Fhrozen commited on
Commit
e182686
1 Parent(s): dd82ed2
README.md ADDED
@@ -0,0 +1,86 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ tags:
3
+ - espnet
4
+ - audio
5
+ - audio-source-separation
6
+ language: en
7
+ datasets:
8
+ - dns_ins20
9
+ license: cc-by-4.0
10
+ inference: false
11
+ ---
12
+
13
+ # ESPnet2 ENH pretrained model
14
+
15
+ ## `neillu23/dns_ins20_enh_train_enh_blstm_tf_raw_valid.loss.best, fs=16k, lang=en`
16
+
17
+ ♻️ Imported from <https://zenodo.org/record/4923697#.YOAOIpozZH4>.
18
+
19
+ This model was trained by neillu23 using dns_ins20 recipe in [espnet](https://github.com/espnet/espnet/).
20
+
21
+ ### Python API
22
+
23
+ ```text
24
+ See https://github.com/espnet/espnet_model_zoo
25
+ ```
26
+
27
+ ### Evaluate in the recipe
28
+
29
+ ```python
30
+ # coming soon
31
+ ```
32
+
33
+ ### Results
34
+
35
+ ```bash
36
+ # RESULTS
37
+ ## Environments
38
+ - date: `Wed Jun 9 09:49:34 CST 2021`
39
+ - python version: `3.8.10 (default, May 19 2021, 18:05:58) [GCC 7.3.0]`
40
+ - espnet version: `espnet 0.9.9`
41
+ - pytorch version: `pytorch 1.4.0`
42
+ - Git hash: `c1dfefb98bf59f654e0907b9681668eaca8ddfcc`
43
+ - Commit date: `Tue Jun 8 17:23:26 2021 +0800`
44
+
45
+
46
+ ## enh_train_enh_blstm_tf_raw
47
+
48
+ config: ./conf/tuning/train_enh_blstm_tf.yaml
49
+
50
+ |dataset|STOI|SAR|SDR|SIR|
51
+ |---|---|---|---|---|
52
+ |enhanced_cv_synthetic|0.98|23.87|23.87|0.00|
53
+ |enhanced_tt_synthetic_no_reverb|0.96|15.94|15.94|0.00|
54
+ |enhanced_tt_synthetic_with_reverb|0.84|11.86|11.86|0.00|
55
+ ```
56
+
57
+ ### Training config
58
+
59
+ See full config in [`config.yaml`](./exp/enh_train_enh_blstm_tf_raw/config.yaml)
60
+
61
+ ```yaml
62
+ config: ./conf/tuning/train_enh_blstm_tf.yaml
63
+ print_config: false
64
+ log_level: INFO
65
+ dry_run: false
66
+ iterator_type: sequence
67
+ output_dir: exp/enh_train_enh_blstm_tf_raw
68
+ ngpu: 1
69
+ seed: 0
70
+ num_workers: 4
71
+ num_att_plot: 3
72
+ dist_backend: nccl
73
+ dist_init_method: env://
74
+ dist_world_size: 2
75
+ dist_rank: 0
76
+ local_rank: 0
77
+ dist_master_addr: localhost
78
+ dist_master_port: 45398
79
+ dist_launcher: null
80
+ multiprocessing_distributed: true
81
+ unused_parameters: false
82
+ sharded_ddp: false
83
+ cudnn_enabled: true
84
+ cudnn_benchmark: false
85
+ cudnn_deterministic: true
86
+ ```
exp/enh_stats_16k/train/feats_stats.npz ADDED
Binary file
exp/enh_train_enh_blstm_tf_raw/100epoch.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d15900c767995f257821bd952592b45177165d2e049a3bc3795646b3d2d4bd6e
3
+ size 190594814
exp/enh_train_enh_blstm_tf_raw/RESULTS.md ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ <!-- Generated by ./scripts/utils/show_enh_score.sh -->
2
+ # RESULTS
3
+ ## Environments
4
+ - date: `Wed Jun 9 09:49:34 CST 2021`
5
+ - python version: `3.8.10 (default, May 19 2021, 18:05:58) [GCC 7.3.0]`
6
+ - espnet version: `espnet 0.9.9`
7
+ - pytorch version: `pytorch 1.4.0`
8
+ - Git hash: `c1dfefb98bf59f654e0907b9681668eaca8ddfcc`
9
+ - Commit date: `Tue Jun 8 17:23:26 2021 +0800`
10
+
11
+
12
+ ## enh_train_enh_blstm_tf_raw
13
+
14
+ config: ./conf/tuning/train_enh_blstm_tf.yaml
15
+
16
+ |dataset|STOI|SAR|SDR|SIR|
17
+ |---|---|---|---|---|
18
+ |enhanced_cv_synthetic|0.98|23.87|23.87|0.00|
19
+ |enhanced_tt_synthetic_no_reverb|0.96|15.94|15.94|0.00|
20
+ |enhanced_tt_synthetic_with_reverb|0.84|11.86|11.86|0.00|
21
+
exp/enh_train_enh_blstm_tf_raw/config.yaml ADDED
@@ -0,0 +1,145 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ config: ./conf/tuning/train_enh_blstm_tf.yaml
2
+ print_config: false
3
+ log_level: INFO
4
+ dry_run: false
5
+ iterator_type: sequence
6
+ output_dir: exp/enh_train_enh_blstm_tf_raw
7
+ ngpu: 1
8
+ seed: 0
9
+ num_workers: 4
10
+ num_att_plot: 3
11
+ dist_backend: nccl
12
+ dist_init_method: env://
13
+ dist_world_size: 2
14
+ dist_rank: 0
15
+ local_rank: 0
16
+ dist_master_addr: localhost
17
+ dist_master_port: 45398
18
+ dist_launcher: null
19
+ multiprocessing_distributed: true
20
+ unused_parameters: false
21
+ sharded_ddp: false
22
+ cudnn_enabled: true
23
+ cudnn_benchmark: false
24
+ cudnn_deterministic: true
25
+ collect_stats: false
26
+ write_collected_feats: false
27
+ max_epoch: 100
28
+ patience: 10
29
+ val_scheduler_criterion:
30
+ - valid
31
+ - loss
32
+ early_stopping_criterion:
33
+ - valid
34
+ - loss
35
+ - min
36
+ best_model_criterion:
37
+ - - valid
38
+ - si_snr
39
+ - max
40
+ - - valid
41
+ - loss
42
+ - min
43
+ keep_nbest_models: 1
44
+ grad_clip: 5.0
45
+ grad_clip_type: 2.0
46
+ grad_noise: false
47
+ accum_grad: 1
48
+ no_forward_run: false
49
+ resume: true
50
+ train_dtype: float32
51
+ use_amp: false
52
+ log_interval: null
53
+ use_tensorboard: true
54
+ use_wandb: false
55
+ wandb_project: null
56
+ wandb_id: null
57
+ detect_anomaly: false
58
+ pretrain_path: null
59
+ init_param: []
60
+ ignore_init_mismatch: false
61
+ freeze_param: []
62
+ num_iters_per_epoch: null
63
+ batch_size: 64
64
+ valid_batch_size: null
65
+ batch_bins: 1000000
66
+ valid_batch_bins: null
67
+ train_shape_file:
68
+ - exp/enh_stats_16k/train/speech_mix_shape
69
+ - exp/enh_stats_16k/train/speech_ref1_shape
70
+ - exp/enh_stats_16k/train/noise_ref1_shape
71
+ valid_shape_file:
72
+ - exp/enh_stats_16k/valid/speech_mix_shape
73
+ - exp/enh_stats_16k/valid/speech_ref1_shape
74
+ - exp/enh_stats_16k/valid/noise_ref1_shape
75
+ batch_type: folded
76
+ valid_batch_type: null
77
+ fold_length:
78
+ - 80000
79
+ - 80000
80
+ - 80000
81
+ sort_in_batch: descending
82
+ sort_batch: descending
83
+ multiple_iterator: false
84
+ chunk_length: 500
85
+ chunk_shift_ratio: 0.5
86
+ num_cache_chunks: 1024
87
+ train_data_path_and_name_and_type:
88
+ - - dump/raw/tr_synthetic/wav.scp
89
+ - speech_mix
90
+ - sound
91
+ - - dump/raw/tr_synthetic/spk1.scp
92
+ - speech_ref1
93
+ - sound
94
+ - - dump/raw/tr_synthetic/noise1.scp
95
+ - noise_ref1
96
+ - sound
97
+ valid_data_path_and_name_and_type:
98
+ - - dump/raw/cv_synthetic/wav.scp
99
+ - speech_mix
100
+ - sound
101
+ - - dump/raw/cv_synthetic/spk1.scp
102
+ - speech_ref1
103
+ - sound
104
+ - - dump/raw/cv_synthetic/noise1.scp
105
+ - noise_ref1
106
+ - sound
107
+ allow_variable_data_keys: false
108
+ max_cache_size: 0.0
109
+ max_cache_fd: 32
110
+ valid_max_cache_size: null
111
+ optim: adam
112
+ optim_conf:
113
+ lr: 0.001
114
+ eps: 1.0e-08
115
+ weight_decay: 1.0e-07
116
+ scheduler: reducelronplateau
117
+ scheduler_conf:
118
+ mode: min
119
+ factor: 0.7
120
+ patience: 1
121
+ init: xavier_uniform
122
+ model_conf:
123
+ loss_type: mask_mse
124
+ mask_type: psm
125
+ use_preprocessor: false
126
+ encoder: stft
127
+ encoder_conf:
128
+ n_fft: 256
129
+ hop_length: 128
130
+ separator: rnn
131
+ separator_conf:
132
+ rnn_type: blstm
133
+ num_spk: 1
134
+ nonlinear: relu
135
+ layer: 3
136
+ unit: 896
137
+ dropout: 0.5
138
+ decoder: stft
139
+ decoder_conf:
140
+ n_fft: 256
141
+ hop_length: 128
142
+ required:
143
+ - output_dir
144
+ version: 0.9.9
145
+ distributed: true
exp/enh_train_enh_blstm_tf_raw/images/backward_time.png ADDED
Binary file
exp/enh_train_enh_blstm_tf_raw/images/forward_time.png ADDED
Binary file
exp/enh_train_enh_blstm_tf_raw/images/gpu_max_cached_mem_GB.png ADDED
Binary file
exp/enh_train_enh_blstm_tf_raw/images/iter_time.png ADDED
Binary file
exp/enh_train_enh_blstm_tf_raw/images/loss.png ADDED
Binary file
exp/enh_train_enh_blstm_tf_raw/images/optim0_lr0.png ADDED
Binary file
exp/enh_train_enh_blstm_tf_raw/images/optim_step_time.png ADDED
Binary file
exp/enh_train_enh_blstm_tf_raw/images/si_snr.png ADDED
Binary file
exp/enh_train_enh_blstm_tf_raw/images/train_time.png ADDED
Binary file
meta.yaml ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
1
+ espnet: 0.9.9
2
+ files:
3
+ model_file: exp/enh_train_enh_blstm_tf_raw/100epoch.pth
4
+ python: "3.8.10 (default, May 19 2021, 18:05:58) \n[GCC 7.3.0]"
5
+ timestamp: 1623331009.137679
6
+ torch: 1.4.0
7
+ yaml_files:
8
+ train_config: exp/enh_train_enh_blstm_tf_raw/config.yaml