csukuangfj
commited on
Commit
•
ca6b51a
1
Parent(s):
e4e9ed6
Add pretrained model for stateless transducer.
Browse files- data/lang_bpe_500/bpe.model +3 -0
- exp/export.sh +15 -0
- exp/log/log-train-2022-03-23-08-46-38-0 +0 -0
- exp/log/log-train-2022-03-23-08-46-38-1 +0 -0
- exp/log/log-train-2022-03-23-08-46-38-2 +0 -0
- exp/log/log-train-2022-03-23-08-46-38-3 +0 -0
- exp/log/log-train-2022-03-23-08-46-38-4 +0 -0
- exp/log/log-train-2022-03-23-08-46-38-5 +0 -0
- exp/log/log-train-2022-03-23-08-46-38-6 +0 -0
- exp/pretrained.pt +3 -0
- exp/run-ot-modified.sh +13 -0
- exp/tensorboard/events.out.tfevents.1647996398.de-74279-k2-train-7-0309102938-68688b4cbd-xhtcg.3022858.0 +3 -0
- log/errs-test-clean-beam_4-epoch-59-avg-19-beam-4.txt +0 -0
- log/errs-test-clean-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-1.txt +0 -0
- log/errs-test-clean-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-2.txt +0 -0
- log/errs-test-clean-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-3.txt +0 -0
- log/errs-test-other-beam_4-epoch-59-avg-19-beam-4.txt +0 -0
- log/errs-test-other-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-1.txt +0 -0
- log/errs-test-other-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-2.txt +0 -0
- log/errs-test-other-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-3.txt +0 -0
- log/log-decode-epoch-59-avg-19-beam-4-2022-03-27-14-53-41 +49 -0
- log/log-decode-epoch-59-avg-19-context-2-max-sym-per-frame-1-2022-03-27-14-13-30 +25 -0
- log/log-decode-epoch-59-avg-19-context-2-max-sym-per-frame-2-2022-03-27-15-20-19 +25 -0
- log/log-decode-epoch-59-avg-19-context-2-max-sym-per-frame-3-2022-03-27-15-27-37 +25 -0
- log/recogs-test-clean-beam_4-epoch-59-avg-19-beam-4.txt +0 -0
- log/recogs-test-clean-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-1.txt +0 -0
- log/recogs-test-clean-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-2.txt +0 -0
- log/recogs-test-clean-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-3.txt +0 -0
- log/recogs-test-other-beam_4-epoch-59-avg-19-beam-4.txt +0 -0
- log/recogs-test-other-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-1.txt +0 -0
- log/recogs-test-other-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-2.txt +0 -0
- log/recogs-test-other-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-3.txt +0 -0
- test_wavs/1089-134686-0001.wav +0 -0
- test_wavs/1221-135766-0001.wav +0 -0
- test_wavs/1221-135766-0002.wav +0 -0
- test_wavs/trans.txt +3 -0
data/lang_bpe_500/bpe.model
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c53433de083c4a6ad12d034550ef22de68cec62c4f58932a7b6b8b2f1e743fa5
|
3 |
+
size 244865
|
exp/export.sh
ADDED
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#!/usr/bin/env bash
|
2 |
+
|
3 |
+
set -e
|
4 |
+
|
5 |
+
export CUDA_VISIBLE_DEVICES=0
|
6 |
+
|
7 |
+
epoch=59
|
8 |
+
avg=19
|
9 |
+
|
10 |
+
./transducer_stateless/export.py \
|
11 |
+
--exp-dir ./transducer_stateless/exp-2 \
|
12 |
+
--bpe-model data/lang_bpe_500/bpe.model \
|
13 |
+
--epoch $epoch \
|
14 |
+
--avg $avg
|
15 |
+
|
exp/log/log-train-2022-03-23-08-46-38-0
ADDED
The diff for this file is too large to render.
See raw diff
|
|
exp/log/log-train-2022-03-23-08-46-38-1
ADDED
The diff for this file is too large to render.
See raw diff
|
|
exp/log/log-train-2022-03-23-08-46-38-2
ADDED
The diff for this file is too large to render.
See raw diff
|
|
exp/log/log-train-2022-03-23-08-46-38-3
ADDED
The diff for this file is too large to render.
See raw diff
|
|
exp/log/log-train-2022-03-23-08-46-38-4
ADDED
The diff for this file is too large to render.
See raw diff
|
|
exp/log/log-train-2022-03-23-08-46-38-5
ADDED
The diff for this file is too large to render.
See raw diff
|
|
exp/log/log-train-2022-03-23-08-46-38-6
ADDED
The diff for this file is too large to render.
See raw diff
|
|
exp/pretrained.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ef09ee7e27c6bc20152c85773f0555bee74a0f8cef22940be3edda0b268ceec8
|
3 |
+
size 336185613
|
exp/run-ot-modified.sh
ADDED
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#!/usr/bin/env bash
|
2 |
+
|
3 |
+
export CUDA_VISIBLE_DEVICES="1,2,3,4,5,6,7"
|
4 |
+
|
5 |
+
./transducer_stateless/train.py \
|
6 |
+
--world-size 7 \
|
7 |
+
--num-epochs 60 \
|
8 |
+
--start-epoch 0 \
|
9 |
+
--exp-dir transducer_stateless/exp-2 \
|
10 |
+
--full-libri 1 \
|
11 |
+
--max-duration 300 \
|
12 |
+
--lr-factor 5 \
|
13 |
+
--modified-transducer-prob 0.25
|
exp/tensorboard/events.out.tfevents.1647996398.de-74279-k2-train-7-0309102938-68688b4cbd-xhtcg.3022858.0
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:293a5b0d9fe5f50dbbf01cd131271d4088cd22b824e7b83966ec62bebad3ebf3
|
3 |
+
size 694759
|
log/errs-test-clean-beam_4-epoch-59-avg-19-beam-4.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
log/errs-test-clean-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-1.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
log/errs-test-clean-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-2.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
log/errs-test-clean-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-3.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
log/errs-test-other-beam_4-epoch-59-avg-19-beam-4.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
log/errs-test-other-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-1.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
log/errs-test-other-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-2.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
log/errs-test-other-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-3.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
log/log-decode-epoch-59-avg-19-beam-4-2022-03-27-14-53-41
ADDED
@@ -0,0 +1,49 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
2022-03-27 14:53:41,806 INFO [decode.py:375] Decoding started
|
2 |
+
2022-03-27 14:53:41,806 INFO [decode.py:381] Device: cuda:0
|
3 |
+
2022-03-27 14:53:41,809 INFO [decode.py:390] {'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 0, 'log_interval': 50, 'reset_interval': 200, 'valid_interval': 3000, 'feature_dim': 80, 'encoder_out_dim': 512, 'subsampling_factor': 4, 'attention_dim': 512, 'nhead': 8, 'dim_feedforward': 2048, 'num_encoder_layers': 12, 'vgg_frontend': False, 'warm_step': 80000, 'env_info': {'k2-version': '1.13', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '5ee082ea55f50e8bd42203ba266945ea5a236ab8', 'k2-git-date': 'Sun Feb 27 09:00:48 2022', 'lhotse-version': '1.0.0.dev+missing.version.file', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'master', 'icefall-git-sha1': '395a3f9-dirty', 'icefall-git-date': 'Wed Mar 23 19:11:34 2022', 'icefall-path': '/ceph-fj/fangjun/open-source-2/icefall-ot-batch-search', 'k2-path': '/ceph-fj/fangjun/open-source-2/k2-rnnt-ali/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-fj/fangjun/open-source-2/lhotse-rnnt-ali/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-3-0307202051-57dc848959-8tmmp', 'IP address': '10.177.24.138'}, 'epoch': 59, 'avg': 19, 'exp_dir': PosixPath('transducer_stateless/exp-2'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'decoding_method': 'modified_beam_search', 'beam_size': 4, 'context_size': 2, 'max_sym_per_frame': 1, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 1000, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'res_dir': PosixPath('transducer_stateless/exp-2/modified_beam_search'), 'suffix': 'epoch-59-avg-19-beam-4', 'blank_id': 0, 'vocab_size': 500}
|
4 |
+
2022-03-27 14:53:41,809 INFO [decode.py:392] About to create model
|
5 |
+
2022-03-27 14:53:42,317 INFO [decode.py:403] averaging ['transducer_stateless/exp-2/epoch-41.pt', 'transducer_stateless/exp-2/epoch-42.pt', 'transducer_stateless/exp-2/epoch-43.pt', 'transducer_stateless/exp-2/epoch-44.pt', 'transducer_stateless/exp-2/epoch-45.pt', 'transducer_stateless/exp-2/epoch-46.pt', 'transducer_stateless/exp-2/epoch-47.pt', 'transducer_stateless/exp-2/epoch-48.pt', 'transducer_stateless/exp-2/epoch-49.pt', 'transducer_stateless/exp-2/epoch-50.pt', 'transducer_stateless/exp-2/epoch-51.pt', 'transducer_stateless/exp-2/epoch-52.pt', 'transducer_stateless/exp-2/epoch-53.pt', 'transducer_stateless/exp-2/epoch-54.pt', 'transducer_stateless/exp-2/epoch-55.pt', 'transducer_stateless/exp-2/epoch-56.pt', 'transducer_stateless/exp-2/epoch-57.pt', 'transducer_stateless/exp-2/epoch-58.pt', 'transducer_stateless/exp-2/epoch-59.pt']
|
6 |
+
2022-03-27 14:53:57,944 INFO [decode.py:412] Number of model parameters: 84007924
|
7 |
+
2022-03-27 14:53:57,944 INFO [asr_datamodule.py:406] About to get test-clean cuts
|
8 |
+
2022-03-27 14:53:58,046 INFO [asr_datamodule.py:411] About to get test-other cuts
|
9 |
+
2022-03-27 14:54:06,972 INFO [decode.py:300] batch 0/?, cuts processed until now is 220
|
10 |
+
2022-03-27 14:54:21,038 INFO [decode.py:300] batch 2/?, cuts processed until now is 633
|
11 |
+
2022-03-27 14:54:36,379 INFO [decode.py:300] batch 4/?, cuts processed until now is 827
|
12 |
+
2022-03-27 14:54:49,088 INFO [decode.py:300] batch 6/?, cuts processed until now is 1034
|
13 |
+
2022-03-27 14:55:04,500 INFO [decode.py:300] batch 8/?, cuts processed until now is 1197
|
14 |
+
2022-03-27 14:55:19,967 INFO [decode.py:300] batch 10/?, cuts processed until now is 1594
|
15 |
+
2022-03-27 14:55:39,010 INFO [decode.py:300] batch 12/?, cuts processed until now is 1675
|
16 |
+
2022-03-27 14:55:55,063 INFO [decode.py:300] batch 14/?, cuts processed until now is 1911
|
17 |
+
2022-03-27 14:56:03,084 INFO [decode.py:300] batch 16/?, cuts processed until now is 2021
|
18 |
+
2022-03-27 14:56:06,593 INFO [decode.py:300] batch 18/?, cuts processed until now is 2091
|
19 |
+
2022-03-27 14:56:14,459 INFO [decode.py:300] batch 20/?, cuts processed until now is 2431
|
20 |
+
2022-03-27 14:56:26,411 INFO [decode.py:300] batch 22/?, cuts processed until now is 2550
|
21 |
+
2022-03-27 14:56:35,070 INFO [decode.py:300] batch 24/?, cuts processed until now is 2620
|
22 |
+
2022-03-27 14:56:35,201 INFO [decode.py:317] The transcripts are stored in transducer_stateless/exp-2/modified_beam_search/recogs-test-clean-beam_4-epoch-59-avg-19-beam-4.txt
|
23 |
+
2022-03-27 14:56:35,267 INFO [utils.py:406] [test-clean-beam_4] %WER 2.56% [1346 / 52576, 159 ins, 105 del, 1082 sub ]
|
24 |
+
2022-03-27 14:56:35,430 INFO [decode.py:330] Wrote detailed error stats to transducer_stateless/exp-2/modified_beam_search/errs-test-clean-beam_4-epoch-59-avg-19-beam-4.txt
|
25 |
+
2022-03-27 14:56:35,431 INFO [decode.py:347]
|
26 |
+
For test-clean, WER of different settings are:
|
27 |
+
beam_4 2.56 best for test-clean
|
28 |
+
|
29 |
+
2022-03-27 14:56:43,931 INFO [decode.py:300] batch 0/?, cuts processed until now is 245
|
30 |
+
2022-03-27 14:56:58,374 INFO [decode.py:300] batch 2/?, cuts processed until now is 708
|
31 |
+
2022-03-27 14:57:13,829 INFO [decode.py:300] batch 4/?, cuts processed until now is 931
|
32 |
+
2022-03-27 14:57:26,530 INFO [decode.py:300] batch 6/?, cuts processed until now is 1170
|
33 |
+
2022-03-27 14:57:41,932 INFO [decode.py:300] batch 8/?, cuts processed until now is 1365
|
34 |
+
2022-03-27 14:57:57,647 INFO [decode.py:300] batch 10/?, cuts processed until now is 1811
|
35 |
+
2022-03-27 14:58:17,812 INFO [decode.py:300] batch 12/?, cuts processed until now is 1904
|
36 |
+
2022-03-27 14:58:33,962 INFO [decode.py:300] batch 14/?, cuts processed until now is 2170
|
37 |
+
2022-03-27 14:58:35,223 INFO [decode.py:300] batch 16/?, cuts processed until now is 2187
|
38 |
+
2022-03-27 14:58:45,603 INFO [decode.py:300] batch 18/?, cuts processed until now is 2398
|
39 |
+
2022-03-27 14:58:46,938 INFO [decode.py:300] batch 20/?, cuts processed until now is 2448
|
40 |
+
2022-03-27 14:59:00,554 INFO [decode.py:300] batch 22/?, cuts processed until now is 2809
|
41 |
+
2022-03-27 14:59:11,297 INFO [decode.py:300] batch 24/?, cuts processed until now is 2939
|
42 |
+
2022-03-27 14:59:11,437 INFO [decode.py:317] The transcripts are stored in transducer_stateless/exp-2/modified_beam_search/recogs-test-other-beam_4-epoch-59-avg-19-beam-4.txt
|
43 |
+
2022-03-27 14:59:11,504 INFO [utils.py:406] [test-other-beam_4] %WER 6.20% [3245 / 52343, 328 ins, 294 del, 2623 sub ]
|
44 |
+
2022-03-27 14:59:11,670 INFO [decode.py:330] Wrote detailed error stats to transducer_stateless/exp-2/modified_beam_search/errs-test-other-beam_4-epoch-59-avg-19-beam-4.txt
|
45 |
+
2022-03-27 14:59:11,671 INFO [decode.py:347]
|
46 |
+
For test-other, WER of different settings are:
|
47 |
+
beam_4 6.2 best for test-other
|
48 |
+
|
49 |
+
2022-03-27 14:59:11,671 INFO [decode.py:439] Done!
|
log/log-decode-epoch-59-avg-19-context-2-max-sym-per-frame-1-2022-03-27-14-13-30
ADDED
@@ -0,0 +1,25 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
2022-03-27 14:13:30,652 INFO [decode.py:375] Decoding started
|
2 |
+
2022-03-27 14:13:30,652 INFO [decode.py:381] Device: cuda:0
|
3 |
+
2022-03-27 14:13:30,654 INFO [decode.py:390] {'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 0, 'log_interval': 50, 'reset_interval': 200, 'valid_interval': 3000, 'feature_dim': 80, 'encoder_out_dim': 512, 'subsampling_factor': 4, 'attention_dim': 512, 'nhead': 8, 'dim_feedforward': 2048, 'num_encoder_layers': 12, 'vgg_frontend': False, 'warm_step': 80000, 'env_info': {'k2-version': '1.13', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '5ee082ea55f50e8bd42203ba266945ea5a236ab8', 'k2-git-date': 'Sun Feb 27 09:00:48 2022', 'lhotse-version': '1.0.0.dev+missing.version.file', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'master', 'icefall-git-sha1': '395a3f9-dirty', 'icefall-git-date': 'Wed Mar 23 19:11:34 2022', 'icefall-path': '/ceph-fj/fangjun/open-source-2/icefall-ot-batch-search', 'k2-path': '/ceph-fj/fangjun/open-source-2/k2-rnnt-ali/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-fj/fangjun/open-source-2/lhotse-rnnt-ali/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-3-0307202051-57dc848959-8tmmp', 'IP address': '10.177.24.138'}, 'epoch': 59, 'avg': 19, 'exp_dir': PosixPath('transducer_stateless/exp-2'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'decoding_method': 'greedy_search', 'beam_size': 4, 'context_size': 2, 'max_sym_per_frame': 1, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 1000, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'res_dir': PosixPath('transducer_stateless/exp-2/greedy_search'), 'suffix': 'epoch-59-avg-19-context-2-max-sym-per-frame-1', 'blank_id': 0, 'vocab_size': 500}
|
4 |
+
2022-03-27 14:13:30,654 INFO [decode.py:392] About to create model
|
5 |
+
2022-03-27 14:13:31,158 INFO [decode.py:403] averaging ['transducer_stateless/exp-2/epoch-41.pt', 'transducer_stateless/exp-2/epoch-42.pt', 'transducer_stateless/exp-2/epoch-43.pt', 'transducer_stateless/exp-2/epoch-44.pt', 'transducer_stateless/exp-2/epoch-45.pt', 'transducer_stateless/exp-2/epoch-46.pt', 'transducer_stateless/exp-2/epoch-47.pt', 'transducer_stateless/exp-2/epoch-48.pt', 'transducer_stateless/exp-2/epoch-49.pt', 'transducer_stateless/exp-2/epoch-50.pt', 'transducer_stateless/exp-2/epoch-51.pt', 'transducer_stateless/exp-2/epoch-52.pt', 'transducer_stateless/exp-2/epoch-53.pt', 'transducer_stateless/exp-2/epoch-54.pt', 'transducer_stateless/exp-2/epoch-55.pt', 'transducer_stateless/exp-2/epoch-56.pt', 'transducer_stateless/exp-2/epoch-57.pt', 'transducer_stateless/exp-2/epoch-58.pt', 'transducer_stateless/exp-2/epoch-59.pt']
|
6 |
+
2022-03-27 14:13:50,788 INFO [decode.py:412] Number of model parameters: 84007924
|
7 |
+
2022-03-27 14:13:50,788 INFO [asr_datamodule.py:406] About to get test-clean cuts
|
8 |
+
2022-03-27 14:13:50,892 INFO [asr_datamodule.py:411] About to get test-other cuts
|
9 |
+
2022-03-27 14:13:53,237 INFO [decode.py:300] batch 0/?, cuts processed until now is 220
|
10 |
+
2022-03-27 14:14:22,335 INFO [decode.py:317] The transcripts are stored in transducer_stateless/exp-2/greedy_search/recogs-test-clean-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-1.txt
|
11 |
+
2022-03-27 14:14:22,402 INFO [utils.py:406] [test-clean-greedy_search] %WER 2.60% [1367 / 52576, 157 ins, 110 del, 1100 sub ]
|
12 |
+
2022-03-27 14:14:22,576 INFO [decode.py:330] Wrote detailed error stats to transducer_stateless/exp-2/greedy_search/errs-test-clean-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-1.txt
|
13 |
+
2022-03-27 14:14:22,577 INFO [decode.py:347]
|
14 |
+
For test-clean, WER of different settings are:
|
15 |
+
greedy_search 2.6 best for test-clean
|
16 |
+
|
17 |
+
2022-03-27 14:14:24,657 INFO [decode.py:300] batch 0/?, cuts processed until now is 245
|
18 |
+
2022-03-27 14:14:53,478 INFO [decode.py:317] The transcripts are stored in transducer_stateless/exp-2/greedy_search/recogs-test-other-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-1.txt
|
19 |
+
2022-03-27 14:14:53,568 INFO [utils.py:406] [test-other-greedy_search] %WER 6.33% [3314 / 52343, 335 ins, 320 del, 2659 sub ]
|
20 |
+
2022-03-27 14:14:53,743 INFO [decode.py:330] Wrote detailed error stats to transducer_stateless/exp-2/greedy_search/errs-test-other-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-1.txt
|
21 |
+
2022-03-27 14:14:53,744 INFO [decode.py:347]
|
22 |
+
For test-other, WER of different settings are:
|
23 |
+
greedy_search 6.33 best for test-other
|
24 |
+
|
25 |
+
2022-03-27 14:14:53,744 INFO [decode.py:439] Done!
|
log/log-decode-epoch-59-avg-19-context-2-max-sym-per-frame-2-2022-03-27-15-20-19
ADDED
@@ -0,0 +1,25 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
2022-03-27 15:20:19,337 INFO [decode.py:375] Decoding started
|
2 |
+
2022-03-27 15:20:19,337 INFO [decode.py:381] Device: cuda:0
|
3 |
+
2022-03-27 15:20:19,339 INFO [decode.py:390] {'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 0, 'log_interval': 50, 'reset_interval': 200, 'valid_interval': 3000, 'feature_dim': 80, 'encoder_out_dim': 512, 'subsampling_factor': 4, 'attention_dim': 512, 'nhead': 8, 'dim_feedforward': 2048, 'num_encoder_layers': 12, 'vgg_frontend': False, 'warm_step': 80000, 'env_info': {'k2-version': '1.13', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '5ee082ea55f50e8bd42203ba266945ea5a236ab8', 'k2-git-date': 'Sun Feb 27 09:00:48 2022', 'lhotse-version': '1.0.0.dev+missing.version.file', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'master', 'icefall-git-sha1': '395a3f9-dirty', 'icefall-git-date': 'Wed Mar 23 19:11:34 2022', 'icefall-path': '/ceph-fj/fangjun/open-source-2/icefall-ot-batch-search', 'k2-path': '/ceph-fj/fangjun/open-source-2/k2-rnnt-ali/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-fj/fangjun/open-source-2/lhotse-rnnt-ali/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-3-0307202051-57dc848959-8tmmp', 'IP address': '10.177.24.138'}, 'epoch': 59, 'avg': 19, 'exp_dir': PosixPath('transducer_stateless/exp-2'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'decoding_method': 'greedy_search', 'beam_size': 4, 'context_size': 2, 'max_sym_per_frame': 2, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 1000, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'res_dir': PosixPath('transducer_stateless/exp-2/greedy_search'), 'suffix': 'epoch-59-avg-19-context-2-max-sym-per-frame-2', 'blank_id': 0, 'vocab_size': 500}
|
4 |
+
2022-03-27 15:20:19,340 INFO [decode.py:392] About to create model
|
5 |
+
2022-03-27 15:20:19,840 INFO [decode.py:403] averaging ['transducer_stateless/exp-2/epoch-41.pt', 'transducer_stateless/exp-2/epoch-42.pt', 'transducer_stateless/exp-2/epoch-43.pt', 'transducer_stateless/exp-2/epoch-44.pt', 'transducer_stateless/exp-2/epoch-45.pt', 'transducer_stateless/exp-2/epoch-46.pt', 'transducer_stateless/exp-2/epoch-47.pt', 'transducer_stateless/exp-2/epoch-48.pt', 'transducer_stateless/exp-2/epoch-49.pt', 'transducer_stateless/exp-2/epoch-50.pt', 'transducer_stateless/exp-2/epoch-51.pt', 'transducer_stateless/exp-2/epoch-52.pt', 'transducer_stateless/exp-2/epoch-53.pt', 'transducer_stateless/exp-2/epoch-54.pt', 'transducer_stateless/exp-2/epoch-55.pt', 'transducer_stateless/exp-2/epoch-56.pt', 'transducer_stateless/exp-2/epoch-57.pt', 'transducer_stateless/exp-2/epoch-58.pt', 'transducer_stateless/exp-2/epoch-59.pt']
|
6 |
+
2022-03-27 15:20:35,399 INFO [decode.py:412] Number of model parameters: 84007924
|
7 |
+
2022-03-27 15:20:35,399 INFO [asr_datamodule.py:406] About to get test-clean cuts
|
8 |
+
2022-03-27 15:20:35,500 INFO [asr_datamodule.py:411] About to get test-other cuts
|
9 |
+
2022-03-27 15:20:41,821 INFO [decode.py:300] batch 0/?, cuts processed until now is 220
|
10 |
+
2022-03-27 15:22:25,607 INFO [decode.py:317] The transcripts are stored in transducer_stateless/exp-2/greedy_search/recogs-test-clean-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-2.txt
|
11 |
+
2022-03-27 15:22:25,674 INFO [utils.py:406] [test-clean-greedy_search] %WER 2.60% [1367 / 52576, 157 ins, 110 del, 1100 sub ]
|
12 |
+
2022-03-27 15:22:25,841 INFO [decode.py:330] Wrote detailed error stats to transducer_stateless/exp-2/greedy_search/errs-test-clean-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-2.txt
|
13 |
+
2022-03-27 15:22:25,841 INFO [decode.py:347]
|
14 |
+
For test-clean, WER of different settings are:
|
15 |
+
greedy_search 2.6 best for test-clean
|
16 |
+
|
17 |
+
2022-03-27 15:22:31,949 INFO [decode.py:300] batch 0/?, cuts processed until now is 245
|
18 |
+
2022-03-27 15:24:12,888 INFO [decode.py:317] The transcripts are stored in transducer_stateless/exp-2/greedy_search/recogs-test-other-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-2.txt
|
19 |
+
2022-03-27 15:24:12,955 INFO [utils.py:406] [test-other-greedy_search] %WER 6.32% [3306 / 52343, 335 ins, 317 del, 2654 sub ]
|
20 |
+
2022-03-27 15:24:13,123 INFO [decode.py:330] Wrote detailed error stats to transducer_stateless/exp-2/greedy_search/errs-test-other-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-2.txt
|
21 |
+
2022-03-27 15:24:13,124 INFO [decode.py:347]
|
22 |
+
For test-other, WER of different settings are:
|
23 |
+
greedy_search 6.32 best for test-other
|
24 |
+
|
25 |
+
2022-03-27 15:24:13,125 INFO [decode.py:439] Done!
|
log/log-decode-epoch-59-avg-19-context-2-max-sym-per-frame-3-2022-03-27-15-27-37
ADDED
@@ -0,0 +1,25 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
2022-03-27 15:27:37,288 INFO [decode.py:375] Decoding started
|
2 |
+
2022-03-27 15:27:37,288 INFO [decode.py:381] Device: cuda:0
|
3 |
+
2022-03-27 15:27:37,290 INFO [decode.py:390] {'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 0, 'log_interval': 50, 'reset_interval': 200, 'valid_interval': 3000, 'feature_dim': 80, 'encoder_out_dim': 512, 'subsampling_factor': 4, 'attention_dim': 512, 'nhead': 8, 'dim_feedforward': 2048, 'num_encoder_layers': 12, 'vgg_frontend': False, 'warm_step': 80000, 'env_info': {'k2-version': '1.13', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '5ee082ea55f50e8bd42203ba266945ea5a236ab8', 'k2-git-date': 'Sun Feb 27 09:00:48 2022', 'lhotse-version': '1.0.0.dev+missing.version.file', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'master', 'icefall-git-sha1': '395a3f9-dirty', 'icefall-git-date': 'Wed Mar 23 19:11:34 2022', 'icefall-path': '/ceph-fj/fangjun/open-source-2/icefall-ot-batch-search', 'k2-path': '/ceph-fj/fangjun/open-source-2/k2-rnnt-ali/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-fj/fangjun/open-source-2/lhotse-rnnt-ali/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-3-0307202051-57dc848959-8tmmp', 'IP address': '10.177.24.138'}, 'epoch': 59, 'avg': 19, 'exp_dir': PosixPath('transducer_stateless/exp-2'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'decoding_method': 'greedy_search', 'beam_size': 4, 'context_size': 2, 'max_sym_per_frame': 3, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 1000, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'res_dir': PosixPath('transducer_stateless/exp-2/greedy_search'), 'suffix': 'epoch-59-avg-19-context-2-max-sym-per-frame-3', 'blank_id': 0, 'vocab_size': 500}
|
4 |
+
2022-03-27 15:27:37,290 INFO [decode.py:392] About to create model
|
5 |
+
2022-03-27 15:27:37,788 INFO [decode.py:403] averaging ['transducer_stateless/exp-2/epoch-41.pt', 'transducer_stateless/exp-2/epoch-42.pt', 'transducer_stateless/exp-2/epoch-43.pt', 'transducer_stateless/exp-2/epoch-44.pt', 'transducer_stateless/exp-2/epoch-45.pt', 'transducer_stateless/exp-2/epoch-46.pt', 'transducer_stateless/exp-2/epoch-47.pt', 'transducer_stateless/exp-2/epoch-48.pt', 'transducer_stateless/exp-2/epoch-49.pt', 'transducer_stateless/exp-2/epoch-50.pt', 'transducer_stateless/exp-2/epoch-51.pt', 'transducer_stateless/exp-2/epoch-52.pt', 'transducer_stateless/exp-2/epoch-53.pt', 'transducer_stateless/exp-2/epoch-54.pt', 'transducer_stateless/exp-2/epoch-55.pt', 'transducer_stateless/exp-2/epoch-56.pt', 'transducer_stateless/exp-2/epoch-57.pt', 'transducer_stateless/exp-2/epoch-58.pt', 'transducer_stateless/exp-2/epoch-59.pt']
|
6 |
+
2022-03-27 15:27:53,677 INFO [decode.py:412] Number of model parameters: 84007924
|
7 |
+
2022-03-27 15:27:53,677 INFO [asr_datamodule.py:406] About to get test-clean cuts
|
8 |
+
2022-03-27 15:27:53,780 INFO [asr_datamodule.py:411] About to get test-other cuts
|
9 |
+
2022-03-27 15:28:00,130 INFO [decode.py:300] batch 0/?, cuts processed until now is 220
|
10 |
+
2022-03-27 15:29:44,793 INFO [decode.py:317] The transcripts are stored in transducer_stateless/exp-2/greedy_search/recogs-test-clean-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-3.txt
|
11 |
+
2022-03-27 15:29:44,860 INFO [utils.py:406] [test-clean-greedy_search] %WER 2.60% [1367 / 52576, 157 ins, 110 del, 1100 sub ]
|
12 |
+
2022-03-27 15:29:45,025 INFO [decode.py:330] Wrote detailed error stats to transducer_stateless/exp-2/greedy_search/errs-test-clean-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-3.txt
|
13 |
+
2022-03-27 15:29:45,025 INFO [decode.py:347]
|
14 |
+
For test-clean, WER of different settings are:
|
15 |
+
greedy_search 2.6 best for test-clean
|
16 |
+
|
17 |
+
2022-03-27 15:29:51,251 INFO [decode.py:300] batch 0/?, cuts processed until now is 245
|
18 |
+
2022-03-27 15:31:33,871 INFO [decode.py:317] The transcripts are stored in transducer_stateless/exp-2/greedy_search/recogs-test-other-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-3.txt
|
19 |
+
2022-03-27 15:31:33,939 INFO [utils.py:406] [test-other-greedy_search] %WER 6.32% [3306 / 52343, 335 ins, 317 del, 2654 sub ]
|
20 |
+
2022-03-27 15:31:34,109 INFO [decode.py:330] Wrote detailed error stats to transducer_stateless/exp-2/greedy_search/errs-test-other-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-3.txt
|
21 |
+
2022-03-27 15:31:34,110 INFO [decode.py:347]
|
22 |
+
For test-other, WER of different settings are:
|
23 |
+
greedy_search 6.32 best for test-other
|
24 |
+
|
25 |
+
2022-03-27 15:31:34,110 INFO [decode.py:439] Done!
|
log/recogs-test-clean-beam_4-epoch-59-avg-19-beam-4.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
log/recogs-test-clean-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-1.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
log/recogs-test-clean-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-2.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
log/recogs-test-clean-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-3.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
log/recogs-test-other-beam_4-epoch-59-avg-19-beam-4.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
log/recogs-test-other-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-1.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
log/recogs-test-other-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-2.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
log/recogs-test-other-greedy_search-epoch-59-avg-19-context-2-max-sym-per-frame-3.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|
test_wavs/1089-134686-0001.wav
ADDED
Binary file (212 kB). View file
|
|
test_wavs/1221-135766-0001.wav
ADDED
Binary file (535 kB). View file
|
|
test_wavs/1221-135766-0002.wav
ADDED
Binary file (154 kB). View file
|
|
test_wavs/trans.txt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
1089-134686-0001 AFTER EARLY NIGHTFALL THE YELLOW LAMPS WOULD LIGHT UP HERE AND THERE THE SQUALID QUARTER OF THE BROTHELS
|
2 |
+
1221-135766-0001 GOD AS A DIRECT CONSEQUENCE OF THE SIN WHICH MAN THUS PUNISHED HAD GIVEN HER A LOVELY CHILD WHOSE PLACE WAS ON THAT SAME DISHONOURED BOSOM TO CONNECT HER PARENT FOR EVER WITH THE RACE AND DESCENT OF MORTALS AND TO BE FINALLY A BLESSED SOUL IN HEAVEN
|
3 |
+
1221-135766-0002 YET THESE THOUGHTS AFFECTED HESTER PRYNNE LESS WITH HOPE THAN APPREHENSION
|