Hugging Face's logo

m3hrdadfi
/
hubert-large-greek-speech-emotion-recognition

Transformers
PyTorch
Greek
hubert
audio
speech
speech-emotion-recognition
Inference Endpoints
Model card Files Community
m3hrdadfi commited on
Commit
8af3d4c
•
1 Parent(s): dd9f7e9

Hello hubert

Browse files
Files changed (5) hide show
  1. config.json +93 -0
  2. preprocessor_config.json +9 -0
  3. pytorch_model.bin +3 -0
  4. test.csv +122 -0
  5. trainer_state.json +100 -0
config.json ADDED
@@ -0,0 +1,93 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "facebook/hubert-large-ll60k",
3
+ "activation_dropout": 0.0,
4
+ "apply_spec_augment": true,
5
+ "architectures": [
6
+ "HubertForSpeechClassification"
7
+ ],
8
+ "attention_dropout": 0.1,
9
+ "bos_token_id": 1,
10
+ "conv_bias": true,
11
+ "conv_dim": [
12
+ 512,
13
+ 512,
14
+ 512,
15
+ 512,
16
+ 512,
17
+ 512,
18
+ 512
19
+ ],
20
+ "conv_kernel": [
21
+ 10,
22
+ 3,
23
+ 3,
24
+ 3,
25
+ 3,
26
+ 2,
27
+ 2
28
+ ],
29
+ "conv_stride": [
30
+ 5,
31
+ 2,
32
+ 2,
33
+ 2,
34
+ 2,
35
+ 2,
36
+ 2
37
+ ],
38
+ "ctc_loss_reduction": "sum",
39
+ "ctc_zero_infinity": false,
40
+ "do_stable_layer_norm": true,
41
+ "eos_token_id": 2,
42
+ "feat_extract_activation": "gelu",
43
+ "feat_extract_dropout": 0.0,
44
+ "feat_extract_norm": "layer",
45
+ "feat_proj_dropout": 0.1,
46
+ "final_dropout": 0.0,
47
+ "finetuning_task": "wav2vec2_clf",
48
+ "gradient_checkpointing": false,
49
+ "hidden_act": "gelu",
50
+ "hidden_dropout": 0.1,
51
+ "hidden_size": 1024,
52
+ "id2label": {
53
+ "0": "anger",
54
+ "1": "disgust",
55
+ "2": "fear",
56
+ "3": "happiness",
57
+ "4": "sadness"
58
+ },
59
+ "initializer_range": 0.02,
60
+ "intermediate_size": 4096,
61
+ "label2id": {
62
+ "anger": 0,
63
+ "disgust": 1,
64
+ "fear": 2,
65
+ "happiness": 3,
66
+ "sadness": 4
67
+ },
68
+ "layer_norm_eps": 1e-05,
69
+ "layerdrop": 0.1,
70
+ "mask_channel_length": 10,
71
+ "mask_channel_min_space": 1,
72
+ "mask_channel_other": 0.0,
73
+ "mask_channel_prob": 0.0,
74
+ "mask_channel_selection": "static",
75
+ "mask_feature_length": 10,
76
+ "mask_feature_prob": 0.0,
77
+ "mask_time_length": 10,
78
+ "mask_time_min_space": 1,
79
+ "mask_time_other": 0.0,
80
+ "mask_time_prob": 0.075,
81
+ "mask_time_selection": "static",
82
+ "model_type": "hubert",
83
+ "num_attention_heads": 16,
84
+ "num_conv_pos_embedding_groups": 16,
85
+ "num_conv_pos_embeddings": 128,
86
+ "num_feat_extract_layers": 7,
87
+ "num_hidden_layers": 24,
88
+ "pad_token_id": 0,
89
+ "pooling_mode": "mean",
90
+ "problem_type": "single_label_classification",
91
+ "transformers_version": "4.7.0.dev0",
92
+ "vocab_size": 32
93
+ }
preprocessor_config.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "do_normalize": true,
3
+ "feature_extractor_type": "Wav2Vec2FeatureExtractor",
4
+ "feature_size": 1,
5
+ "padding_side": "right",
6
+ "padding_value": 0,
7
+ "return_attention_mask": true,
8
+ "sampling_rate": 16000
9
+ }
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a67e242ae3683318af3b6bbaf392128f4d058aeae49fa1d37fdc4685a0fa0a97
3
+ size 1266141741
test.csv ADDED
@@ -0,0 +1,122 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ name path emotion
2
+ d17 (2) /content/data/aesdd/disgust/d17 (2).wav disgust
3
+ s15 (2) /content/data/aesdd/sadness/s15 (2).wav sadness
4
+ a03 (2) /content/data/aesdd/anger/a03 (2).wav anger
5
+ s07 (5) /content/data/aesdd/sadness/s07 (5).wav sadness
6
+ h07 (2) /content/data/aesdd/happiness/h07 (2).wav happiness
7
+ f08 (4) /content/data/aesdd/fear/f08 (4).wav fear
8
+ f19 (1) /content/data/aesdd/fear/f19 (1).wav fear
9
+ s15 (4) /content/data/aesdd/sadness/s15 (4).wav sadness
10
+ f12 (5) /content/data/aesdd/fear/f12 (5).wav fear
11
+ s04 (6) /content/data/aesdd/sadness/s04 (6).wav sadness
12
+ a15 (4) /content/data/aesdd/anger/a15 (4).wav anger
13
+ h13 (5) /content/data/aesdd/happiness/h13 (5).wav happiness
14
+ d11 (6) /content/data/aesdd/disgust/d11 (6).wav disgust
15
+ a09 (6) /content/data/aesdd/anger/a09 (6).wav anger
16
+ h03 (1) /content/data/aesdd/happiness/h03 (1).wav happiness
17
+ s17 (6) /content/data/aesdd/sadness/s17 (6).wav sadness
18
+ d15 (5) /content/data/aesdd/disgust/d15 (5).wav disgust
19
+ s13 (4) /content/data/aesdd/sadness/s13 (4).wav sadness
20
+ f15 (3) /content/data/aesdd/fear/f15 (3).wav fear
21
+ h05 (3) /content/data/aesdd/happiness/h05 (3).wav happiness
22
+ s16 (4) /content/data/aesdd/sadness/s16 (4).wav sadness
23
+ f01 (6) /content/data/aesdd/fear/f01 (6).wav fear
24
+ d01 (1) /content/data/aesdd/disgust/d01 (1).wav disgust
25
+ d11 (5) /content/data/aesdd/disgust/d11 (5).wav disgust
26
+ a20 (3) /content/data/aesdd/anger/a20 (3).wav anger
27
+ a13 (5) /content/data/aesdd/anger/a13 (5).wav anger
28
+ a01 (2) /content/data/aesdd/anger/a01 (2).wav anger
29
+ f14 (4) /content/data/aesdd/fear/f14 (4).wav fear
30
+ a05 (3) /content/data/aesdd/anger/a05 (3).wav anger
31
+ h04 (5) /content/data/aesdd/happiness/h04 (5).wav happiness
32
+ h05 (4) /content/data/aesdd/happiness/h05 (4).wav happiness
33
+ f15 (4) /content/data/aesdd/fear/f15 (4).wav fear
34
+ h03 (3) /content/data/aesdd/happiness/h03 (3).wav happiness
35
+ f18 (4) /content/data/aesdd/fear/f18 (4).wav fear
36
+ f16 (3) /content/data/aesdd/fear/f16 (3).wav fear
37
+ d13 (5) /content/data/aesdd/disgust/d13 (5).wav disgust
38
+ a15 (2) /content/data/aesdd/anger/a15 (2).wav anger
39
+ a10 (6) /content/data/aesdd/anger/a10 (6).wav anger
40
+ a14 (1) /content/data/aesdd/anger/a14 (1).wav anger
41
+ d18 (1) /content/data/aesdd/disgust/d18 (1).wav disgust
42
+ f13 (01) /content/data/aesdd/fear/f13 (01).wav fear
43
+ h05 (1) /content/data/aesdd/happiness/h05 (1).wav happiness
44
+ h08 (4) /content/data/aesdd/happiness/h08 (4).wav happiness
45
+ f11 (01) /content/data/aesdd/fear/f11 (01).wav fear
46
+ s17 (5) /content/data/aesdd/sadness/s17 (5).wav sadness
47
+ d04 (2) /content/data/aesdd/disgust/d04 (2).wav disgust
48
+ d18 (4) /content/data/aesdd/disgust/d18 (4).wav disgust
49
+ h04 (6) /content/data/aesdd/happiness/h04 (6).wav happiness
50
+ s19 (1) /content/data/aesdd/sadness/s19 (1).wav sadness
51
+ a06 (5)b /content/data/aesdd/anger/a06 (5)b.wav anger
52
+ d14 (2) /content/data/aesdd/disgust/d14 (2).wav disgust
53
+ f09 (6) /content/data/aesdd/fear/f09 (6).wav fear
54
+ f17 (4) /content/data/aesdd/fear/f17 (4).wav fear
55
+ a06 (6) /content/data/aesdd/anger/a06 (6).wav anger
56
+ d08 (2) /content/data/aesdd/disgust/d08 (2).wav disgust
57
+ h07 (4) /content/data/aesdd/happiness/h07 (4).wav happiness
58
+ d09 (3) /content/data/aesdd/disgust/d09 (3).wav disgust
59
+ a12 (4) /content/data/aesdd/anger/a12 (4).wav anger
60
+ s07 (3) /content/data/aesdd/sadness/s07 (3).wav sadness
61
+ f16 (6) /content/data/aesdd/fear/f16 (6).wav fear
62
+ f13 (5) /content/data/aesdd/fear/f13 (5).wav fear
63
+ d17 (4) /content/data/aesdd/disgust/d17 (4).wav disgust
64
+ s07 (1) /content/data/aesdd/sadness/s07 (1).wav sadness
65
+ f02 (1) /content/data/aesdd/fear/f02 (1).wav fear
66
+ h10 (1) /content/data/aesdd/happiness/h10 (1).wav happiness
67
+ h10 (2) /content/data/aesdd/happiness/h10 (2).wav happiness
68
+ s04 (1) /content/data/aesdd/sadness/s04 (1).wav sadness
69
+ d10 (6) /content/data/aesdd/disgust/d10 (6).wav disgust
70
+ a04 (2) /content/data/aesdd/anger/a04 (2).wav anger
71
+ a18 (2) /content/data/aesdd/anger/a18 (2).wav anger
72
+ a18 (6) /content/data/aesdd/anger/a18 (6).wav anger
73
+ s12 (2) /content/data/aesdd/sadness/s12 (2).wav sadness
74
+ s15 (5) /content/data/aesdd/sadness/s15 (5).wav sadness
75
+ a08 (1) /content/data/aesdd/anger/a08 (1).wav anger
76
+ d02 (2) /content/data/aesdd/disgust/d02 (2).wav disgust
77
+ h12 (4) /content/data/aesdd/happiness/h12 (4).wav happiness
78
+ h08 (3) /content/data/aesdd/happiness/h08 (3).wav happiness
79
+ s10 (2) /content/data/aesdd/sadness/s10 (2).wav sadness
80
+ f19 (2) /content/data/aesdd/fear/f19 (2).wav fear
81
+ s10 (5) /content/data/aesdd/sadness/s10 (5).wav sadness
82
+ h09 (4) /content/data/aesdd/happiness/h09 (4).wav happiness
83
+ s13 (6) /content/data/aesdd/sadness/s13 (6).wav sadness
84
+ h18 (2) /content/data/aesdd/happiness/h18 (2).wav happiness
85
+ d03 (5) /content/data/aesdd/disgust/d03 (5).wav disgust
86
+ d14 (1) /content/data/aesdd/disgust/d14 (1).wav disgust
87
+ s02 (3) /content/data/aesdd/sadness/s02 (3).wav sadness
88
+ f14 (1) /content/data/aesdd/fear/f14 (1).wav fear
89
+ a01 (3) /content/data/aesdd/anger/a01 (3).wav anger
90
+ a07 (3) /content/data/aesdd/anger/a07 (3).wav anger
91
+ f17 (1) /content/data/aesdd/fear/f17 (1).wav fear
92
+ h14 (2) /content/data/aesdd/happiness/h14 (2).wav happiness
93
+ d01 (6) /content/data/aesdd/disgust/d01 (6).wav disgust
94
+ s01 (1) /content/data/aesdd/sadness/s01 (1).wav sadness
95
+ s17 (3) /content/data/aesdd/sadness/s17 (3).wav sadness
96
+ f15 (5) /content/data/aesdd/fear/f15 (5).wav fear
97
+ h01 (4) /content/data/aesdd/happiness/h01 (4).wav happiness
98
+ s09 (1) /content/data/aesdd/sadness/s09 (1).wav sadness
99
+ f06 (5) /content/data/aesdd/fear/f06 (5).wav fear
100
+ d05 (6) /content/data/aesdd/disgust/d05 (6).wav disgust
101
+ a08 (2) /content/data/aesdd/anger/a08 (2).wav anger
102
+ s13 (2) /content/data/aesdd/sadness/s13 (2).wav sadness
103
+ d18 (2) /content/data/aesdd/disgust/d18 (2).wav disgust
104
+ a09 (2) /content/data/aesdd/anger/a09 (2).wav anger
105
+ f05 (6) /content/data/aesdd/fear/f05 (6).wav fear
106
+ s04 (5) /content/data/aesdd/sadness/s04 (5).wav sadness
107
+ h06 (5) /content/data/aesdd/happiness/h06 (5).wav happiness
108
+ a14 (6) /content/data/aesdd/anger/a14 (6).wav anger
109
+ a13 (2) /content/data/aesdd/anger/a13 (2).wav anger
110
+ h10 (5) /content/data/aesdd/happiness/h10 (5).wav happiness
111
+ a13 (1) /content/data/aesdd/anger/a13 (1).wav anger
112
+ d12 (3) /content/data/aesdd/disgust/d12 (3).wav disgust
113
+ h18 (3) /content/data/aesdd/happiness/h18 (3).wav happiness
114
+ d15 (1) /content/data/aesdd/disgust/d15 (1).wav disgust
115
+ h19 (3) /content/data/aesdd/happiness/h19 (3).wav happiness
116
+ s05 (4) /content/data/aesdd/sadness/s05 (4).wav sadness
117
+ d17 (6) /content/data/aesdd/disgust/d17 (6).wav disgust
118
+ d12 (2) /content/data/aesdd/disgust/d12 (2).wav disgust
119
+ s06 (4) /content/data/aesdd/sadness/s06 (4).wav sadness
120
+ f19 (3) /content/data/aesdd/fear/f19 (3).wav fear
121
+ f09 (2) /content/data/aesdd/fear/f09 (2).wav fear
122
+ h08 (2) /content/data/aesdd/happiness/h08 (2).wav happiness
trainer_state.json ADDED
@@ -0,0 +1,100 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 6.991735537190083,
5
+ "global_step": 560,
6
+ "is_hyper_param_search": false,
7
+ "is_local_process_zero": true,
8
+ "is_world_process_zero": true,
9
+ "log_history": [
10
+ {
11
+ "epoch": 1.25,
12
+ "learning_rate": 0.00016428571428571428,
13
+ "loss": 1.4242,
14
+ "step": 100
15
+ },
16
+ {
17
+ "epoch": 1.25,
18
+ "eval_accuracy": 0.6776859760284424,
19
+ "eval_loss": 0.9407395124435425,
20
+ "eval_runtime": 15.4142,
21
+ "eval_samples_per_second": 7.85,
22
+ "eval_steps_per_second": 3.957,
23
+ "step": 100
24
+ },
25
+ {
26
+ "epoch": 2.5,
27
+ "learning_rate": 0.00012857142857142858,
28
+ "loss": 0.884,
29
+ "step": 200
30
+ },
31
+ {
32
+ "epoch": 2.5,
33
+ "eval_accuracy": 0.6528925895690918,
34
+ "eval_loss": 0.8700841665267944,
35
+ "eval_runtime": 23.5759,
36
+ "eval_samples_per_second": 5.132,
37
+ "eval_steps_per_second": 2.587,
38
+ "step": 200
39
+ },
40
+ {
41
+ "epoch": 3.74,
42
+ "learning_rate": 9.357142857142858e-05,
43
+ "loss": 0.5967,
44
+ "step": 300
45
+ },
46
+ {
47
+ "epoch": 3.74,
48
+ "eval_accuracy": 0.8181818127632141,
49
+ "eval_loss": 0.7293241620063782,
50
+ "eval_runtime": 15.1537,
51
+ "eval_samples_per_second": 7.985,
52
+ "eval_steps_per_second": 4.025,
53
+ "step": 300
54
+ },
55
+ {
56
+ "epoch": 4.99,
57
+ "learning_rate": 5.785714285714287e-05,
58
+ "loss": 0.3024,
59
+ "step": 400
60
+ },
61
+ {
62
+ "epoch": 4.99,
63
+ "eval_accuracy": 0.8595041036605835,
64
+ "eval_loss": 0.5951272249221802,
65
+ "eval_runtime": 23.2801,
66
+ "eval_samples_per_second": 5.198,
67
+ "eval_steps_per_second": 2.62,
68
+ "step": 400
69
+ },
70
+ {
71
+ "epoch": 6.25,
72
+ "learning_rate": 2.214285714285714e-05,
73
+ "loss": 0.1382,
74
+ "step": 500
75
+ },
76
+ {
77
+ "epoch": 6.25,
78
+ "eval_accuracy": 0.9090909361839294,
79
+ "eval_loss": 0.3326501250267029,
80
+ "eval_runtime": 23.5413,
81
+ "eval_samples_per_second": 5.14,
82
+ "eval_steps_per_second": 2.591,
83
+ "step": 500
84
+ },
85
+ {
86
+ "epoch": 6.99,
87
+ "step": 560,
88
+ "total_flos": 5.1637585435657395e+17,
89
+ "train_loss": 0.6040960873876299,
90
+ "train_runtime": 1484.0947,
91
+ "train_samples_per_second": 2.278,
92
+ "train_steps_per_second": 0.377
93
+ }
94
+ ],
95
+ "max_steps": 560,
96
+ "num_train_epochs": 7,
97
+ "total_flos": 5.1637585435657395e+17,
98
+ "trial_name": null,
99
+ "trial_params": null
100
+ }