabe==0.6.1
+mistune==0.8.4
+more-itertools==4.2.0
+multiprocess==0.70.12.2
+nbclient==0.5.3
+nbconvert==6.1.0
+nbformat==5.1.3
+nest-asyncio==1.5.1
+netifaces==0.10.4
+nodeenv==1.6.0
+notebook==6.4.0
+numpy==1.21.2
+nvidia-ml-py3==7.352.0
+oauthlib==3.1.0
+packaging==21.0
+pandas==1.3.0
+pandocfilters==1.4.3
+parso==0.8.2
+pathspec==0.9.0
+pexpect==4.6.0
+pickleshare==0.7.5
+Pillow==8.3.1
+platformdirs==2.0.2
+pluggy==0.13.1
+pre-commit==2.15.0
+prometheus-client==0.11.0
+prompt-toolkit==3.0.19
+protobuf==3.6.1
+psutil==5.8.0
+ptyprocess==0.7.0
+py==1.10.0
+pyarrow==5.0.0
+pyasn1==0.4.2
+pyasn1-modules==0.2.1
+pycairo==1.16.2
+pycodestyle==2.5.0
+pycparser==2.20
+pycups==1.9.73
+pyflakes==2.1.1
+Pygments==2.9.0
+PyGObject==3.36.0
+PyHamcrest==1.9.0
+PyJWT==1.7.1
+pymacaroons==0.13.0
+PyMySQL==1.0.2
+PyNaCl==1.3.0
+pyOpenSSL==19.0.0
+pyparsing==2.4.7
+pyRFC3339==1.1
+pyrsistent==0.15.5
+pyserial==3.4
+pytest==5.4.1
+python-apt==2.0.0+ubuntu0.20.4.5
+python-dateutil==2.8.2
+python-debian===0.1.36ubuntu1
+pytube==10.9.3
+pytz==2019.3
+PyYAML==5.3.1
+pyzmq==22.1.0
+qtconsole==5.1.1
+QtPy==1.9.0
+regex==2021.7.6
+requests==2.22.0
+requests-unixsocket==0.2.0
+ruamel.yaml==0.17.10
+ruamel.yaml.clib==0.2.6
+scipy==1.7.1
+screen-resolution-extra==0.0.0
+SecretStorage==2.3.1
+Send2Trash==1.7.1
+sentencepiece==0.1.96
+service-identity==18.1.0
+simplejson==3.16.0
+six==1.14.0
+sos==4.1
+-e git+https://github.com/speechbrain/speechbrain.git@1d194bfc51ae20b9e38596d220cdf0f4977e69de#egg=speechbrain
+ssh-import-id==5.10
+supervisor==4.1.0
+systemd-python==234
+terminado==0.10.1
+testpath==0.5.0
+toml==0.10.2
+torch==1.8.1
+torchaudio==0.8.1
+torchvision==0.10.0
+tornado==6.1
+tqdm==4.62.2
+traitlets==5.0.5
+Twisted==18.9.0
+typed-ast==1.4.3
+typing-extensions==3.10.0.0
+ubuntu-advantage-tools==27.2
+ufw==0.36
+unattended-upgrades==0.1
+urllib3==1.25.8
+virtualenv==20.6.0
+wadllib==1.3.3
+wcwidth==0.2.5
+webencodings==0.5.1
+widgetsnbextension==3.5.1
+xkit==0.0.0
+xxhash==2.0.2
+yamllint==1.23.0
+zipp==1.0.0
+zope.interface==4.7.1
+
+
+2021-10-04 19:38:41,186 - speechbrain.utils.superpowers - DEBUG - e3e51338
+
+
+2021-10-04 19:38:43,292 - speechbrain.tokenizers.SentencePiece - INFO - Tokenizer is already trained.
+2021-10-04 19:38:43,292 - speechbrain.tokenizers.SentencePiece - INFO - ==== Loading Tokenizer ===
+2021-10-04 19:38:43,292 - speechbrain.tokenizers.SentencePiece - INFO - Tokenizer path: results/tokenizer_seg_bpe5k_char/5000_char.model
+2021-10-04 19:38:43,292 - speechbrain.tokenizers.SentencePiece - INFO - Tokenizer vocab_size: 5000
+2021-10-04 19:38:43,292 - speechbrain.tokenizers.SentencePiece - INFO - Tokenizer type: char
+2021-10-04 19:38:43,294 - speechbrain.tokenizers.SentencePiece - INFO - ==== Accuracy checking for recovering text from tokenizer ===
+2021-10-04 19:38:54,904 - speechbrain.tokenizers.SentencePiece - INFO - recover words from: results/prepare_seg/dev.json
+2021-10-04 19:38:54,904 - speechbrain.tokenizers.SentencePiece - INFO - Wrong recover words: 0
+2021-10-04 19:38:54,904 - speechbrain.tokenizers.SentencePiece - WARNING - accuracy recovering words: 1.0
+2021-10-04 19:38:54,904 - speechbrain.tokenizers.SentencePiece - INFO - ==== Accuracy checking for recovering text from tokenizer ===
+2021-10-04 19:39:06,509 - speechbrain.tokenizers.SentencePiece - INFO - recover words from: results/prepare_seg/eval.json
+2021-10-04 19:39:06,510 - speechbrain.tokenizers.SentencePiece - INFO - Wrong recover words: 0
+2021-10-04 19:39:06,510 - speechbrain.tokenizers.SentencePiece - WARNING - accuracy recovering words: 1.0
+2021-10-04 19:39:06,510 - speechbrain.tokenizers.SentencePiece - INFO - ==== Accuracy checking for recovering text from tokenizer ===
+2021-10-04 19:39:18,349 - speechbrain.tokenizers.SentencePiece - INFO - recover words from: results/prepare_seg/test.json
+2021-10-04 19:39:18,349 - speechbrain.tokenizers.SentencePiece - INFO - Wrong recover words: 0
+2021-10-04 19:39:18,349 - speechbrain.tokenizers.SentencePiece - WARNING - accuracy recovering words: 1.0

ASR-model/tokenizer_seg_bpe5k_char/train.py

@@ -0,0 +1,30 @@
+import sys
+
+import speechbrain as sb
+from hyperpyyaml import load_hyperpyyaml
+from speechbrain.utils.distributed import run_on_main
+
+if __name__ == "__main__":
+    hparams_file_path, run_opts, overrides = sb.parse_arguments(sys.argv[1:])
+
+    with open(hparams_file_path) as hparams_file:
+        hparams = load_hyperpyyaml(hparams_file, overrides)
+
+    sb.create_experiment_directory(
+        experiment_directory=hparams["output_folder"],
+        hyperparams_to_save=hparams_file_path,
+        overrides=overrides,
+    )
+
+    from matbn_prepare import prepare_matbn
+
+    run_on_main(
+        prepare_matbn,
+        kwargs={
+            "dataset_folder": hparams["dataset_folder"],
+            "save_folder": hparams["prepare_folder"],
+            "keep_unk": hparams["keep_unk"],
+        },
+    )
+
+    hparams["tokenizer"]()