Hugging Face's logo

Bhaveen
/
epoch_musical_instruments_identification

PyTorch
wav2vec2
Model card Files
xet
Community
Bhaveen commited on
Commit
54d7bfa
·
verified ·
1 Parent(s): 1d0750e

Upload folder using huggingface_hub

Browse files
Files changed (12) hide show
  1. checkpoint-720/config.json +131 -0
  2. checkpoint-720/optimizer.pt +3 -0
  3. checkpoint-720/preprocessor_config.json +9 -0
  4. checkpoint-720/pytorch_model.bin +3 -0
  5. checkpoint-720/rng_state.pth +3 -0
  6. checkpoint-720/scheduler.pt +3 -0
  7. checkpoint-720/trainer_state.json +4356 -0
  8. checkpoint-720/training_args.bin +3 -0
  9. config.json +131 -0
  10. preprocessor_config.json +9 -0
  11. pytorch_model.bin +3 -0
  12. training_args.bin +3 -0
checkpoint-720/config.json ADDED
@@ -0,0 +1,131 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "facebook/wav2vec2-base-960h",
3
+ "activation_dropout": 0.1,
4
+ "adapter_kernel_size": 3,
5
+ "adapter_stride": 2,
6
+ "add_adapter": false,
7
+ "apply_spec_augment": true,
8
+ "architectures": [
9
+ "Wav2Vec2ForSequenceClassification"
10
+ ],
11
+ "attention_dropout": 0.1,
12
+ "bos_token_id": 1,
13
+ "classifier_proj_size": 256,
14
+ "codevector_dim": 256,
15
+ "contrastive_logits_temperature": 0.1,
16
+ "conv_bias": false,
17
+ "conv_dim": [
18
+ 512,
19
+ 512,
20
+ 512,
21
+ 512,
22
+ 512,
23
+ 512,
24
+ 512
25
+ ],
26
+ "conv_kernel": [
27
+ 10,
28
+ 3,
29
+ 3,
30
+ 3,
31
+ 3,
32
+ 2,
33
+ 2
34
+ ],
35
+ "conv_stride": [
36
+ 5,
37
+ 2,
38
+ 2,
39
+ 2,
40
+ 2,
41
+ 2,
42
+ 2
43
+ ],
44
+ "ctc_loss_reduction": "sum",
45
+ "ctc_zero_infinity": false,
46
+ "diversity_loss_weight": 0.1,
47
+ "do_stable_layer_norm": false,
48
+ "eos_token_id": 2,
49
+ "feat_extract_activation": "gelu",
50
+ "feat_extract_dropout": 0.0,
51
+ "feat_extract_norm": "group",
52
+ "feat_proj_dropout": 0.1,
53
+ "feat_quantizer_dropout": 0.0,
54
+ "final_dropout": 0.1,
55
+ "gradient_checkpointing": false,
56
+ "hidden_act": "gelu",
57
+ "hidden_dropout": 0.1,
58
+ "hidden_dropout_prob": 0.1,
59
+ "hidden_size": 768,
60
+ "id2label": {
61
+ "0": "Acoustic_Guitar",
62
+ "1": "Bass_Guitar",
63
+ "2": "Drum_set",
64
+ "3": "Electro_Guitar",
65
+ "4": "flute",
66
+ "5": "Hi_Hats",
67
+ "6": "Keyboard",
68
+ "7": "Trumpet",
69
+ "8": "Violin"
70
+ },
71
+ "initializer_range": 0.02,
72
+ "intermediate_size": 3072,
73
+ "label2id": {
74
+ "LABEL_0": 0,
75
+ "LABEL_1": 1,
76
+ "LABEL_2": 2,
77
+ "LABEL_3": 3,
78
+ "LABEL_4": 4,
79
+ "LABEL_5": 5,
80
+ "LABEL_6": 6,
81
+ "LABEL_7": 7,
82
+ "LABEL_8": 8
83
+ },
84
+ "layer_norm_eps": 1e-05,
85
+ "layerdrop": 0.1,
86
+ "mask_feature_length": 10,
87
+ "mask_feature_min_masks": 0,
88
+ "mask_feature_prob": 0.0,
89
+ "mask_time_length": 10,
90
+ "mask_time_min_masks": 2,
91
+ "mask_time_prob": 0.05,
92
+ "model_type": "wav2vec2",
93
+ "num_adapter_layers": 3,
94
+ "num_attention_heads": 12,
95
+ "num_codevector_groups": 2,
96
+ "num_codevectors_per_group": 320,
97
+ "num_conv_pos_embedding_groups": 16,
98
+ "num_conv_pos_embeddings": 128,
99
+ "num_feat_extract_layers": 7,
100
+ "num_hidden_layers": 12,
101
+ "num_negatives": 100,
102
+ "output_hidden_size": 768,
103
+ "pad_token_id": 0,
104
+ "proj_codevector_dim": 256,
105
+ "tdnn_dilation": [
106
+ 1,
107
+ 2,
108
+ 3,
109
+ 1,
110
+ 1
111
+ ],
112
+ "tdnn_dim": [
113
+ 512,
114
+ 512,
115
+ 512,
116
+ 512,
117
+ 1500
118
+ ],
119
+ "tdnn_kernel": [
120
+ 5,
121
+ 3,
122
+ 3,
123
+ 1,
124
+ 1
125
+ ],
126
+ "torch_dtype": "float32",
127
+ "transformers_version": "4.28.1",
128
+ "use_weighted_layer_sum": false,
129
+ "vocab_size": 32,
130
+ "xvector_output_dim": 512
131
+ }
checkpoint-720/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:857533b30f3b22216d4cda977fabe42a90842346780df6b9a46b71a132060f04
3
+ size 756693690
checkpoint-720/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": 0.0,
7
+ "return_attention_mask": false,
8
+ "sampling_rate": 16000
9
+ }
checkpoint-720/pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:529b4dcf7630acfa3889e1a4d06f86c9168b131ff064132ec44c1a3222b8ff2e
3
+ size 378358054
checkpoint-720/rng_state.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59fc09be6858fe09668606be579d4f40c8aa28b04c59c008768520f69c719cfb
3
+ size 14372
checkpoint-720/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9de366a0d80182d5daa97a4be9f3d51bf8dc72e03c63ec1e956193c792addf13
3
+ size 1064
checkpoint-720/trainer_state.json ADDED
@@ -0,0 +1,4356 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": 1.7899529933929443,
3
+ "best_model_checkpoint": "epoch_musical_instruments_identification/checkpoint-720",
4
+ "epoch": 2.0,
5
+ "global_step": 720,
6
+ "is_hyper_param_search": false,
7
+ "is_local_process_zero": true,
8
+ "is_world_process_zero": true,
9
+ "log_history": [
10
+ {
11
+ "epoch": 0.0,
12
+ "learning_rate": 1.0000000000000001e-07,
13
+ "loss": 2.1798,
14
+ "step": 1
15
+ },
16
+ {
17
+ "epoch": 0.01,
18
+ "learning_rate": 2.0000000000000002e-07,
19
+ "loss": 2.1949,
20
+ "step": 2
21
+ },
22
+ {
23
+ "epoch": 0.01,
24
+ "learning_rate": 3.0000000000000004e-07,
25
+ "loss": 2.1421,
26
+ "step": 3
27
+ },
28
+ {
29
+ "epoch": 0.01,
30
+ "learning_rate": 4.0000000000000003e-07,
31
+ "loss": 2.2,
32
+ "step": 4
33
+ },
34
+ {
35
+ "epoch": 0.01,
36
+ "learning_rate": 5.000000000000001e-07,
37
+ "loss": 2.1972,
38
+ "step": 5
39
+ },
40
+ {
41
+ "epoch": 0.02,
42
+ "learning_rate": 6.000000000000001e-07,
43
+ "loss": 2.1965,
44
+ "step": 6
45
+ },
46
+ {
47
+ "epoch": 0.02,
48
+ "learning_rate": 7.000000000000001e-07,
49
+ "loss": 2.1932,
50
+ "step": 7
51
+ },
52
+ {
53
+ "epoch": 0.02,
54
+ "learning_rate": 8.000000000000001e-07,
55
+ "loss": 2.1834,
56
+ "step": 8
57
+ },
58
+ {
59
+ "epoch": 0.03,
60
+ "learning_rate": 9.000000000000001e-07,
61
+ "loss": 2.2029,
62
+ "step": 9
63
+ },
64
+ {
65
+ "epoch": 0.03,
66
+ "learning_rate": 1.0000000000000002e-06,
67
+ "loss": 2.169,
68
+ "step": 10
69
+ },
70
+ {
71
+ "epoch": 0.03,
72
+ "learning_rate": 1.1e-06,
73
+ "loss": 2.198,
74
+ "step": 11
75
+ },
76
+ {
77
+ "epoch": 0.03,
78
+ "learning_rate": 1.2000000000000002e-06,
79
+ "loss": 2.1692,
80
+ "step": 12
81
+ },
82
+ {
83
+ "epoch": 0.04,
84
+ "learning_rate": 1.3e-06,
85
+ "loss": 2.1486,
86
+ "step": 13
87
+ },
88
+ {
89
+ "epoch": 0.04,
90
+ "learning_rate": 1.4000000000000001e-06,
91
+ "loss": 2.2162,
92
+ "step": 14
93
+ },
94
+ {
95
+ "epoch": 0.04,
96
+ "learning_rate": 1.5e-06,
97
+ "loss": 2.1936,
98
+ "step": 15
99
+ },
100
+ {
101
+ "epoch": 0.04,
102
+ "learning_rate": 1.6000000000000001e-06,
103
+ "loss": 2.1845,
104
+ "step": 16
105
+ },
106
+ {
107
+ "epoch": 0.05,
108
+ "learning_rate": 1.7000000000000002e-06,
109
+ "loss": 2.1702,
110
+ "step": 17
111
+ },
112
+ {
113
+ "epoch": 0.05,
114
+ "learning_rate": 1.8000000000000001e-06,
115
+ "loss": 2.1505,
116
+ "step": 18
117
+ },
118
+ {
119
+ "epoch": 0.05,
120
+ "learning_rate": 1.9000000000000002e-06,
121
+ "loss": 2.1599,
122
+ "step": 19
123
+ },
124
+ {
125
+ "epoch": 0.06,
126
+ "learning_rate": 2.0000000000000003e-06,
127
+ "loss": 2.1775,
128
+ "step": 20
129
+ },
130
+ {
131
+ "epoch": 0.06,
132
+ "learning_rate": 2.1000000000000002e-06,
133
+ "loss": 2.1894,
134
+ "step": 21
135
+ },
136
+ {
137
+ "epoch": 0.06,
138
+ "learning_rate": 2.2e-06,
139
+ "loss": 2.2096,
140
+ "step": 22
141
+ },
142
+ {
143
+ "epoch": 0.06,
144
+ "learning_rate": 2.3000000000000004e-06,
145
+ "loss": 2.1961,
146
+ "step": 23
147
+ },
148
+ {
149
+ "epoch": 0.07,
150
+ "learning_rate": 2.4000000000000003e-06,
151
+ "loss": 2.178,
152
+ "step": 24
153
+ },
154
+ {
155
+ "epoch": 0.07,
156
+ "learning_rate": 2.5e-06,
157
+ "loss": 2.1424,
158
+ "step": 25
159
+ },
160
+ {
161
+ "epoch": 0.07,
162
+ "learning_rate": 2.6e-06,
163
+ "loss": 2.1992,
164
+ "step": 26
165
+ },
166
+ {
167
+ "epoch": 0.07,
168
+ "learning_rate": 2.7000000000000004e-06,
169
+ "loss": 2.1574,
170
+ "step": 27
171
+ },
172
+ {
173
+ "epoch": 0.08,
174
+ "learning_rate": 2.8000000000000003e-06,
175
+ "loss": 2.2056,
176
+ "step": 28
177
+ },
178
+ {
179
+ "epoch": 0.08,
180
+ "learning_rate": 2.9e-06,
181
+ "loss": 2.1986,
182
+ "step": 29
183
+ },
184
+ {
185
+ "epoch": 0.08,
186
+ "learning_rate": 3e-06,
187
+ "loss": 2.193,
188
+ "step": 30
189
+ },
190
+ {
191
+ "epoch": 0.09,
192
+ "learning_rate": 3.1000000000000004e-06,
193
+ "loss": 2.2004,
194
+ "step": 31
195
+ },
196
+ {
197
+ "epoch": 0.09,
198
+ "learning_rate": 3.2000000000000003e-06,
199
+ "loss": 2.1936,
200
+ "step": 32
201
+ },
202
+ {
203
+ "epoch": 0.09,
204
+ "learning_rate": 3.3000000000000006e-06,
205
+ "loss": 2.191,
206
+ "step": 33
207
+ },
208
+ {
209
+ "epoch": 0.09,
210
+ "learning_rate": 3.4000000000000005e-06,
211
+ "loss": 2.1971,
212
+ "step": 34
213
+ },
214
+ {
215
+ "epoch": 0.1,
216
+ "learning_rate": 3.5e-06,
217
+ "loss": 2.1668,
218
+ "step": 35
219
+ },
220
+ {
221
+ "epoch": 0.1,
222
+ "learning_rate": 3.6000000000000003e-06,
223
+ "loss": 2.1731,
224
+ "step": 36
225
+ },
226
+ {
227
+ "epoch": 0.1,
228
+ "learning_rate": 3.7e-06,
229
+ "loss": 2.2396,
230
+ "step": 37
231
+ },
232
+ {
233
+ "epoch": 0.11,
234
+ "learning_rate": 3.8000000000000005e-06,
235
+ "loss": 2.1974,
236
+ "step": 38
237
+ },
238
+ {
239
+ "epoch": 0.11,
240
+ "learning_rate": 3.900000000000001e-06,
241
+ "loss": 2.1959,
242
+ "step": 39
243
+ },
244
+ {
245
+ "epoch": 0.11,
246
+ "learning_rate": 4.000000000000001e-06,
247
+ "loss": 2.1969,
248
+ "step": 40
249
+ },
250
+ {
251
+ "epoch": 0.11,
252
+ "learning_rate": 4.1e-06,
253
+ "loss": 2.1947,
254
+ "step": 41
255
+ },
256
+ {
257
+ "epoch": 0.12,
258
+ "learning_rate": 4.2000000000000004e-06,
259
+ "loss": 2.2025,
260
+ "step": 42
261
+ },
262
+ {
263
+ "epoch": 0.12,
264
+ "learning_rate": 4.3e-06,
265
+ "loss": 2.1958,
266
+ "step": 43
267
+ },
268
+ {
269
+ "epoch": 0.12,
270
+ "learning_rate": 4.4e-06,
271
+ "loss": 2.1837,
272
+ "step": 44
273
+ },
274
+ {
275
+ "epoch": 0.12,
276
+ "learning_rate": 4.5e-06,
277
+ "loss": 2.156,
278
+ "step": 45
279
+ },
280
+ {
281
+ "epoch": 0.13,
282
+ "learning_rate": 4.600000000000001e-06,
283
+ "loss": 2.182,
284
+ "step": 46
285
+ },
286
+ {
287
+ "epoch": 0.13,
288
+ "learning_rate": 4.7e-06,
289
+ "loss": 2.2051,
290
+ "step": 47
291
+ },
292
+ {
293
+ "epoch": 0.13,
294
+ "learning_rate": 4.800000000000001e-06,
295
+ "loss": 2.1977,
296
+ "step": 48
297
+ },
298
+ {
299
+ "epoch": 0.14,
300
+ "learning_rate": 4.9000000000000005e-06,
301
+ "loss": 2.1975,
302
+ "step": 49
303
+ },
304
+ {
305
+ "epoch": 0.14,
306
+ "learning_rate": 5e-06,
307
+ "loss": 2.1796,
308
+ "step": 50
309
+ },
310
+ {
311
+ "epoch": 0.14,
312
+ "learning_rate": 4.992537313432837e-06,
313
+ "loss": 2.1908,
314
+ "step": 51
315
+ },
316
+ {
317
+ "epoch": 0.14,
318
+ "learning_rate": 4.985074626865672e-06,
319
+ "loss": 2.1862,
320
+ "step": 52
321
+ },
322
+ {
323
+ "epoch": 0.15,
324
+ "learning_rate": 4.977611940298508e-06,
325
+ "loss": 2.1912,
326
+ "step": 53
327
+ },
328
+ {
329
+ "epoch": 0.15,
330
+ "learning_rate": 4.970149253731344e-06,
331
+ "loss": 2.1712,
332
+ "step": 54
333
+ },
334
+ {
335
+ "epoch": 0.15,
336
+ "learning_rate": 4.9626865671641796e-06,
337
+ "loss": 2.1952,
338
+ "step": 55
339
+ },
340
+ {
341
+ "epoch": 0.16,
342
+ "learning_rate": 4.955223880597016e-06,
343
+ "loss": 2.1987,
344
+ "step": 56
345
+ },
346
+ {
347
+ "epoch": 0.16,
348
+ "learning_rate": 4.947761194029851e-06,
349
+ "loss": 2.1515,
350
+ "step": 57
351
+ },
352
+ {
353
+ "epoch": 0.16,
354
+ "learning_rate": 4.940298507462687e-06,
355
+ "loss": 2.1981,
356
+ "step": 58
357
+ },
358
+ {
359
+ "epoch": 0.16,
360
+ "learning_rate": 4.9328358208955226e-06,
361
+ "loss": 2.1974,
362
+ "step": 59
363
+ },
364
+ {
365
+ "epoch": 0.17,
366
+ "learning_rate": 4.925373134328359e-06,
367
+ "loss": 2.1728,
368
+ "step": 60
369
+ },
370
+ {
371
+ "epoch": 0.17,
372
+ "learning_rate": 4.917910447761194e-06,
373
+ "loss": 2.1944,
374
+ "step": 61
375
+ },
376
+ {
377
+ "epoch": 0.17,
378
+ "learning_rate": 4.91044776119403e-06,
379
+ "loss": 2.1841,
380
+ "step": 62
381
+ },
382
+ {
383
+ "epoch": 0.17,
384
+ "learning_rate": 4.9029850746268655e-06,
385
+ "loss": 2.1788,
386
+ "step": 63
387
+ },
388
+ {
389
+ "epoch": 0.18,
390
+ "learning_rate": 4.895522388059702e-06,
391
+ "loss": 2.1788,
392
+ "step": 64
393
+ },
394
+ {
395
+ "epoch": 0.18,
396
+ "learning_rate": 4.888059701492538e-06,
397
+ "loss": 2.1885,
398
+ "step": 65
399
+ },
400
+ {
401
+ "epoch": 0.18,
402
+ "learning_rate": 4.880597014925373e-06,
403
+ "loss": 2.2013,
404
+ "step": 66
405
+ },
406
+ {
407
+ "epoch": 0.19,
408
+ "learning_rate": 4.873134328358209e-06,
409
+ "loss": 2.1819,
410
+ "step": 67
411
+ },
412
+ {
413
+ "epoch": 0.19,
414
+ "learning_rate": 4.865671641791045e-06,
415
+ "loss": 2.1918,
416
+ "step": 68
417
+ },
418
+ {
419
+ "epoch": 0.19,
420
+ "learning_rate": 4.858208955223881e-06,
421
+ "loss": 2.1814,
422
+ "step": 69
423
+ },
424
+ {
425
+ "epoch": 0.19,
426
+ "learning_rate": 4.850746268656717e-06,
427
+ "loss": 2.1669,
428
+ "step": 70
429
+ },
430
+ {
431
+ "epoch": 0.2,
432
+ "learning_rate": 4.843283582089552e-06,
433
+ "loss": 2.1769,
434
+ "step": 71
435
+ },
436
+ {
437
+ "epoch": 0.2,
438
+ "learning_rate": 4.8358208955223885e-06,
439
+ "loss": 2.0641,
440
+ "step": 72
441
+ },
442
+ {
443
+ "epoch": 0.2,
444
+ "learning_rate": 4.828358208955225e-06,
445
+ "loss": 2.1581,
446
+ "step": 73
447
+ },
448
+ {
449
+ "epoch": 0.21,
450
+ "learning_rate": 4.82089552238806e-06,
451
+ "loss": 2.2104,
452
+ "step": 74
453
+ },
454
+ {
455
+ "epoch": 0.21,
456
+ "learning_rate": 4.813432835820896e-06,
457
+ "loss": 2.1883,
458
+ "step": 75
459
+ },
460
+ {
461
+ "epoch": 0.21,
462
+ "learning_rate": 4.805970149253732e-06,
463
+ "loss": 2.18,
464
+ "step": 76
465
+ },
466
+ {
467
+ "epoch": 0.21,
468
+ "learning_rate": 4.798507462686568e-06,
469
+ "loss": 2.1975,
470
+ "step": 77
471
+ },
472
+ {
473
+ "epoch": 0.22,
474
+ "learning_rate": 4.791044776119403e-06,
475
+ "loss": 2.1636,
476
+ "step": 78
477
+ },
478
+ {
479
+ "epoch": 0.22,
480
+ "learning_rate": 4.783582089552239e-06,
481
+ "loss": 2.1418,
482
+ "step": 79
483
+ },
484
+ {
485
+ "epoch": 0.22,
486
+ "learning_rate": 4.7761194029850745e-06,
487
+ "loss": 2.1028,
488
+ "step": 80
489
+ },
490
+ {
491
+ "epoch": 0.23,
492
+ "learning_rate": 4.768656716417911e-06,
493
+ "loss": 2.2149,
494
+ "step": 81
495
+ },
496
+ {
497
+ "epoch": 0.23,
498
+ "learning_rate": 4.761194029850746e-06,
499
+ "loss": 2.2023,
500
+ "step": 82
501
+ },
502
+ {
503
+ "epoch": 0.23,
504
+ "learning_rate": 4.753731343283582e-06,
505
+ "loss": 2.157,
506
+ "step": 83
507
+ },
508
+ {
509
+ "epoch": 0.23,
510
+ "learning_rate": 4.746268656716418e-06,
511
+ "loss": 2.1837,
512
+ "step": 84
513
+ },
514
+ {
515
+ "epoch": 0.24,
516
+ "learning_rate": 4.738805970149254e-06,
517
+ "loss": 2.2039,
518
+ "step": 85
519
+ },
520
+ {
521
+ "epoch": 0.24,
522
+ "learning_rate": 4.73134328358209e-06,
523
+ "loss": 2.1747,
524
+ "step": 86
525
+ },
526
+ {
527
+ "epoch": 0.24,
528
+ "learning_rate": 4.723880597014926e-06,
529
+ "loss": 2.1713,
530
+ "step": 87
531
+ },
532
+ {
533
+ "epoch": 0.24,
534
+ "learning_rate": 4.716417910447761e-06,
535
+ "loss": 2.2014,
536
+ "step": 88
537
+ },
538
+ {
539
+ "epoch": 0.25,
540
+ "learning_rate": 4.7089552238805975e-06,
541
+ "loss": 2.16,
542
+ "step": 89
543
+ },
544
+ {
545
+ "epoch": 0.25,
546
+ "learning_rate": 4.701492537313434e-06,
547
+ "loss": 2.1637,
548
+ "step": 90
549
+ },
550
+ {
551
+ "epoch": 0.25,
552
+ "learning_rate": 4.694029850746269e-06,
553
+ "loss": 2.1522,
554
+ "step": 91
555
+ },
556
+ {
557
+ "epoch": 0.26,
558
+ "learning_rate": 4.686567164179105e-06,
559
+ "loss": 2.1512,
560
+ "step": 92
561
+ },
562
+ {
563
+ "epoch": 0.26,
564
+ "learning_rate": 4.6791044776119405e-06,
565
+ "loss": 2.1399,
566
+ "step": 93
567
+ },
568
+ {
569
+ "epoch": 0.26,
570
+ "learning_rate": 4.671641791044777e-06,
571
+ "loss": 2.1774,
572
+ "step": 94
573
+ },
574
+ {
575
+ "epoch": 0.26,
576
+ "learning_rate": 4.664179104477613e-06,
577
+ "loss": 2.1562,
578
+ "step": 95
579
+ },
580
+ {
581
+ "epoch": 0.27,
582
+ "learning_rate": 4.656716417910448e-06,
583
+ "loss": 2.1495,
584
+ "step": 96
585
+ },
586
+ {
587
+ "epoch": 0.27,
588
+ "learning_rate": 4.649253731343284e-06,
589
+ "loss": 2.1463,
590
+ "step": 97
591
+ },
592
+ {
593
+ "epoch": 0.27,
594
+ "learning_rate": 4.64179104477612e-06,
595
+ "loss": 2.0873,
596
+ "step": 98
597
+ },
598
+ {
599
+ "epoch": 0.28,
600
+ "learning_rate": 4.634328358208956e-06,
601
+ "loss": 2.2128,
602
+ "step": 99
603
+ },
604
+ {
605
+ "epoch": 0.28,
606
+ "learning_rate": 4.626865671641791e-06,
607
+ "loss": 2.1939,
608
+ "step": 100
609
+ },
610
+ {
611
+ "epoch": 0.28,
612
+ "learning_rate": 4.619402985074627e-06,
613
+ "loss": 2.1955,
614
+ "step": 101
615
+ },
616
+ {
617
+ "epoch": 0.28,
618
+ "learning_rate": 4.611940298507463e-06,
619
+ "loss": 2.1607,
620
+ "step": 102
621
+ },
622
+ {
623
+ "epoch": 0.29,
624
+ "learning_rate": 4.604477611940299e-06,
625
+ "loss": 2.1646,
626
+ "step": 103
627
+ },
628
+ {
629
+ "epoch": 0.29,
630
+ "learning_rate": 4.597014925373134e-06,
631
+ "loss": 2.209,
632
+ "step": 104
633
+ },
634
+ {
635
+ "epoch": 0.29,
636
+ "learning_rate": 4.58955223880597e-06,
637
+ "loss": 2.1192,
638
+ "step": 105
639
+ },
640
+ {
641
+ "epoch": 0.29,
642
+ "learning_rate": 4.5820895522388065e-06,
643
+ "loss": 2.1559,
644
+ "step": 106
645
+ },
646
+ {
647
+ "epoch": 0.3,
648
+ "learning_rate": 4.574626865671642e-06,
649
+ "loss": 2.1715,
650
+ "step": 107
651
+ },
652
+ {
653
+ "epoch": 0.3,
654
+ "learning_rate": 4.567164179104478e-06,
655
+ "loss": 2.1405,
656
+ "step": 108
657
+ },
658
+ {
659
+ "epoch": 0.3,
660
+ "learning_rate": 4.559701492537314e-06,
661
+ "loss": 2.1512,
662
+ "step": 109
663
+ },
664
+ {
665
+ "epoch": 0.31,
666
+ "learning_rate": 4.5522388059701495e-06,
667
+ "loss": 2.0992,
668
+ "step": 110
669
+ },
670
+ {
671
+ "epoch": 0.31,
672
+ "learning_rate": 4.544776119402986e-06,
673
+ "loss": 2.1689,
674
+ "step": 111
675
+ },
676
+ {
677
+ "epoch": 0.31,
678
+ "learning_rate": 4.537313432835822e-06,
679
+ "loss": 2.192,
680
+ "step": 112
681
+ },
682
+ {
683
+ "epoch": 0.31,
684
+ "learning_rate": 4.529850746268657e-06,
685
+ "loss": 2.1623,
686
+ "step": 113
687
+ },
688
+ {
689
+ "epoch": 0.32,
690
+ "learning_rate": 4.522388059701493e-06,
691
+ "loss": 2.1558,
692
+ "step": 114
693
+ },
694
+ {
695
+ "epoch": 0.32,
696
+ "learning_rate": 4.514925373134329e-06,
697
+ "loss": 2.1688,
698
+ "step": 115
699
+ },
700
+ {
701
+ "epoch": 0.32,
702
+ "learning_rate": 4.507462686567165e-06,
703
+ "loss": 2.1712,
704
+ "step": 116
705
+ },
706
+ {
707
+ "epoch": 0.33,
708
+ "learning_rate": 4.5e-06,
709
+ "loss": 2.1972,
710
+ "step": 117
711
+ },
712
+ {
713
+ "epoch": 0.33,
714
+ "learning_rate": 4.492537313432836e-06,
715
+ "loss": 2.1291,
716
+ "step": 118
717
+ },
718
+ {
719
+ "epoch": 0.33,
720
+ "learning_rate": 4.485074626865672e-06,
721
+ "loss": 2.1888,
722
+ "step": 119
723
+ },
724
+ {
725
+ "epoch": 0.33,
726
+ "learning_rate": 4.477611940298508e-06,
727
+ "loss": 2.1274,
728
+ "step": 120
729
+ },
730
+ {
731
+ "epoch": 0.34,
732
+ "learning_rate": 4.470149253731343e-06,
733
+ "loss": 2.15,
734
+ "step": 121
735
+ },
736
+ {
737
+ "epoch": 0.34,
738
+ "learning_rate": 4.462686567164179e-06,
739
+ "loss": 2.1682,
740
+ "step": 122
741
+ },
742
+ {
743
+ "epoch": 0.34,
744
+ "learning_rate": 4.4552238805970154e-06,
745
+ "loss": 2.1079,
746
+ "step": 123
747
+ },
748
+ {
749
+ "epoch": 0.34,
750
+ "learning_rate": 4.447761194029851e-06,
751
+ "loss": 2.1104,
752
+ "step": 124
753
+ },
754
+ {
755
+ "epoch": 0.35,
756
+ "learning_rate": 4.440298507462687e-06,
757
+ "loss": 2.1409,
758
+ "step": 125
759
+ },
760
+ {
761
+ "epoch": 0.35,
762
+ "learning_rate": 4.432835820895523e-06,
763
+ "loss": 2.1611,
764
+ "step": 126
765
+ },
766
+ {
767
+ "epoch": 0.35,
768
+ "learning_rate": 4.4253731343283584e-06,
769
+ "loss": 2.1206,
770
+ "step": 127
771
+ },
772
+ {
773
+ "epoch": 0.36,
774
+ "learning_rate": 4.417910447761195e-06,
775
+ "loss": 2.1372,
776
+ "step": 128
777
+ },
778
+ {
779
+ "epoch": 0.36,
780
+ "learning_rate": 4.41044776119403e-06,
781
+ "loss": 2.1301,
782
+ "step": 129
783
+ },
784
+ {
785
+ "epoch": 0.36,
786
+ "learning_rate": 4.402985074626866e-06,
787
+ "loss": 2.1106,
788
+ "step": 130
789
+ },
790
+ {
791
+ "epoch": 0.36,
792
+ "learning_rate": 4.395522388059702e-06,
793
+ "loss": 2.1445,
794
+ "step": 131
795
+ },
796
+ {
797
+ "epoch": 0.37,
798
+ "learning_rate": 4.388059701492538e-06,
799
+ "loss": 2.1115,
800
+ "step": 132
801
+ },
802
+ {
803
+ "epoch": 0.37,
804
+ "learning_rate": 4.380597014925374e-06,
805
+ "loss": 2.1086,
806
+ "step": 133
807
+ },
808
+ {
809
+ "epoch": 0.37,
810
+ "learning_rate": 4.373134328358209e-06,
811
+ "loss": 2.1346,
812
+ "step": 134
813
+ },
814
+ {
815
+ "epoch": 0.38,
816
+ "learning_rate": 4.365671641791045e-06,
817
+ "loss": 2.1469,
818
+ "step": 135
819
+ },
820
+ {
821
+ "epoch": 0.38,
822
+ "learning_rate": 4.358208955223881e-06,
823
+ "loss": 2.1265,
824
+ "step": 136
825
+ },
826
+ {
827
+ "epoch": 0.38,
828
+ "learning_rate": 4.350746268656717e-06,
829
+ "loss": 2.1215,
830
+ "step": 137
831
+ },
832
+ {
833
+ "epoch": 0.38,
834
+ "learning_rate": 4.343283582089552e-06,
835
+ "loss": 2.1903,
836
+ "step": 138
837
+ },
838
+ {
839
+ "epoch": 0.39,
840
+ "learning_rate": 4.335820895522388e-06,
841
+ "loss": 2.1719,
842
+ "step": 139
843
+ },
844
+ {
845
+ "epoch": 0.39,
846
+ "learning_rate": 4.3283582089552236e-06,
847
+ "loss": 2.0859,
848
+ "step": 140
849
+ },
850
+ {
851
+ "epoch": 0.39,
852
+ "learning_rate": 4.32089552238806e-06,
853
+ "loss": 2.1733,
854
+ "step": 141
855
+ },
856
+ {
857
+ "epoch": 0.39,
858
+ "learning_rate": 4.313432835820896e-06,
859
+ "loss": 2.185,
860
+ "step": 142
861
+ },
862
+ {
863
+ "epoch": 0.4,
864
+ "learning_rate": 4.305970149253731e-06,
865
+ "loss": 2.1999,
866
+ "step": 143
867
+ },
868
+ {
869
+ "epoch": 0.4,
870
+ "learning_rate": 4.298507462686567e-06,
871
+ "loss": 2.1753,
872
+ "step": 144
873
+ },
874
+ {
875
+ "epoch": 0.4,
876
+ "learning_rate": 4.2910447761194036e-06,
877
+ "loss": 2.0894,
878
+ "step": 145
879
+ },
880
+ {
881
+ "epoch": 0.41,
882
+ "learning_rate": 4.283582089552239e-06,
883
+ "loss": 2.1322,
884
+ "step": 146
885
+ },
886
+ {
887
+ "epoch": 0.41,
888
+ "learning_rate": 4.276119402985075e-06,
889
+ "loss": 2.1918,
890
+ "step": 147
891
+ },
892
+ {
893
+ "epoch": 0.41,
894
+ "learning_rate": 4.268656716417911e-06,
895
+ "loss": 2.0494,
896
+ "step": 148
897
+ },
898
+ {
899
+ "epoch": 0.41,
900
+ "learning_rate": 4.2611940298507466e-06,
901
+ "loss": 2.1328,
902
+ "step": 149
903
+ },
904
+ {
905
+ "epoch": 0.42,
906
+ "learning_rate": 4.253731343283583e-06,
907
+ "loss": 2.1002,
908
+ "step": 150
909
+ },
910
+ {
911
+ "epoch": 0.42,
912
+ "learning_rate": 4.246268656716419e-06,
913
+ "loss": 2.1895,
914
+ "step": 151
915
+ },
916
+ {
917
+ "epoch": 0.42,
918
+ "learning_rate": 4.238805970149254e-06,
919
+ "loss": 2.1706,
920
+ "step": 152
921
+ },
922
+ {
923
+ "epoch": 0.42,
924
+ "learning_rate": 4.23134328358209e-06,
925
+ "loss": 2.1581,
926
+ "step": 153
927
+ },
928
+ {
929
+ "epoch": 0.43,
930
+ "learning_rate": 4.223880597014926e-06,
931
+ "loss": 2.1329,
932
+ "step": 154
933
+ },
934
+ {
935
+ "epoch": 0.43,
936
+ "learning_rate": 4.216417910447762e-06,
937
+ "loss": 2.0279,
938
+ "step": 155
939
+ },
940
+ {
941
+ "epoch": 0.43,
942
+ "learning_rate": 4.208955223880597e-06,
943
+ "loss": 2.0756,
944
+ "step": 156
945
+ },
946
+ {
947
+ "epoch": 0.44,
948
+ "learning_rate": 4.201492537313433e-06,
949
+ "loss": 2.1489,
950
+ "step": 157
951
+ },
952
+ {
953
+ "epoch": 0.44,
954
+ "learning_rate": 4.194029850746269e-06,
955
+ "loss": 2.0792,
956
+ "step": 158
957
+ },
958
+ {
959
+ "epoch": 0.44,
960
+ "learning_rate": 4.186567164179105e-06,
961
+ "loss": 2.1095,
962
+ "step": 159
963
+ },
964
+ {
965
+ "epoch": 0.44,
966
+ "learning_rate": 4.17910447761194e-06,
967
+ "loss": 2.1122,
968
+ "step": 160
969
+ },
970
+ {
971
+ "epoch": 0.45,
972
+ "learning_rate": 4.171641791044776e-06,
973
+ "loss": 2.1414,
974
+ "step": 161
975
+ },
976
+ {
977
+ "epoch": 0.45,
978
+ "learning_rate": 4.1641791044776125e-06,
979
+ "loss": 2.1077,
980
+ "step": 162
981
+ },
982
+ {
983
+ "epoch": 0.45,
984
+ "learning_rate": 4.156716417910448e-06,
985
+ "loss": 2.04,
986
+ "step": 163
987
+ },
988
+ {
989
+ "epoch": 0.46,
990
+ "learning_rate": 4.149253731343284e-06,
991
+ "loss": 2.0884,
992
+ "step": 164
993
+ },
994
+ {
995
+ "epoch": 0.46,
996
+ "learning_rate": 4.141791044776119e-06,
997
+ "loss": 2.1742,
998
+ "step": 165
999
+ },
1000
+ {
1001
+ "epoch": 0.46,
1002
+ "learning_rate": 4.1343283582089555e-06,
1003
+ "loss": 2.1544,
1004
+ "step": 166
1005
+ },
1006
+ {
1007
+ "epoch": 0.46,
1008
+ "learning_rate": 4.126865671641792e-06,
1009
+ "loss": 2.1542,
1010
+ "step": 167
1011
+ },
1012
+ {
1013
+ "epoch": 0.47,
1014
+ "learning_rate": 4.119402985074627e-06,
1015
+ "loss": 2.149,
1016
+ "step": 168
1017
+ },
1018
+ {
1019
+ "epoch": 0.47,
1020
+ "learning_rate": 4.111940298507463e-06,
1021
+ "loss": 2.1546,
1022
+ "step": 169
1023
+ },
1024
+ {
1025
+ "epoch": 0.47,
1026
+ "learning_rate": 4.104477611940299e-06,
1027
+ "loss": 2.1103,
1028
+ "step": 170
1029
+ },
1030
+ {
1031
+ "epoch": 0.47,
1032
+ "learning_rate": 4.097014925373135e-06,
1033
+ "loss": 2.1966,
1034
+ "step": 171
1035
+ },
1036
+ {
1037
+ "epoch": 0.48,
1038
+ "learning_rate": 4.089552238805971e-06,
1039
+ "loss": 2.1643,
1040
+ "step": 172
1041
+ },
1042
+ {
1043
+ "epoch": 0.48,
1044
+ "learning_rate": 4.082089552238806e-06,
1045
+ "loss": 2.1643,
1046
+ "step": 173
1047
+ },
1048
+ {
1049
+ "epoch": 0.48,
1050
+ "learning_rate": 4.074626865671642e-06,
1051
+ "loss": 2.0986,
1052
+ "step": 174
1053
+ },
1054
+ {
1055
+ "epoch": 0.49,
1056
+ "learning_rate": 4.067164179104478e-06,
1057
+ "loss": 2.116,
1058
+ "step": 175
1059
+ },
1060
+ {
1061
+ "epoch": 0.49,
1062
+ "learning_rate": 4.059701492537314e-06,
1063
+ "loss": 2.0683,
1064
+ "step": 176
1065
+ },
1066
+ {
1067
+ "epoch": 0.49,
1068
+ "learning_rate": 4.052238805970149e-06,
1069
+ "loss": 2.1523,
1070
+ "step": 177
1071
+ },
1072
+ {
1073
+ "epoch": 0.49,
1074
+ "learning_rate": 4.044776119402985e-06,
1075
+ "loss": 2.1446,
1076
+ "step": 178
1077
+ },
1078
+ {
1079
+ "epoch": 0.5,
1080
+ "learning_rate": 4.037313432835821e-06,
1081
+ "loss": 2.0941,
1082
+ "step": 179
1083
+ },
1084
+ {
1085
+ "epoch": 0.5,
1086
+ "learning_rate": 4.029850746268657e-06,
1087
+ "loss": 2.0787,
1088
+ "step": 180
1089
+ },
1090
+ {
1091
+ "epoch": 0.5,
1092
+ "learning_rate": 4.022388059701493e-06,
1093
+ "loss": 2.177,
1094
+ "step": 181
1095
+ },
1096
+ {
1097
+ "epoch": 0.51,
1098
+ "learning_rate": 4.014925373134328e-06,
1099
+ "loss": 2.1608,
1100
+ "step": 182
1101
+ },
1102
+ {
1103
+ "epoch": 0.51,
1104
+ "learning_rate": 4.0074626865671645e-06,
1105
+ "loss": 2.1292,
1106
+ "step": 183
1107
+ },
1108
+ {
1109
+ "epoch": 0.51,
1110
+ "learning_rate": 4.000000000000001e-06,
1111
+ "loss": 2.0893,
1112
+ "step": 184
1113
+ },
1114
+ {
1115
+ "epoch": 0.51,
1116
+ "learning_rate": 3.992537313432836e-06,
1117
+ "loss": 2.0865,
1118
+ "step": 185
1119
+ },
1120
+ {
1121
+ "epoch": 0.52,
1122
+ "learning_rate": 3.985074626865672e-06,
1123
+ "loss": 2.1232,
1124
+ "step": 186
1125
+ },
1126
+ {
1127
+ "epoch": 0.52,
1128
+ "learning_rate": 3.977611940298508e-06,
1129
+ "loss": 2.0578,
1130
+ "step": 187
1131
+ },
1132
+ {
1133
+ "epoch": 0.52,
1134
+ "learning_rate": 3.970149253731344e-06,
1135
+ "loss": 2.0968,
1136
+ "step": 188
1137
+ },
1138
+ {
1139
+ "epoch": 0.53,
1140
+ "learning_rate": 3.96268656716418e-06,
1141
+ "loss": 2.1495,
1142
+ "step": 189
1143
+ },
1144
+ {
1145
+ "epoch": 0.53,
1146
+ "learning_rate": 3.955223880597015e-06,
1147
+ "loss": 2.0749,
1148
+ "step": 190
1149
+ },
1150
+ {
1151
+ "epoch": 0.53,
1152
+ "learning_rate": 3.947761194029851e-06,
1153
+ "loss": 2.0804,
1154
+ "step": 191
1155
+ },
1156
+ {
1157
+ "epoch": 0.53,
1158
+ "learning_rate": 3.940298507462687e-06,
1159
+ "loss": 2.1234,
1160
+ "step": 192
1161
+ },
1162
+ {
1163
+ "epoch": 0.54,
1164
+ "learning_rate": 3.932835820895523e-06,
1165
+ "loss": 2.06,
1166
+ "step": 193
1167
+ },
1168
+ {
1169
+ "epoch": 0.54,
1170
+ "learning_rate": 3.925373134328358e-06,
1171
+ "loss": 2.1024,
1172
+ "step": 194
1173
+ },
1174
+ {
1175
+ "epoch": 0.54,
1176
+ "learning_rate": 3.917910447761194e-06,
1177
+ "loss": 2.1366,
1178
+ "step": 195
1179
+ },
1180
+ {
1181
+ "epoch": 0.54,
1182
+ "learning_rate": 3.91044776119403e-06,
1183
+ "loss": 2.2053,
1184
+ "step": 196
1185
+ },
1186
+ {
1187
+ "epoch": 0.55,
1188
+ "learning_rate": 3.902985074626866e-06,
1189
+ "loss": 2.0121,
1190
+ "step": 197
1191
+ },
1192
+ {
1193
+ "epoch": 0.55,
1194
+ "learning_rate": 3.895522388059702e-06,
1195
+ "loss": 2.0567,
1196
+ "step": 198
1197
+ },
1198
+ {
1199
+ "epoch": 0.55,
1200
+ "learning_rate": 3.888059701492537e-06,
1201
+ "loss": 2.0644,
1202
+ "step": 199
1203
+ },
1204
+ {
1205
+ "epoch": 0.56,
1206
+ "learning_rate": 3.8805970149253735e-06,
1207
+ "loss": 2.037,
1208
+ "step": 200
1209
+ },
1210
+ {
1211
+ "epoch": 0.56,
1212
+ "learning_rate": 3.873134328358209e-06,
1213
+ "loss": 2.018,
1214
+ "step": 201
1215
+ },
1216
+ {
1217
+ "epoch": 0.56,
1218
+ "learning_rate": 3.865671641791045e-06,
1219
+ "loss": 2.0937,
1220
+ "step": 202
1221
+ },
1222
+ {
1223
+ "epoch": 0.56,
1224
+ "learning_rate": 3.858208955223881e-06,
1225
+ "loss": 2.0954,
1226
+ "step": 203
1227
+ },
1228
+ {
1229
+ "epoch": 0.57,
1230
+ "learning_rate": 3.8507462686567165e-06,
1231
+ "loss": 1.966,
1232
+ "step": 204
1233
+ },
1234
+ {
1235
+ "epoch": 0.57,
1236
+ "learning_rate": 3.843283582089553e-06,
1237
+ "loss": 2.0463,
1238
+ "step": 205
1239
+ },
1240
+ {
1241
+ "epoch": 0.57,
1242
+ "learning_rate": 3.835820895522389e-06,
1243
+ "loss": 2.0863,
1244
+ "step": 206
1245
+ },
1246
+ {
1247
+ "epoch": 0.57,
1248
+ "learning_rate": 3.828358208955224e-06,
1249
+ "loss": 2.2103,
1250
+ "step": 207
1251
+ },
1252
+ {
1253
+ "epoch": 0.58,
1254
+ "learning_rate": 3.82089552238806e-06,
1255
+ "loss": 2.1,
1256
+ "step": 208
1257
+ },
1258
+ {
1259
+ "epoch": 0.58,
1260
+ "learning_rate": 3.813432835820896e-06,
1261
+ "loss": 2.1174,
1262
+ "step": 209
1263
+ },
1264
+ {
1265
+ "epoch": 0.58,
1266
+ "learning_rate": 3.8059701492537314e-06,
1267
+ "loss": 2.0624,
1268
+ "step": 210
1269
+ },
1270
+ {
1271
+ "epoch": 0.59,
1272
+ "learning_rate": 3.7985074626865675e-06,
1273
+ "loss": 2.0201,
1274
+ "step": 211
1275
+ },
1276
+ {
1277
+ "epoch": 0.59,
1278
+ "learning_rate": 3.791044776119403e-06,
1279
+ "loss": 2.0903,
1280
+ "step": 212
1281
+ },
1282
+ {
1283
+ "epoch": 0.59,
1284
+ "learning_rate": 3.783582089552239e-06,
1285
+ "loss": 2.1576,
1286
+ "step": 213
1287
+ },
1288
+ {
1289
+ "epoch": 0.59,
1290
+ "learning_rate": 3.776119402985075e-06,
1291
+ "loss": 2.1386,
1292
+ "step": 214
1293
+ },
1294
+ {
1295
+ "epoch": 0.6,
1296
+ "learning_rate": 3.7686567164179105e-06,
1297
+ "loss": 2.1276,
1298
+ "step": 215
1299
+ },
1300
+ {
1301
+ "epoch": 0.6,
1302
+ "learning_rate": 3.7611940298507467e-06,
1303
+ "loss": 2.1071,
1304
+ "step": 216
1305
+ },
1306
+ {
1307
+ "epoch": 0.6,
1308
+ "learning_rate": 3.7537313432835824e-06,
1309
+ "loss": 2.0719,
1310
+ "step": 217
1311
+ },
1312
+ {
1313
+ "epoch": 0.61,
1314
+ "learning_rate": 3.746268656716418e-06,
1315
+ "loss": 2.1039,
1316
+ "step": 218
1317
+ },
1318
+ {
1319
+ "epoch": 0.61,
1320
+ "learning_rate": 3.738805970149254e-06,
1321
+ "loss": 1.9856,
1322
+ "step": 219
1323
+ },
1324
+ {
1325
+ "epoch": 0.61,
1326
+ "learning_rate": 3.73134328358209e-06,
1327
+ "loss": 2.1071,
1328
+ "step": 220
1329
+ },
1330
+ {
1331
+ "epoch": 0.61,
1332
+ "learning_rate": 3.7238805970149254e-06,
1333
+ "loss": 2.0637,
1334
+ "step": 221
1335
+ },
1336
+ {
1337
+ "epoch": 0.62,
1338
+ "learning_rate": 3.7164179104477616e-06,
1339
+ "loss": 2.0905,
1340
+ "step": 222
1341
+ },
1342
+ {
1343
+ "epoch": 0.62,
1344
+ "learning_rate": 3.7089552238805978e-06,
1345
+ "loss": 2.1093,
1346
+ "step": 223
1347
+ },
1348
+ {
1349
+ "epoch": 0.62,
1350
+ "learning_rate": 3.701492537313433e-06,
1351
+ "loss": 1.9111,
1352
+ "step": 224
1353
+ },
1354
+ {
1355
+ "epoch": 0.62,
1356
+ "learning_rate": 3.6940298507462693e-06,
1357
+ "loss": 2.1346,
1358
+ "step": 225
1359
+ },
1360
+ {
1361
+ "epoch": 0.63,
1362
+ "learning_rate": 3.6865671641791046e-06,
1363
+ "loss": 2.0913,
1364
+ "step": 226
1365
+ },
1366
+ {
1367
+ "epoch": 0.63,
1368
+ "learning_rate": 3.6791044776119407e-06,
1369
+ "loss": 1.9708,
1370
+ "step": 227
1371
+ },
1372
+ {
1373
+ "epoch": 0.63,
1374
+ "learning_rate": 3.6716417910447765e-06,
1375
+ "loss": 2.0772,
1376
+ "step": 228
1377
+ },
1378
+ {
1379
+ "epoch": 0.64,
1380
+ "learning_rate": 3.6641791044776122e-06,
1381
+ "loss": 1.9808,
1382
+ "step": 229
1383
+ },
1384
+ {
1385
+ "epoch": 0.64,
1386
+ "learning_rate": 3.656716417910448e-06,
1387
+ "loss": 2.0212,
1388
+ "step": 230
1389
+ },
1390
+ {
1391
+ "epoch": 0.64,
1392
+ "learning_rate": 3.649253731343284e-06,
1393
+ "loss": 2.0217,
1394
+ "step": 231
1395
+ },
1396
+ {
1397
+ "epoch": 0.64,
1398
+ "learning_rate": 3.6417910447761195e-06,
1399
+ "loss": 2.0808,
1400
+ "step": 232
1401
+ },
1402
+ {
1403
+ "epoch": 0.65,
1404
+ "learning_rate": 3.6343283582089557e-06,
1405
+ "loss": 2.0378,
1406
+ "step": 233
1407
+ },
1408
+ {
1409
+ "epoch": 0.65,
1410
+ "learning_rate": 3.6268656716417914e-06,
1411
+ "loss": 2.069,
1412
+ "step": 234
1413
+ },
1414
+ {
1415
+ "epoch": 0.65,
1416
+ "learning_rate": 3.619402985074627e-06,
1417
+ "loss": 2.1461,
1418
+ "step": 235
1419
+ },
1420
+ {
1421
+ "epoch": 0.66,
1422
+ "learning_rate": 3.611940298507463e-06,
1423
+ "loss": 2.1297,
1424
+ "step": 236
1425
+ },
1426
+ {
1427
+ "epoch": 0.66,
1428
+ "learning_rate": 3.6044776119402986e-06,
1429
+ "loss": 2.0674,
1430
+ "step": 237
1431
+ },
1432
+ {
1433
+ "epoch": 0.66,
1434
+ "learning_rate": 3.5970149253731344e-06,
1435
+ "loss": 1.9524,
1436
+ "step": 238
1437
+ },
1438
+ {
1439
+ "epoch": 0.66,
1440
+ "learning_rate": 3.5895522388059706e-06,
1441
+ "loss": 2.1357,
1442
+ "step": 239
1443
+ },
1444
+ {
1445
+ "epoch": 0.67,
1446
+ "learning_rate": 3.582089552238806e-06,
1447
+ "loss": 2.0455,
1448
+ "step": 240
1449
+ },
1450
+ {
1451
+ "epoch": 0.67,
1452
+ "learning_rate": 3.574626865671642e-06,
1453
+ "loss": 2.0855,
1454
+ "step": 241
1455
+ },
1456
+ {
1457
+ "epoch": 0.67,
1458
+ "learning_rate": 3.5671641791044782e-06,
1459
+ "loss": 1.9894,
1460
+ "step": 242
1461
+ },
1462
+ {
1463
+ "epoch": 0.68,
1464
+ "learning_rate": 3.5597014925373135e-06,
1465
+ "loss": 2.1611,
1466
+ "step": 243
1467
+ },
1468
+ {
1469
+ "epoch": 0.68,
1470
+ "learning_rate": 3.5522388059701497e-06,
1471
+ "loss": 2.0324,
1472
+ "step": 244
1473
+ },
1474
+ {
1475
+ "epoch": 0.68,
1476
+ "learning_rate": 3.5447761194029855e-06,
1477
+ "loss": 2.0957,
1478
+ "step": 245
1479
+ },
1480
+ {
1481
+ "epoch": 0.68,
1482
+ "learning_rate": 3.5373134328358212e-06,
1483
+ "loss": 2.1491,
1484
+ "step": 246
1485
+ },
1486
+ {
1487
+ "epoch": 0.69,
1488
+ "learning_rate": 3.529850746268657e-06,
1489
+ "loss": 2.1214,
1490
+ "step": 247
1491
+ },
1492
+ {
1493
+ "epoch": 0.69,
1494
+ "learning_rate": 3.5223880597014927e-06,
1495
+ "loss": 1.9777,
1496
+ "step": 248
1497
+ },
1498
+ {
1499
+ "epoch": 0.69,
1500
+ "learning_rate": 3.5149253731343285e-06,
1501
+ "loss": 2.0506,
1502
+ "step": 249
1503
+ },
1504
+ {
1505
+ "epoch": 0.69,
1506
+ "learning_rate": 3.5074626865671646e-06,
1507
+ "loss": 2.0092,
1508
+ "step": 250
1509
+ },
1510
+ {
1511
+ "epoch": 0.7,
1512
+ "learning_rate": 3.5e-06,
1513
+ "loss": 1.9975,
1514
+ "step": 251
1515
+ },
1516
+ {
1517
+ "epoch": 0.7,
1518
+ "learning_rate": 3.492537313432836e-06,
1519
+ "loss": 1.9842,
1520
+ "step": 252
1521
+ },
1522
+ {
1523
+ "epoch": 0.7,
1524
+ "learning_rate": 3.4850746268656723e-06,
1525
+ "loss": 1.9896,
1526
+ "step": 253
1527
+ },
1528
+ {
1529
+ "epoch": 0.71,
1530
+ "learning_rate": 3.4776119402985076e-06,
1531
+ "loss": 2.139,
1532
+ "step": 254
1533
+ },
1534
+ {
1535
+ "epoch": 0.71,
1536
+ "learning_rate": 3.4701492537313438e-06,
1537
+ "loss": 2.059,
1538
+ "step": 255
1539
+ },
1540
+ {
1541
+ "epoch": 0.71,
1542
+ "learning_rate": 3.4626865671641795e-06,
1543
+ "loss": 2.1195,
1544
+ "step": 256
1545
+ },
1546
+ {
1547
+ "epoch": 0.71,
1548
+ "learning_rate": 3.4552238805970153e-06,
1549
+ "loss": 2.0374,
1550
+ "step": 257
1551
+ },
1552
+ {
1553
+ "epoch": 0.72,
1554
+ "learning_rate": 3.447761194029851e-06,
1555
+ "loss": 2.0747,
1556
+ "step": 258
1557
+ },
1558
+ {
1559
+ "epoch": 0.72,
1560
+ "learning_rate": 3.440298507462687e-06,
1561
+ "loss": 2.0196,
1562
+ "step": 259
1563
+ },
1564
+ {
1565
+ "epoch": 0.72,
1566
+ "learning_rate": 3.4328358208955225e-06,
1567
+ "loss": 2.1209,
1568
+ "step": 260
1569
+ },
1570
+ {
1571
+ "epoch": 0.72,
1572
+ "learning_rate": 3.4253731343283587e-06,
1573
+ "loss": 2.0032,
1574
+ "step": 261
1575
+ },
1576
+ {
1577
+ "epoch": 0.73,
1578
+ "learning_rate": 3.417910447761194e-06,
1579
+ "loss": 2.0541,
1580
+ "step": 262
1581
+ },
1582
+ {
1583
+ "epoch": 0.73,
1584
+ "learning_rate": 3.41044776119403e-06,
1585
+ "loss": 1.9661,
1586
+ "step": 263
1587
+ },
1588
+ {
1589
+ "epoch": 0.73,
1590
+ "learning_rate": 3.402985074626866e-06,
1591
+ "loss": 2.0738,
1592
+ "step": 264
1593
+ },
1594
+ {
1595
+ "epoch": 0.74,
1596
+ "learning_rate": 3.3955223880597017e-06,
1597
+ "loss": 2.0677,
1598
+ "step": 265
1599
+ },
1600
+ {
1601
+ "epoch": 0.74,
1602
+ "learning_rate": 3.3880597014925374e-06,
1603
+ "loss": 2.029,
1604
+ "step": 266
1605
+ },
1606
+ {
1607
+ "epoch": 0.74,
1608
+ "learning_rate": 3.3805970149253736e-06,
1609
+ "loss": 2.1059,
1610
+ "step": 267
1611
+ },
1612
+ {
1613
+ "epoch": 0.74,
1614
+ "learning_rate": 3.373134328358209e-06,
1615
+ "loss": 2.1733,
1616
+ "step": 268
1617
+ },
1618
+ {
1619
+ "epoch": 0.75,
1620
+ "learning_rate": 3.365671641791045e-06,
1621
+ "loss": 2.0441,
1622
+ "step": 269
1623
+ },
1624
+ {
1625
+ "epoch": 0.75,
1626
+ "learning_rate": 3.3582089552238813e-06,
1627
+ "loss": 1.971,
1628
+ "step": 270
1629
+ },
1630
+ {
1631
+ "epoch": 0.75,
1632
+ "learning_rate": 3.3507462686567166e-06,
1633
+ "loss": 1.9794,
1634
+ "step": 271
1635
+ },
1636
+ {
1637
+ "epoch": 0.76,
1638
+ "learning_rate": 3.3432835820895528e-06,
1639
+ "loss": 2.0214,
1640
+ "step": 272
1641
+ },
1642
+ {
1643
+ "epoch": 0.76,
1644
+ "learning_rate": 3.335820895522388e-06,
1645
+ "loss": 2.0053,
1646
+ "step": 273
1647
+ },
1648
+ {
1649
+ "epoch": 0.76,
1650
+ "learning_rate": 3.3283582089552242e-06,
1651
+ "loss": 2.0227,
1652
+ "step": 274
1653
+ },
1654
+ {
1655
+ "epoch": 0.76,
1656
+ "learning_rate": 3.32089552238806e-06,
1657
+ "loss": 2.1288,
1658
+ "step": 275
1659
+ },
1660
+ {
1661
+ "epoch": 0.77,
1662
+ "learning_rate": 3.3134328358208957e-06,
1663
+ "loss": 1.9847,
1664
+ "step": 276
1665
+ },
1666
+ {
1667
+ "epoch": 0.77,
1668
+ "learning_rate": 3.3059701492537315e-06,
1669
+ "loss": 1.8497,
1670
+ "step": 277
1671
+ },
1672
+ {
1673
+ "epoch": 0.77,
1674
+ "learning_rate": 3.2985074626865677e-06,
1675
+ "loss": 2.2032,
1676
+ "step": 278
1677
+ },
1678
+ {
1679
+ "epoch": 0.78,
1680
+ "learning_rate": 3.291044776119403e-06,
1681
+ "loss": 1.9626,
1682
+ "step": 279
1683
+ },
1684
+ {
1685
+ "epoch": 0.78,
1686
+ "learning_rate": 3.283582089552239e-06,
1687
+ "loss": 1.9461,
1688
+ "step": 280
1689
+ },
1690
+ {
1691
+ "epoch": 0.78,
1692
+ "learning_rate": 3.2761194029850753e-06,
1693
+ "loss": 1.9711,
1694
+ "step": 281
1695
+ },
1696
+ {
1697
+ "epoch": 0.78,
1698
+ "learning_rate": 3.2686567164179106e-06,
1699
+ "loss": 2.0287,
1700
+ "step": 282
1701
+ },
1702
+ {
1703
+ "epoch": 0.79,
1704
+ "learning_rate": 3.261194029850747e-06,
1705
+ "loss": 2.1258,
1706
+ "step": 283
1707
+ },
1708
+ {
1709
+ "epoch": 0.79,
1710
+ "learning_rate": 3.253731343283582e-06,
1711
+ "loss": 2.1334,
1712
+ "step": 284
1713
+ },
1714
+ {
1715
+ "epoch": 0.79,
1716
+ "learning_rate": 3.2462686567164183e-06,
1717
+ "loss": 1.9876,
1718
+ "step": 285
1719
+ },
1720
+ {
1721
+ "epoch": 0.79,
1722
+ "learning_rate": 3.238805970149254e-06,
1723
+ "loss": 2.044,
1724
+ "step": 286
1725
+ },
1726
+ {
1727
+ "epoch": 0.8,
1728
+ "learning_rate": 3.23134328358209e-06,
1729
+ "loss": 1.9858,
1730
+ "step": 287
1731
+ },
1732
+ {
1733
+ "epoch": 0.8,
1734
+ "learning_rate": 3.2238805970149255e-06,
1735
+ "loss": 2.0305,
1736
+ "step": 288
1737
+ },
1738
+ {
1739
+ "epoch": 0.8,
1740
+ "learning_rate": 3.2164179104477617e-06,
1741
+ "loss": 1.9659,
1742
+ "step": 289
1743
+ },
1744
+ {
1745
+ "epoch": 0.81,
1746
+ "learning_rate": 3.208955223880597e-06,
1747
+ "loss": 2.0716,
1748
+ "step": 290
1749
+ },
1750
+ {
1751
+ "epoch": 0.81,
1752
+ "learning_rate": 3.2014925373134332e-06,
1753
+ "loss": 2.1441,
1754
+ "step": 291
1755
+ },
1756
+ {
1757
+ "epoch": 0.81,
1758
+ "learning_rate": 3.194029850746269e-06,
1759
+ "loss": 2.0301,
1760
+ "step": 292
1761
+ },
1762
+ {
1763
+ "epoch": 0.81,
1764
+ "learning_rate": 3.1865671641791047e-06,
1765
+ "loss": 1.8711,
1766
+ "step": 293
1767
+ },
1768
+ {
1769
+ "epoch": 0.82,
1770
+ "learning_rate": 3.1791044776119405e-06,
1771
+ "loss": 2.0777,
1772
+ "step": 294
1773
+ },
1774
+ {
1775
+ "epoch": 0.82,
1776
+ "learning_rate": 3.1716417910447766e-06,
1777
+ "loss": 2.0008,
1778
+ "step": 295
1779
+ },
1780
+ {
1781
+ "epoch": 0.82,
1782
+ "learning_rate": 3.164179104477612e-06,
1783
+ "loss": 1.9443,
1784
+ "step": 296
1785
+ },
1786
+ {
1787
+ "epoch": 0.82,
1788
+ "learning_rate": 3.156716417910448e-06,
1789
+ "loss": 2.0198,
1790
+ "step": 297
1791
+ },
1792
+ {
1793
+ "epoch": 0.83,
1794
+ "learning_rate": 3.1492537313432834e-06,
1795
+ "loss": 1.9773,
1796
+ "step": 298
1797
+ },
1798
+ {
1799
+ "epoch": 0.83,
1800
+ "learning_rate": 3.1417910447761196e-06,
1801
+ "loss": 1.9517,
1802
+ "step": 299
1803
+ },
1804
+ {
1805
+ "epoch": 0.83,
1806
+ "learning_rate": 3.1343283582089558e-06,
1807
+ "loss": 1.9766,
1808
+ "step": 300
1809
+ },
1810
+ {
1811
+ "epoch": 0.84,
1812
+ "learning_rate": 3.126865671641791e-06,
1813
+ "loss": 1.9886,
1814
+ "step": 301
1815
+ },
1816
+ {
1817
+ "epoch": 0.84,
1818
+ "learning_rate": 3.1194029850746273e-06,
1819
+ "loss": 1.9225,
1820
+ "step": 302
1821
+ },
1822
+ {
1823
+ "epoch": 0.84,
1824
+ "learning_rate": 3.111940298507463e-06,
1825
+ "loss": 2.0075,
1826
+ "step": 303
1827
+ },
1828
+ {
1829
+ "epoch": 0.84,
1830
+ "learning_rate": 3.1044776119402988e-06,
1831
+ "loss": 1.9959,
1832
+ "step": 304
1833
+ },
1834
+ {
1835
+ "epoch": 0.85,
1836
+ "learning_rate": 3.0970149253731345e-06,
1837
+ "loss": 1.8774,
1838
+ "step": 305
1839
+ },
1840
+ {
1841
+ "epoch": 0.85,
1842
+ "learning_rate": 3.0895522388059707e-06,
1843
+ "loss": 1.9782,
1844
+ "step": 306
1845
+ },
1846
+ {
1847
+ "epoch": 0.85,
1848
+ "learning_rate": 3.082089552238806e-06,
1849
+ "loss": 1.9044,
1850
+ "step": 307
1851
+ },
1852
+ {
1853
+ "epoch": 0.86,
1854
+ "learning_rate": 3.074626865671642e-06,
1855
+ "loss": 2.0563,
1856
+ "step": 308
1857
+ },
1858
+ {
1859
+ "epoch": 0.86,
1860
+ "learning_rate": 3.0671641791044775e-06,
1861
+ "loss": 2.0289,
1862
+ "step": 309
1863
+ },
1864
+ {
1865
+ "epoch": 0.86,
1866
+ "learning_rate": 3.0597014925373137e-06,
1867
+ "loss": 1.9767,
1868
+ "step": 310
1869
+ },
1870
+ {
1871
+ "epoch": 0.86,
1872
+ "learning_rate": 3.05223880597015e-06,
1873
+ "loss": 2.0064,
1874
+ "step": 311
1875
+ },
1876
+ {
1877
+ "epoch": 0.87,
1878
+ "learning_rate": 3.044776119402985e-06,
1879
+ "loss": 1.9287,
1880
+ "step": 312
1881
+ },
1882
+ {
1883
+ "epoch": 0.87,
1884
+ "learning_rate": 3.0373134328358213e-06,
1885
+ "loss": 1.9814,
1886
+ "step": 313
1887
+ },
1888
+ {
1889
+ "epoch": 0.87,
1890
+ "learning_rate": 3.029850746268657e-06,
1891
+ "loss": 2.0669,
1892
+ "step": 314
1893
+ },
1894
+ {
1895
+ "epoch": 0.88,
1896
+ "learning_rate": 3.022388059701493e-06,
1897
+ "loss": 2.0184,
1898
+ "step": 315
1899
+ },
1900
+ {
1901
+ "epoch": 0.88,
1902
+ "learning_rate": 3.0149253731343286e-06,
1903
+ "loss": 2.0662,
1904
+ "step": 316
1905
+ },
1906
+ {
1907
+ "epoch": 0.88,
1908
+ "learning_rate": 3.0074626865671648e-06,
1909
+ "loss": 2.0909,
1910
+ "step": 317
1911
+ },
1912
+ {
1913
+ "epoch": 0.88,
1914
+ "learning_rate": 3e-06,
1915
+ "loss": 1.8902,
1916
+ "step": 318
1917
+ },
1918
+ {
1919
+ "epoch": 0.89,
1920
+ "learning_rate": 2.9925373134328362e-06,
1921
+ "loss": 1.8977,
1922
+ "step": 319
1923
+ },
1924
+ {
1925
+ "epoch": 0.89,
1926
+ "learning_rate": 2.9850746268656716e-06,
1927
+ "loss": 2.0049,
1928
+ "step": 320
1929
+ },
1930
+ {
1931
+ "epoch": 0.89,
1932
+ "learning_rate": 2.9776119402985077e-06,
1933
+ "loss": 1.9111,
1934
+ "step": 321
1935
+ },
1936
+ {
1937
+ "epoch": 0.89,
1938
+ "learning_rate": 2.9701492537313435e-06,
1939
+ "loss": 1.959,
1940
+ "step": 322
1941
+ },
1942
+ {
1943
+ "epoch": 0.9,
1944
+ "learning_rate": 2.9626865671641792e-06,
1945
+ "loss": 1.9366,
1946
+ "step": 323
1947
+ },
1948
+ {
1949
+ "epoch": 0.9,
1950
+ "learning_rate": 2.955223880597015e-06,
1951
+ "loss": 2.1039,
1952
+ "step": 324
1953
+ },
1954
+ {
1955
+ "epoch": 0.9,
1956
+ "learning_rate": 2.947761194029851e-06,
1957
+ "loss": 1.8799,
1958
+ "step": 325
1959
+ },
1960
+ {
1961
+ "epoch": 0.91,
1962
+ "learning_rate": 2.9402985074626865e-06,
1963
+ "loss": 1.9662,
1964
+ "step": 326
1965
+ },
1966
+ {
1967
+ "epoch": 0.91,
1968
+ "learning_rate": 2.9328358208955226e-06,
1969
+ "loss": 2.05,
1970
+ "step": 327
1971
+ },
1972
+ {
1973
+ "epoch": 0.91,
1974
+ "learning_rate": 2.925373134328359e-06,
1975
+ "loss": 1.9833,
1976
+ "step": 328
1977
+ },
1978
+ {
1979
+ "epoch": 0.91,
1980
+ "learning_rate": 2.917910447761194e-06,
1981
+ "loss": 2.0645,
1982
+ "step": 329
1983
+ },
1984
+ {
1985
+ "epoch": 0.92,
1986
+ "learning_rate": 2.9104477611940303e-06,
1987
+ "loss": 1.8323,
1988
+ "step": 330
1989
+ },
1990
+ {
1991
+ "epoch": 0.92,
1992
+ "learning_rate": 2.902985074626866e-06,
1993
+ "loss": 2.017,
1994
+ "step": 331
1995
+ },
1996
+ {
1997
+ "epoch": 0.92,
1998
+ "learning_rate": 2.895522388059702e-06,
1999
+ "loss": 1.9408,
2000
+ "step": 332
2001
+ },
2002
+ {
2003
+ "epoch": 0.93,
2004
+ "learning_rate": 2.8880597014925376e-06,
2005
+ "loss": 1.8791,
2006
+ "step": 333
2007
+ },
2008
+ {
2009
+ "epoch": 0.93,
2010
+ "learning_rate": 2.8805970149253733e-06,
2011
+ "loss": 2.03,
2012
+ "step": 334
2013
+ },
2014
+ {
2015
+ "epoch": 0.93,
2016
+ "learning_rate": 2.873134328358209e-06,
2017
+ "loss": 2.0441,
2018
+ "step": 335
2019
+ },
2020
+ {
2021
+ "epoch": 0.93,
2022
+ "learning_rate": 2.8656716417910452e-06,
2023
+ "loss": 2.057,
2024
+ "step": 336
2025
+ },
2026
+ {
2027
+ "epoch": 0.94,
2028
+ "learning_rate": 2.8582089552238805e-06,
2029
+ "loss": 2.0373,
2030
+ "step": 337
2031
+ },
2032
+ {
2033
+ "epoch": 0.94,
2034
+ "learning_rate": 2.8507462686567167e-06,
2035
+ "loss": 1.8639,
2036
+ "step": 338
2037
+ },
2038
+ {
2039
+ "epoch": 0.94,
2040
+ "learning_rate": 2.843283582089553e-06,
2041
+ "loss": 1.9165,
2042
+ "step": 339
2043
+ },
2044
+ {
2045
+ "epoch": 0.94,
2046
+ "learning_rate": 2.835820895522388e-06,
2047
+ "loss": 1.8682,
2048
+ "step": 340
2049
+ },
2050
+ {
2051
+ "epoch": 0.95,
2052
+ "learning_rate": 2.8283582089552244e-06,
2053
+ "loss": 1.8921,
2054
+ "step": 341
2055
+ },
2056
+ {
2057
+ "epoch": 0.95,
2058
+ "learning_rate": 2.82089552238806e-06,
2059
+ "loss": 1.8587,
2060
+ "step": 342
2061
+ },
2062
+ {
2063
+ "epoch": 0.95,
2064
+ "learning_rate": 2.813432835820896e-06,
2065
+ "loss": 1.9945,
2066
+ "step": 343
2067
+ },
2068
+ {
2069
+ "epoch": 0.96,
2070
+ "learning_rate": 2.8059701492537316e-06,
2071
+ "loss": 2.0896,
2072
+ "step": 344
2073
+ },
2074
+ {
2075
+ "epoch": 0.96,
2076
+ "learning_rate": 2.798507462686567e-06,
2077
+ "loss": 1.9366,
2078
+ "step": 345
2079
+ },
2080
+ {
2081
+ "epoch": 0.96,
2082
+ "learning_rate": 2.791044776119403e-06,
2083
+ "loss": 1.9651,
2084
+ "step": 346
2085
+ },
2086
+ {
2087
+ "epoch": 0.96,
2088
+ "learning_rate": 2.7835820895522393e-06,
2089
+ "loss": 2.0134,
2090
+ "step": 347
2091
+ },
2092
+ {
2093
+ "epoch": 0.97,
2094
+ "learning_rate": 2.7761194029850746e-06,
2095
+ "loss": 1.9445,
2096
+ "step": 348
2097
+ },
2098
+ {
2099
+ "epoch": 0.97,
2100
+ "learning_rate": 2.7686567164179108e-06,
2101
+ "loss": 2.0142,
2102
+ "step": 349
2103
+ },
2104
+ {
2105
+ "epoch": 0.97,
2106
+ "learning_rate": 2.7611940298507465e-06,
2107
+ "loss": 1.8989,
2108
+ "step": 350
2109
+ },
2110
+ {
2111
+ "epoch": 0.97,
2112
+ "learning_rate": 2.7537313432835823e-06,
2113
+ "loss": 2.0269,
2114
+ "step": 351
2115
+ },
2116
+ {
2117
+ "epoch": 0.98,
2118
+ "learning_rate": 2.746268656716418e-06,
2119
+ "loss": 1.8869,
2120
+ "step": 352
2121
+ },
2122
+ {
2123
+ "epoch": 0.98,
2124
+ "learning_rate": 2.738805970149254e-06,
2125
+ "loss": 1.943,
2126
+ "step": 353
2127
+ },
2128
+ {
2129
+ "epoch": 0.98,
2130
+ "learning_rate": 2.7313432835820895e-06,
2131
+ "loss": 1.9522,
2132
+ "step": 354
2133
+ },
2134
+ {
2135
+ "epoch": 0.99,
2136
+ "learning_rate": 2.7238805970149257e-06,
2137
+ "loss": 2.0554,
2138
+ "step": 355
2139
+ },
2140
+ {
2141
+ "epoch": 0.99,
2142
+ "learning_rate": 2.716417910447761e-06,
2143
+ "loss": 1.995,
2144
+ "step": 356
2145
+ },
2146
+ {
2147
+ "epoch": 0.99,
2148
+ "learning_rate": 2.708955223880597e-06,
2149
+ "loss": 1.9536,
2150
+ "step": 357
2151
+ },
2152
+ {
2153
+ "epoch": 0.99,
2154
+ "learning_rate": 2.7014925373134333e-06,
2155
+ "loss": 1.9834,
2156
+ "step": 358
2157
+ },
2158
+ {
2159
+ "epoch": 1.0,
2160
+ "learning_rate": 2.6940298507462687e-06,
2161
+ "loss": 1.9415,
2162
+ "step": 359
2163
+ },
2164
+ {
2165
+ "epoch": 1.0,
2166
+ "learning_rate": 2.686567164179105e-06,
2167
+ "loss": 2.0185,
2168
+ "step": 360
2169
+ },
2170
+ {
2171
+ "epoch": 1.0,
2172
+ "eval_accuracy": 0.5638888888888889,
2173
+ "eval_loss": 1.9300233125686646,
2174
+ "eval_roc_auc": 0.9004854290932072,
2175
+ "eval_runtime": 14.7425,
2176
+ "eval_samples_per_second": 24.419,
2177
+ "eval_steps_per_second": 24.419,
2178
+ "step": 360
2179
+ },
2180
+ {
2181
+ "epoch": 1.0,
2182
+ "learning_rate": 2.6791044776119406e-06,
2183
+ "loss": 1.9965,
2184
+ "step": 361
2185
+ },
2186
+ {
2187
+ "epoch": 1.01,
2188
+ "learning_rate": 2.6716417910447763e-06,
2189
+ "loss": 1.9962,
2190
+ "step": 362
2191
+ },
2192
+ {
2193
+ "epoch": 1.01,
2194
+ "learning_rate": 2.664179104477612e-06,
2195
+ "loss": 1.911,
2196
+ "step": 363
2197
+ },
2198
+ {
2199
+ "epoch": 1.01,
2200
+ "learning_rate": 2.6567164179104482e-06,
2201
+ "loss": 1.8858,
2202
+ "step": 364
2203
+ },
2204
+ {
2205
+ "epoch": 1.01,
2206
+ "learning_rate": 2.6492537313432836e-06,
2207
+ "loss": 2.0619,
2208
+ "step": 365
2209
+ },
2210
+ {
2211
+ "epoch": 1.02,
2212
+ "learning_rate": 2.6417910447761197e-06,
2213
+ "loss": 1.9916,
2214
+ "step": 366
2215
+ },
2216
+ {
2217
+ "epoch": 1.02,
2218
+ "learning_rate": 2.634328358208956e-06,
2219
+ "loss": 1.8988,
2220
+ "step": 367
2221
+ },
2222
+ {
2223
+ "epoch": 1.02,
2224
+ "learning_rate": 2.6268656716417912e-06,
2225
+ "loss": 1.8329,
2226
+ "step": 368
2227
+ },
2228
+ {
2229
+ "epoch": 1.02,
2230
+ "learning_rate": 2.6194029850746274e-06,
2231
+ "loss": 2.0962,
2232
+ "step": 369
2233
+ },
2234
+ {
2235
+ "epoch": 1.03,
2236
+ "learning_rate": 2.6119402985074627e-06,
2237
+ "loss": 1.9337,
2238
+ "step": 370
2239
+ },
2240
+ {
2241
+ "epoch": 1.03,
2242
+ "learning_rate": 2.604477611940299e-06,
2243
+ "loss": 1.9416,
2244
+ "step": 371
2245
+ },
2246
+ {
2247
+ "epoch": 1.03,
2248
+ "learning_rate": 2.5970149253731346e-06,
2249
+ "loss": 1.9493,
2250
+ "step": 372
2251
+ },
2252
+ {
2253
+ "epoch": 1.04,
2254
+ "learning_rate": 2.58955223880597e-06,
2255
+ "loss": 1.9142,
2256
+ "step": 373
2257
+ },
2258
+ {
2259
+ "epoch": 1.04,
2260
+ "learning_rate": 2.582089552238806e-06,
2261
+ "loss": 1.9146,
2262
+ "step": 374
2263
+ },
2264
+ {
2265
+ "epoch": 1.04,
2266
+ "learning_rate": 2.5746268656716423e-06,
2267
+ "loss": 1.8407,
2268
+ "step": 375
2269
+ },
2270
+ {
2271
+ "epoch": 1.04,
2272
+ "learning_rate": 2.5671641791044776e-06,
2273
+ "loss": 1.8227,
2274
+ "step": 376
2275
+ },
2276
+ {
2277
+ "epoch": 1.05,
2278
+ "learning_rate": 2.559701492537314e-06,
2279
+ "loss": 1.9717,
2280
+ "step": 377
2281
+ },
2282
+ {
2283
+ "epoch": 1.05,
2284
+ "learning_rate": 2.5522388059701496e-06,
2285
+ "loss": 1.8316,
2286
+ "step": 378
2287
+ },
2288
+ {
2289
+ "epoch": 1.05,
2290
+ "learning_rate": 2.5447761194029853e-06,
2291
+ "loss": 1.9037,
2292
+ "step": 379
2293
+ },
2294
+ {
2295
+ "epoch": 1.06,
2296
+ "learning_rate": 2.537313432835821e-06,
2297
+ "loss": 2.0702,
2298
+ "step": 380
2299
+ },
2300
+ {
2301
+ "epoch": 1.06,
2302
+ "learning_rate": 2.529850746268657e-06,
2303
+ "loss": 1.9795,
2304
+ "step": 381
2305
+ },
2306
+ {
2307
+ "epoch": 1.06,
2308
+ "learning_rate": 2.5223880597014925e-06,
2309
+ "loss": 2.0151,
2310
+ "step": 382
2311
+ },
2312
+ {
2313
+ "epoch": 1.06,
2314
+ "learning_rate": 2.5149253731343287e-06,
2315
+ "loss": 1.9916,
2316
+ "step": 383
2317
+ },
2318
+ {
2319
+ "epoch": 1.07,
2320
+ "learning_rate": 2.507462686567164e-06,
2321
+ "loss": 1.9577,
2322
+ "step": 384
2323
+ },
2324
+ {
2325
+ "epoch": 1.07,
2326
+ "learning_rate": 2.5e-06,
2327
+ "loss": 2.0164,
2328
+ "step": 385
2329
+ },
2330
+ {
2331
+ "epoch": 1.07,
2332
+ "learning_rate": 2.492537313432836e-06,
2333
+ "loss": 1.9572,
2334
+ "step": 386
2335
+ },
2336
+ {
2337
+ "epoch": 1.07,
2338
+ "learning_rate": 2.485074626865672e-06,
2339
+ "loss": 2.0272,
2340
+ "step": 387
2341
+ },
2342
+ {
2343
+ "epoch": 1.08,
2344
+ "learning_rate": 2.477611940298508e-06,
2345
+ "loss": 1.8226,
2346
+ "step": 388
2347
+ },
2348
+ {
2349
+ "epoch": 1.08,
2350
+ "learning_rate": 2.4701492537313436e-06,
2351
+ "loss": 1.8782,
2352
+ "step": 389
2353
+ },
2354
+ {
2355
+ "epoch": 1.08,
2356
+ "learning_rate": 2.4626865671641794e-06,
2357
+ "loss": 1.9196,
2358
+ "step": 390
2359
+ },
2360
+ {
2361
+ "epoch": 1.09,
2362
+ "learning_rate": 2.455223880597015e-06,
2363
+ "loss": 1.9875,
2364
+ "step": 391
2365
+ },
2366
+ {
2367
+ "epoch": 1.09,
2368
+ "learning_rate": 2.447761194029851e-06,
2369
+ "loss": 1.984,
2370
+ "step": 392
2371
+ },
2372
+ {
2373
+ "epoch": 1.09,
2374
+ "learning_rate": 2.4402985074626866e-06,
2375
+ "loss": 1.8335,
2376
+ "step": 393
2377
+ },
2378
+ {
2379
+ "epoch": 1.09,
2380
+ "learning_rate": 2.4328358208955224e-06,
2381
+ "loss": 1.8524,
2382
+ "step": 394
2383
+ },
2384
+ {
2385
+ "epoch": 1.1,
2386
+ "learning_rate": 2.4253731343283585e-06,
2387
+ "loss": 1.8464,
2388
+ "step": 395
2389
+ },
2390
+ {
2391
+ "epoch": 1.1,
2392
+ "learning_rate": 2.4179104477611943e-06,
2393
+ "loss": 1.9122,
2394
+ "step": 396
2395
+ },
2396
+ {
2397
+ "epoch": 1.1,
2398
+ "learning_rate": 2.41044776119403e-06,
2399
+ "loss": 1.9877,
2400
+ "step": 397
2401
+ },
2402
+ {
2403
+ "epoch": 1.11,
2404
+ "learning_rate": 2.402985074626866e-06,
2405
+ "loss": 1.901,
2406
+ "step": 398
2407
+ },
2408
+ {
2409
+ "epoch": 1.11,
2410
+ "learning_rate": 2.3955223880597015e-06,
2411
+ "loss": 1.9487,
2412
+ "step": 399
2413
+ },
2414
+ {
2415
+ "epoch": 1.11,
2416
+ "learning_rate": 2.3880597014925373e-06,
2417
+ "loss": 2.0606,
2418
+ "step": 400
2419
+ },
2420
+ {
2421
+ "epoch": 1.11,
2422
+ "learning_rate": 2.380597014925373e-06,
2423
+ "loss": 1.8264,
2424
+ "step": 401
2425
+ },
2426
+ {
2427
+ "epoch": 1.12,
2428
+ "learning_rate": 2.373134328358209e-06,
2429
+ "loss": 1.7932,
2430
+ "step": 402
2431
+ },
2432
+ {
2433
+ "epoch": 1.12,
2434
+ "learning_rate": 2.365671641791045e-06,
2435
+ "loss": 2.013,
2436
+ "step": 403
2437
+ },
2438
+ {
2439
+ "epoch": 1.12,
2440
+ "learning_rate": 2.3582089552238807e-06,
2441
+ "loss": 1.9554,
2442
+ "step": 404
2443
+ },
2444
+ {
2445
+ "epoch": 1.12,
2446
+ "learning_rate": 2.350746268656717e-06,
2447
+ "loss": 2.0201,
2448
+ "step": 405
2449
+ },
2450
+ {
2451
+ "epoch": 1.13,
2452
+ "learning_rate": 2.3432835820895526e-06,
2453
+ "loss": 2.0781,
2454
+ "step": 406
2455
+ },
2456
+ {
2457
+ "epoch": 1.13,
2458
+ "learning_rate": 2.3358208955223883e-06,
2459
+ "loss": 1.9608,
2460
+ "step": 407
2461
+ },
2462
+ {
2463
+ "epoch": 1.13,
2464
+ "learning_rate": 2.328358208955224e-06,
2465
+ "loss": 1.9512,
2466
+ "step": 408
2467
+ },
2468
+ {
2469
+ "epoch": 1.14,
2470
+ "learning_rate": 2.32089552238806e-06,
2471
+ "loss": 2.0163,
2472
+ "step": 409
2473
+ },
2474
+ {
2475
+ "epoch": 1.14,
2476
+ "learning_rate": 2.3134328358208956e-06,
2477
+ "loss": 1.8655,
2478
+ "step": 410
2479
+ },
2480
+ {
2481
+ "epoch": 1.14,
2482
+ "learning_rate": 2.3059701492537313e-06,
2483
+ "loss": 1.9806,
2484
+ "step": 411
2485
+ },
2486
+ {
2487
+ "epoch": 1.14,
2488
+ "learning_rate": 2.298507462686567e-06,
2489
+ "loss": 1.8726,
2490
+ "step": 412
2491
+ },
2492
+ {
2493
+ "epoch": 1.15,
2494
+ "learning_rate": 2.2910447761194032e-06,
2495
+ "loss": 1.9621,
2496
+ "step": 413
2497
+ },
2498
+ {
2499
+ "epoch": 1.15,
2500
+ "learning_rate": 2.283582089552239e-06,
2501
+ "loss": 1.9423,
2502
+ "step": 414
2503
+ },
2504
+ {
2505
+ "epoch": 1.15,
2506
+ "learning_rate": 2.2761194029850747e-06,
2507
+ "loss": 2.0209,
2508
+ "step": 415
2509
+ },
2510
+ {
2511
+ "epoch": 1.16,
2512
+ "learning_rate": 2.268656716417911e-06,
2513
+ "loss": 1.8826,
2514
+ "step": 416
2515
+ },
2516
+ {
2517
+ "epoch": 1.16,
2518
+ "learning_rate": 2.2611940298507466e-06,
2519
+ "loss": 1.8739,
2520
+ "step": 417
2521
+ },
2522
+ {
2523
+ "epoch": 1.16,
2524
+ "learning_rate": 2.2537313432835824e-06,
2525
+ "loss": 1.9068,
2526
+ "step": 418
2527
+ },
2528
+ {
2529
+ "epoch": 1.16,
2530
+ "learning_rate": 2.246268656716418e-06,
2531
+ "loss": 2.0084,
2532
+ "step": 419
2533
+ },
2534
+ {
2535
+ "epoch": 1.17,
2536
+ "learning_rate": 2.238805970149254e-06,
2537
+ "loss": 1.916,
2538
+ "step": 420
2539
+ },
2540
+ {
2541
+ "epoch": 1.17,
2542
+ "learning_rate": 2.2313432835820896e-06,
2543
+ "loss": 2.0367,
2544
+ "step": 421
2545
+ },
2546
+ {
2547
+ "epoch": 1.17,
2548
+ "learning_rate": 2.2238805970149254e-06,
2549
+ "loss": 1.8648,
2550
+ "step": 422
2551
+ },
2552
+ {
2553
+ "epoch": 1.18,
2554
+ "learning_rate": 2.2164179104477616e-06,
2555
+ "loss": 1.8548,
2556
+ "step": 423
2557
+ },
2558
+ {
2559
+ "epoch": 1.18,
2560
+ "learning_rate": 2.2089552238805973e-06,
2561
+ "loss": 1.9594,
2562
+ "step": 424
2563
+ },
2564
+ {
2565
+ "epoch": 1.18,
2566
+ "learning_rate": 2.201492537313433e-06,
2567
+ "loss": 1.9486,
2568
+ "step": 425
2569
+ },
2570
+ {
2571
+ "epoch": 1.18,
2572
+ "learning_rate": 2.194029850746269e-06,
2573
+ "loss": 1.9606,
2574
+ "step": 426
2575
+ },
2576
+ {
2577
+ "epoch": 1.19,
2578
+ "learning_rate": 2.1865671641791045e-06,
2579
+ "loss": 1.8467,
2580
+ "step": 427
2581
+ },
2582
+ {
2583
+ "epoch": 1.19,
2584
+ "learning_rate": 2.1791044776119403e-06,
2585
+ "loss": 1.8841,
2586
+ "step": 428
2587
+ },
2588
+ {
2589
+ "epoch": 1.19,
2590
+ "learning_rate": 2.171641791044776e-06,
2591
+ "loss": 1.9361,
2592
+ "step": 429
2593
+ },
2594
+ {
2595
+ "epoch": 1.19,
2596
+ "learning_rate": 2.1641791044776118e-06,
2597
+ "loss": 1.8525,
2598
+ "step": 430
2599
+ },
2600
+ {
2601
+ "epoch": 1.2,
2602
+ "learning_rate": 2.156716417910448e-06,
2603
+ "loss": 1.8965,
2604
+ "step": 431
2605
+ },
2606
+ {
2607
+ "epoch": 1.2,
2608
+ "learning_rate": 2.1492537313432837e-06,
2609
+ "loss": 1.8744,
2610
+ "step": 432
2611
+ },
2612
+ {
2613
+ "epoch": 1.2,
2614
+ "learning_rate": 2.1417910447761194e-06,
2615
+ "loss": 1.7742,
2616
+ "step": 433
2617
+ },
2618
+ {
2619
+ "epoch": 1.21,
2620
+ "learning_rate": 2.1343283582089556e-06,
2621
+ "loss": 1.9439,
2622
+ "step": 434
2623
+ },
2624
+ {
2625
+ "epoch": 1.21,
2626
+ "learning_rate": 2.1268656716417914e-06,
2627
+ "loss": 2.0789,
2628
+ "step": 435
2629
+ },
2630
+ {
2631
+ "epoch": 1.21,
2632
+ "learning_rate": 2.119402985074627e-06,
2633
+ "loss": 1.8909,
2634
+ "step": 436
2635
+ },
2636
+ {
2637
+ "epoch": 1.21,
2638
+ "learning_rate": 2.111940298507463e-06,
2639
+ "loss": 1.8485,
2640
+ "step": 437
2641
+ },
2642
+ {
2643
+ "epoch": 1.22,
2644
+ "learning_rate": 2.1044776119402986e-06,
2645
+ "loss": 1.947,
2646
+ "step": 438
2647
+ },
2648
+ {
2649
+ "epoch": 1.22,
2650
+ "learning_rate": 2.0970149253731344e-06,
2651
+ "loss": 1.8121,
2652
+ "step": 439
2653
+ },
2654
+ {
2655
+ "epoch": 1.22,
2656
+ "learning_rate": 2.08955223880597e-06,
2657
+ "loss": 1.9623,
2658
+ "step": 440
2659
+ },
2660
+ {
2661
+ "epoch": 1.23,
2662
+ "learning_rate": 2.0820895522388063e-06,
2663
+ "loss": 1.9504,
2664
+ "step": 441
2665
+ },
2666
+ {
2667
+ "epoch": 1.23,
2668
+ "learning_rate": 2.074626865671642e-06,
2669
+ "loss": 1.8736,
2670
+ "step": 442
2671
+ },
2672
+ {
2673
+ "epoch": 1.23,
2674
+ "learning_rate": 2.0671641791044778e-06,
2675
+ "loss": 1.7974,
2676
+ "step": 443
2677
+ },
2678
+ {
2679
+ "epoch": 1.23,
2680
+ "learning_rate": 2.0597014925373135e-06,
2681
+ "loss": 1.8689,
2682
+ "step": 444
2683
+ },
2684
+ {
2685
+ "epoch": 1.24,
2686
+ "learning_rate": 2.0522388059701497e-06,
2687
+ "loss": 1.8249,
2688
+ "step": 445
2689
+ },
2690
+ {
2691
+ "epoch": 1.24,
2692
+ "learning_rate": 2.0447761194029854e-06,
2693
+ "loss": 1.8112,
2694
+ "step": 446
2695
+ },
2696
+ {
2697
+ "epoch": 1.24,
2698
+ "learning_rate": 2.037313432835821e-06,
2699
+ "loss": 2.0194,
2700
+ "step": 447
2701
+ },
2702
+ {
2703
+ "epoch": 1.24,
2704
+ "learning_rate": 2.029850746268657e-06,
2705
+ "loss": 1.7701,
2706
+ "step": 448
2707
+ },
2708
+ {
2709
+ "epoch": 1.25,
2710
+ "learning_rate": 2.0223880597014927e-06,
2711
+ "loss": 1.7401,
2712
+ "step": 449
2713
+ },
2714
+ {
2715
+ "epoch": 1.25,
2716
+ "learning_rate": 2.0149253731343284e-06,
2717
+ "loss": 1.9414,
2718
+ "step": 450
2719
+ },
2720
+ {
2721
+ "epoch": 1.25,
2722
+ "learning_rate": 2.007462686567164e-06,
2723
+ "loss": 1.9715,
2724
+ "step": 451
2725
+ },
2726
+ {
2727
+ "epoch": 1.26,
2728
+ "learning_rate": 2.0000000000000003e-06,
2729
+ "loss": 1.9848,
2730
+ "step": 452
2731
+ },
2732
+ {
2733
+ "epoch": 1.26,
2734
+ "learning_rate": 1.992537313432836e-06,
2735
+ "loss": 2.1102,
2736
+ "step": 453
2737
+ },
2738
+ {
2739
+ "epoch": 1.26,
2740
+ "learning_rate": 1.985074626865672e-06,
2741
+ "loss": 1.947,
2742
+ "step": 454
2743
+ },
2744
+ {
2745
+ "epoch": 1.26,
2746
+ "learning_rate": 1.9776119402985076e-06,
2747
+ "loss": 1.8938,
2748
+ "step": 455
2749
+ },
2750
+ {
2751
+ "epoch": 1.27,
2752
+ "learning_rate": 1.9701492537313433e-06,
2753
+ "loss": 1.8564,
2754
+ "step": 456
2755
+ },
2756
+ {
2757
+ "epoch": 1.27,
2758
+ "learning_rate": 1.962686567164179e-06,
2759
+ "loss": 2.0322,
2760
+ "step": 457
2761
+ },
2762
+ {
2763
+ "epoch": 1.27,
2764
+ "learning_rate": 1.955223880597015e-06,
2765
+ "loss": 2.025,
2766
+ "step": 458
2767
+ },
2768
+ {
2769
+ "epoch": 1.27,
2770
+ "learning_rate": 1.947761194029851e-06,
2771
+ "loss": 1.9294,
2772
+ "step": 459
2773
+ },
2774
+ {
2775
+ "epoch": 1.28,
2776
+ "learning_rate": 1.9402985074626867e-06,
2777
+ "loss": 1.885,
2778
+ "step": 460
2779
+ },
2780
+ {
2781
+ "epoch": 1.28,
2782
+ "learning_rate": 1.9328358208955225e-06,
2783
+ "loss": 2.0458,
2784
+ "step": 461
2785
+ },
2786
+ {
2787
+ "epoch": 1.28,
2788
+ "learning_rate": 1.9253731343283582e-06,
2789
+ "loss": 1.6424,
2790
+ "step": 462
2791
+ },
2792
+ {
2793
+ "epoch": 1.29,
2794
+ "learning_rate": 1.9179104477611944e-06,
2795
+ "loss": 2.0316,
2796
+ "step": 463
2797
+ },
2798
+ {
2799
+ "epoch": 1.29,
2800
+ "learning_rate": 1.91044776119403e-06,
2801
+ "loss": 1.8323,
2802
+ "step": 464
2803
+ },
2804
+ {
2805
+ "epoch": 1.29,
2806
+ "learning_rate": 1.9029850746268657e-06,
2807
+ "loss": 1.9308,
2808
+ "step": 465
2809
+ },
2810
+ {
2811
+ "epoch": 1.29,
2812
+ "learning_rate": 1.8955223880597014e-06,
2813
+ "loss": 1.8161,
2814
+ "step": 466
2815
+ },
2816
+ {
2817
+ "epoch": 1.3,
2818
+ "learning_rate": 1.8880597014925376e-06,
2819
+ "loss": 1.9323,
2820
+ "step": 467
2821
+ },
2822
+ {
2823
+ "epoch": 1.3,
2824
+ "learning_rate": 1.8805970149253733e-06,
2825
+ "loss": 2.0803,
2826
+ "step": 468
2827
+ },
2828
+ {
2829
+ "epoch": 1.3,
2830
+ "learning_rate": 1.873134328358209e-06,
2831
+ "loss": 1.9013,
2832
+ "step": 469
2833
+ },
2834
+ {
2835
+ "epoch": 1.31,
2836
+ "learning_rate": 1.865671641791045e-06,
2837
+ "loss": 1.9763,
2838
+ "step": 470
2839
+ },
2840
+ {
2841
+ "epoch": 1.31,
2842
+ "learning_rate": 1.8582089552238808e-06,
2843
+ "loss": 1.9798,
2844
+ "step": 471
2845
+ },
2846
+ {
2847
+ "epoch": 1.31,
2848
+ "learning_rate": 1.8507462686567165e-06,
2849
+ "loss": 1.9376,
2850
+ "step": 472
2851
+ },
2852
+ {
2853
+ "epoch": 1.31,
2854
+ "learning_rate": 1.8432835820895523e-06,
2855
+ "loss": 2.0252,
2856
+ "step": 473
2857
+ },
2858
+ {
2859
+ "epoch": 1.32,
2860
+ "learning_rate": 1.8358208955223882e-06,
2861
+ "loss": 1.7557,
2862
+ "step": 474
2863
+ },
2864
+ {
2865
+ "epoch": 1.32,
2866
+ "learning_rate": 1.828358208955224e-06,
2867
+ "loss": 2.001,
2868
+ "step": 475
2869
+ },
2870
+ {
2871
+ "epoch": 1.32,
2872
+ "learning_rate": 1.8208955223880597e-06,
2873
+ "loss": 2.1158,
2874
+ "step": 476
2875
+ },
2876
+ {
2877
+ "epoch": 1.32,
2878
+ "learning_rate": 1.8134328358208957e-06,
2879
+ "loss": 1.8338,
2880
+ "step": 477
2881
+ },
2882
+ {
2883
+ "epoch": 1.33,
2884
+ "learning_rate": 1.8059701492537314e-06,
2885
+ "loss": 1.8692,
2886
+ "step": 478
2887
+ },
2888
+ {
2889
+ "epoch": 1.33,
2890
+ "learning_rate": 1.7985074626865672e-06,
2891
+ "loss": 1.9515,
2892
+ "step": 479
2893
+ },
2894
+ {
2895
+ "epoch": 1.33,
2896
+ "learning_rate": 1.791044776119403e-06,
2897
+ "loss": 2.073,
2898
+ "step": 480
2899
+ },
2900
+ {
2901
+ "epoch": 1.34,
2902
+ "learning_rate": 1.7835820895522391e-06,
2903
+ "loss": 1.9917,
2904
+ "step": 481
2905
+ },
2906
+ {
2907
+ "epoch": 1.34,
2908
+ "learning_rate": 1.7761194029850749e-06,
2909
+ "loss": 1.7809,
2910
+ "step": 482
2911
+ },
2912
+ {
2913
+ "epoch": 1.34,
2914
+ "learning_rate": 1.7686567164179106e-06,
2915
+ "loss": 1.9147,
2916
+ "step": 483
2917
+ },
2918
+ {
2919
+ "epoch": 1.34,
2920
+ "learning_rate": 1.7611940298507464e-06,
2921
+ "loss": 1.9618,
2922
+ "step": 484
2923
+ },
2924
+ {
2925
+ "epoch": 1.35,
2926
+ "learning_rate": 1.7537313432835823e-06,
2927
+ "loss": 1.9435,
2928
+ "step": 485
2929
+ },
2930
+ {
2931
+ "epoch": 1.35,
2932
+ "learning_rate": 1.746268656716418e-06,
2933
+ "loss": 1.8669,
2934
+ "step": 486
2935
+ },
2936
+ {
2937
+ "epoch": 1.35,
2938
+ "learning_rate": 1.7388059701492538e-06,
2939
+ "loss": 1.8714,
2940
+ "step": 487
2941
+ },
2942
+ {
2943
+ "epoch": 1.36,
2944
+ "learning_rate": 1.7313432835820898e-06,
2945
+ "loss": 1.8984,
2946
+ "step": 488
2947
+ },
2948
+ {
2949
+ "epoch": 1.36,
2950
+ "learning_rate": 1.7238805970149255e-06,
2951
+ "loss": 1.777,
2952
+ "step": 489
2953
+ },
2954
+ {
2955
+ "epoch": 1.36,
2956
+ "learning_rate": 1.7164179104477613e-06,
2957
+ "loss": 1.8989,
2958
+ "step": 490
2959
+ },
2960
+ {
2961
+ "epoch": 1.36,
2962
+ "learning_rate": 1.708955223880597e-06,
2963
+ "loss": 1.9492,
2964
+ "step": 491
2965
+ },
2966
+ {
2967
+ "epoch": 1.37,
2968
+ "learning_rate": 1.701492537313433e-06,
2969
+ "loss": 1.964,
2970
+ "step": 492
2971
+ },
2972
+ {
2973
+ "epoch": 1.37,
2974
+ "learning_rate": 1.6940298507462687e-06,
2975
+ "loss": 1.9117,
2976
+ "step": 493
2977
+ },
2978
+ {
2979
+ "epoch": 1.37,
2980
+ "learning_rate": 1.6865671641791045e-06,
2981
+ "loss": 1.9403,
2982
+ "step": 494
2983
+ },
2984
+ {
2985
+ "epoch": 1.38,
2986
+ "learning_rate": 1.6791044776119406e-06,
2987
+ "loss": 1.7131,
2988
+ "step": 495
2989
+ },
2990
+ {
2991
+ "epoch": 1.38,
2992
+ "learning_rate": 1.6716417910447764e-06,
2993
+ "loss": 1.7807,
2994
+ "step": 496
2995
+ },
2996
+ {
2997
+ "epoch": 1.38,
2998
+ "learning_rate": 1.6641791044776121e-06,
2999
+ "loss": 1.8228,
3000
+ "step": 497
3001
+ },
3002
+ {
3003
+ "epoch": 1.38,
3004
+ "learning_rate": 1.6567164179104479e-06,
3005
+ "loss": 1.9325,
3006
+ "step": 498
3007
+ },
3008
+ {
3009
+ "epoch": 1.39,
3010
+ "learning_rate": 1.6492537313432838e-06,
3011
+ "loss": 1.8643,
3012
+ "step": 499
3013
+ },
3014
+ {
3015
+ "epoch": 1.39,
3016
+ "learning_rate": 1.6417910447761196e-06,
3017
+ "loss": 1.9326,
3018
+ "step": 500
3019
+ },
3020
+ {
3021
+ "epoch": 1.39,
3022
+ "learning_rate": 1.6343283582089553e-06,
3023
+ "loss": 1.921,
3024
+ "step": 501
3025
+ },
3026
+ {
3027
+ "epoch": 1.39,
3028
+ "learning_rate": 1.626865671641791e-06,
3029
+ "loss": 1.8616,
3030
+ "step": 502
3031
+ },
3032
+ {
3033
+ "epoch": 1.4,
3034
+ "learning_rate": 1.619402985074627e-06,
3035
+ "loss": 1.7485,
3036
+ "step": 503
3037
+ },
3038
+ {
3039
+ "epoch": 1.4,
3040
+ "learning_rate": 1.6119402985074628e-06,
3041
+ "loss": 1.8943,
3042
+ "step": 504
3043
+ },
3044
+ {
3045
+ "epoch": 1.4,
3046
+ "learning_rate": 1.6044776119402985e-06,
3047
+ "loss": 1.8202,
3048
+ "step": 505
3049
+ },
3050
+ {
3051
+ "epoch": 1.41,
3052
+ "learning_rate": 1.5970149253731345e-06,
3053
+ "loss": 1.8878,
3054
+ "step": 506
3055
+ },
3056
+ {
3057
+ "epoch": 1.41,
3058
+ "learning_rate": 1.5895522388059702e-06,
3059
+ "loss": 1.8747,
3060
+ "step": 507
3061
+ },
3062
+ {
3063
+ "epoch": 1.41,
3064
+ "learning_rate": 1.582089552238806e-06,
3065
+ "loss": 1.8613,
3066
+ "step": 508
3067
+ },
3068
+ {
3069
+ "epoch": 1.41,
3070
+ "learning_rate": 1.5746268656716417e-06,
3071
+ "loss": 1.9888,
3072
+ "step": 509
3073
+ },
3074
+ {
3075
+ "epoch": 1.42,
3076
+ "learning_rate": 1.5671641791044779e-06,
3077
+ "loss": 1.9108,
3078
+ "step": 510
3079
+ },
3080
+ {
3081
+ "epoch": 1.42,
3082
+ "learning_rate": 1.5597014925373136e-06,
3083
+ "loss": 1.9351,
3084
+ "step": 511
3085
+ },
3086
+ {
3087
+ "epoch": 1.42,
3088
+ "learning_rate": 1.5522388059701494e-06,
3089
+ "loss": 1.8483,
3090
+ "step": 512
3091
+ },
3092
+ {
3093
+ "epoch": 1.43,
3094
+ "learning_rate": 1.5447761194029853e-06,
3095
+ "loss": 1.8947,
3096
+ "step": 513
3097
+ },
3098
+ {
3099
+ "epoch": 1.43,
3100
+ "learning_rate": 1.537313432835821e-06,
3101
+ "loss": 1.8294,
3102
+ "step": 514
3103
+ },
3104
+ {
3105
+ "epoch": 1.43,
3106
+ "learning_rate": 1.5298507462686568e-06,
3107
+ "loss": 1.8969,
3108
+ "step": 515
3109
+ },
3110
+ {
3111
+ "epoch": 1.43,
3112
+ "learning_rate": 1.5223880597014926e-06,
3113
+ "loss": 1.8804,
3114
+ "step": 516
3115
+ },
3116
+ {
3117
+ "epoch": 1.44,
3118
+ "learning_rate": 1.5149253731343285e-06,
3119
+ "loss": 1.9103,
3120
+ "step": 517
3121
+ },
3122
+ {
3123
+ "epoch": 1.44,
3124
+ "learning_rate": 1.5074626865671643e-06,
3125
+ "loss": 1.9362,
3126
+ "step": 518
3127
+ },
3128
+ {
3129
+ "epoch": 1.44,
3130
+ "learning_rate": 1.5e-06,
3131
+ "loss": 1.9531,
3132
+ "step": 519
3133
+ },
3134
+ {
3135
+ "epoch": 1.44,
3136
+ "learning_rate": 1.4925373134328358e-06,
3137
+ "loss": 1.8927,
3138
+ "step": 520
3139
+ },
3140
+ {
3141
+ "epoch": 1.45,
3142
+ "learning_rate": 1.4850746268656717e-06,
3143
+ "loss": 1.8403,
3144
+ "step": 521
3145
+ },
3146
+ {
3147
+ "epoch": 1.45,
3148
+ "learning_rate": 1.4776119402985075e-06,
3149
+ "loss": 1.876,
3150
+ "step": 522
3151
+ },
3152
+ {
3153
+ "epoch": 1.45,
3154
+ "learning_rate": 1.4701492537313432e-06,
3155
+ "loss": 1.8347,
3156
+ "step": 523
3157
+ },
3158
+ {
3159
+ "epoch": 1.46,
3160
+ "learning_rate": 1.4626865671641794e-06,
3161
+ "loss": 1.9134,
3162
+ "step": 524
3163
+ },
3164
+ {
3165
+ "epoch": 1.46,
3166
+ "learning_rate": 1.4552238805970152e-06,
3167
+ "loss": 1.85,
3168
+ "step": 525
3169
+ },
3170
+ {
3171
+ "epoch": 1.46,
3172
+ "learning_rate": 1.447761194029851e-06,
3173
+ "loss": 1.8518,
3174
+ "step": 526
3175
+ },
3176
+ {
3177
+ "epoch": 1.46,
3178
+ "learning_rate": 1.4402985074626866e-06,
3179
+ "loss": 1.8527,
3180
+ "step": 527
3181
+ },
3182
+ {
3183
+ "epoch": 1.47,
3184
+ "learning_rate": 1.4328358208955226e-06,
3185
+ "loss": 1.8301,
3186
+ "step": 528
3187
+ },
3188
+ {
3189
+ "epoch": 1.47,
3190
+ "learning_rate": 1.4253731343283584e-06,
3191
+ "loss": 1.6702,
3192
+ "step": 529
3193
+ },
3194
+ {
3195
+ "epoch": 1.47,
3196
+ "learning_rate": 1.417910447761194e-06,
3197
+ "loss": 1.962,
3198
+ "step": 530
3199
+ },
3200
+ {
3201
+ "epoch": 1.48,
3202
+ "learning_rate": 1.41044776119403e-06,
3203
+ "loss": 1.8934,
3204
+ "step": 531
3205
+ },
3206
+ {
3207
+ "epoch": 1.48,
3208
+ "learning_rate": 1.4029850746268658e-06,
3209
+ "loss": 1.919,
3210
+ "step": 532
3211
+ },
3212
+ {
3213
+ "epoch": 1.48,
3214
+ "learning_rate": 1.3955223880597016e-06,
3215
+ "loss": 1.6701,
3216
+ "step": 533
3217
+ },
3218
+ {
3219
+ "epoch": 1.48,
3220
+ "learning_rate": 1.3880597014925373e-06,
3221
+ "loss": 1.8981,
3222
+ "step": 534
3223
+ },
3224
+ {
3225
+ "epoch": 1.49,
3226
+ "learning_rate": 1.3805970149253733e-06,
3227
+ "loss": 2.0018,
3228
+ "step": 535
3229
+ },
3230
+ {
3231
+ "epoch": 1.49,
3232
+ "learning_rate": 1.373134328358209e-06,
3233
+ "loss": 1.8839,
3234
+ "step": 536
3235
+ },
3236
+ {
3237
+ "epoch": 1.49,
3238
+ "learning_rate": 1.3656716417910448e-06,
3239
+ "loss": 2.068,
3240
+ "step": 537
3241
+ },
3242
+ {
3243
+ "epoch": 1.49,
3244
+ "learning_rate": 1.3582089552238805e-06,
3245
+ "loss": 1.9649,
3246
+ "step": 538
3247
+ },
3248
+ {
3249
+ "epoch": 1.5,
3250
+ "learning_rate": 1.3507462686567167e-06,
3251
+ "loss": 1.913,
3252
+ "step": 539
3253
+ },
3254
+ {
3255
+ "epoch": 1.5,
3256
+ "learning_rate": 1.3432835820895524e-06,
3257
+ "loss": 1.9283,
3258
+ "step": 540
3259
+ },
3260
+ {
3261
+ "epoch": 1.5,
3262
+ "learning_rate": 1.3358208955223882e-06,
3263
+ "loss": 2.0541,
3264
+ "step": 541
3265
+ },
3266
+ {
3267
+ "epoch": 1.51,
3268
+ "learning_rate": 1.3283582089552241e-06,
3269
+ "loss": 1.843,
3270
+ "step": 542
3271
+ },
3272
+ {
3273
+ "epoch": 1.51,
3274
+ "learning_rate": 1.3208955223880599e-06,
3275
+ "loss": 1.8551,
3276
+ "step": 543
3277
+ },
3278
+ {
3279
+ "epoch": 1.51,
3280
+ "learning_rate": 1.3134328358208956e-06,
3281
+ "loss": 1.853,
3282
+ "step": 544
3283
+ },
3284
+ {
3285
+ "epoch": 1.51,
3286
+ "learning_rate": 1.3059701492537314e-06,
3287
+ "loss": 1.9547,
3288
+ "step": 545
3289
+ },
3290
+ {
3291
+ "epoch": 1.52,
3292
+ "learning_rate": 1.2985074626865673e-06,
3293
+ "loss": 1.7693,
3294
+ "step": 546
3295
+ },
3296
+ {
3297
+ "epoch": 1.52,
3298
+ "learning_rate": 1.291044776119403e-06,
3299
+ "loss": 2.13,
3300
+ "step": 547
3301
+ },
3302
+ {
3303
+ "epoch": 1.52,
3304
+ "learning_rate": 1.2835820895522388e-06,
3305
+ "loss": 2.1405,
3306
+ "step": 548
3307
+ },
3308
+ {
3309
+ "epoch": 1.52,
3310
+ "learning_rate": 1.2761194029850748e-06,
3311
+ "loss": 1.7077,
3312
+ "step": 549
3313
+ },
3314
+ {
3315
+ "epoch": 1.53,
3316
+ "learning_rate": 1.2686567164179105e-06,
3317
+ "loss": 1.9839,
3318
+ "step": 550
3319
+ },
3320
+ {
3321
+ "epoch": 1.53,
3322
+ "learning_rate": 1.2611940298507463e-06,
3323
+ "loss": 1.7851,
3324
+ "step": 551
3325
+ },
3326
+ {
3327
+ "epoch": 1.53,
3328
+ "learning_rate": 1.253731343283582e-06,
3329
+ "loss": 2.0055,
3330
+ "step": 552
3331
+ },
3332
+ {
3333
+ "epoch": 1.54,
3334
+ "learning_rate": 1.246268656716418e-06,
3335
+ "loss": 1.8724,
3336
+ "step": 553
3337
+ },
3338
+ {
3339
+ "epoch": 1.54,
3340
+ "learning_rate": 1.238805970149254e-06,
3341
+ "loss": 1.9407,
3342
+ "step": 554
3343
+ },
3344
+ {
3345
+ "epoch": 1.54,
3346
+ "learning_rate": 1.2313432835820897e-06,
3347
+ "loss": 1.7383,
3348
+ "step": 555
3349
+ },
3350
+ {
3351
+ "epoch": 1.54,
3352
+ "learning_rate": 1.2238805970149254e-06,
3353
+ "loss": 1.9087,
3354
+ "step": 556
3355
+ },
3356
+ {
3357
+ "epoch": 1.55,
3358
+ "learning_rate": 1.2164179104477612e-06,
3359
+ "loss": 2.0751,
3360
+ "step": 557
3361
+ },
3362
+ {
3363
+ "epoch": 1.55,
3364
+ "learning_rate": 1.2089552238805971e-06,
3365
+ "loss": 1.7087,
3366
+ "step": 558
3367
+ },
3368
+ {
3369
+ "epoch": 1.55,
3370
+ "learning_rate": 1.201492537313433e-06,
3371
+ "loss": 1.9712,
3372
+ "step": 559
3373
+ },
3374
+ {
3375
+ "epoch": 1.56,
3376
+ "learning_rate": 1.1940298507462686e-06,
3377
+ "loss": 1.7763,
3378
+ "step": 560
3379
+ },
3380
+ {
3381
+ "epoch": 1.56,
3382
+ "learning_rate": 1.1865671641791046e-06,
3383
+ "loss": 1.8555,
3384
+ "step": 561
3385
+ },
3386
+ {
3387
+ "epoch": 1.56,
3388
+ "learning_rate": 1.1791044776119403e-06,
3389
+ "loss": 1.964,
3390
+ "step": 562
3391
+ },
3392
+ {
3393
+ "epoch": 1.56,
3394
+ "learning_rate": 1.1716417910447763e-06,
3395
+ "loss": 1.7212,
3396
+ "step": 563
3397
+ },
3398
+ {
3399
+ "epoch": 1.57,
3400
+ "learning_rate": 1.164179104477612e-06,
3401
+ "loss": 1.844,
3402
+ "step": 564
3403
+ },
3404
+ {
3405
+ "epoch": 1.57,
3406
+ "learning_rate": 1.1567164179104478e-06,
3407
+ "loss": 1.765,
3408
+ "step": 565
3409
+ },
3410
+ {
3411
+ "epoch": 1.57,
3412
+ "learning_rate": 1.1492537313432835e-06,
3413
+ "loss": 2.0413,
3414
+ "step": 566
3415
+ },
3416
+ {
3417
+ "epoch": 1.57,
3418
+ "learning_rate": 1.1417910447761195e-06,
3419
+ "loss": 1.7817,
3420
+ "step": 567
3421
+ },
3422
+ {
3423
+ "epoch": 1.58,
3424
+ "learning_rate": 1.1343283582089555e-06,
3425
+ "loss": 1.9402,
3426
+ "step": 568
3427
+ },
3428
+ {
3429
+ "epoch": 1.58,
3430
+ "learning_rate": 1.1268656716417912e-06,
3431
+ "loss": 1.7291,
3432
+ "step": 569
3433
+ },
3434
+ {
3435
+ "epoch": 1.58,
3436
+ "learning_rate": 1.119402985074627e-06,
3437
+ "loss": 1.693,
3438
+ "step": 570
3439
+ },
3440
+ {
3441
+ "epoch": 1.59,
3442
+ "learning_rate": 1.1119402985074627e-06,
3443
+ "loss": 1.768,
3444
+ "step": 571
3445
+ },
3446
+ {
3447
+ "epoch": 1.59,
3448
+ "learning_rate": 1.1044776119402987e-06,
3449
+ "loss": 1.8089,
3450
+ "step": 572
3451
+ },
3452
+ {
3453
+ "epoch": 1.59,
3454
+ "learning_rate": 1.0970149253731344e-06,
3455
+ "loss": 1.7817,
3456
+ "step": 573
3457
+ },
3458
+ {
3459
+ "epoch": 1.59,
3460
+ "learning_rate": 1.0895522388059701e-06,
3461
+ "loss": 1.853,
3462
+ "step": 574
3463
+ },
3464
+ {
3465
+ "epoch": 1.6,
3466
+ "learning_rate": 1.0820895522388059e-06,
3467
+ "loss": 1.876,
3468
+ "step": 575
3469
+ },
3470
+ {
3471
+ "epoch": 1.6,
3472
+ "learning_rate": 1.0746268656716418e-06,
3473
+ "loss": 1.8657,
3474
+ "step": 576
3475
+ },
3476
+ {
3477
+ "epoch": 1.6,
3478
+ "learning_rate": 1.0671641791044778e-06,
3479
+ "loss": 1.8389,
3480
+ "step": 577
3481
+ },
3482
+ {
3483
+ "epoch": 1.61,
3484
+ "learning_rate": 1.0597014925373136e-06,
3485
+ "loss": 1.9378,
3486
+ "step": 578
3487
+ },
3488
+ {
3489
+ "epoch": 1.61,
3490
+ "learning_rate": 1.0522388059701493e-06,
3491
+ "loss": 2.0328,
3492
+ "step": 579
3493
+ },
3494
+ {
3495
+ "epoch": 1.61,
3496
+ "learning_rate": 1.044776119402985e-06,
3497
+ "loss": 1.7983,
3498
+ "step": 580
3499
+ },
3500
+ {
3501
+ "epoch": 1.61,
3502
+ "learning_rate": 1.037313432835821e-06,
3503
+ "loss": 1.8521,
3504
+ "step": 581
3505
+ },
3506
+ {
3507
+ "epoch": 1.62,
3508
+ "learning_rate": 1.0298507462686568e-06,
3509
+ "loss": 1.9444,
3510
+ "step": 582
3511
+ },
3512
+ {
3513
+ "epoch": 1.62,
3514
+ "learning_rate": 1.0223880597014927e-06,
3515
+ "loss": 1.8302,
3516
+ "step": 583
3517
+ },
3518
+ {
3519
+ "epoch": 1.62,
3520
+ "learning_rate": 1.0149253731343285e-06,
3521
+ "loss": 1.9136,
3522
+ "step": 584
3523
+ },
3524
+ {
3525
+ "epoch": 1.62,
3526
+ "learning_rate": 1.0074626865671642e-06,
3527
+ "loss": 1.7285,
3528
+ "step": 585
3529
+ },
3530
+ {
3531
+ "epoch": 1.63,
3532
+ "learning_rate": 1.0000000000000002e-06,
3533
+ "loss": 1.7922,
3534
+ "step": 586
3535
+ },
3536
+ {
3537
+ "epoch": 1.63,
3538
+ "learning_rate": 9.92537313432836e-07,
3539
+ "loss": 1.8326,
3540
+ "step": 587
3541
+ },
3542
+ {
3543
+ "epoch": 1.63,
3544
+ "learning_rate": 9.850746268656717e-07,
3545
+ "loss": 1.8254,
3546
+ "step": 588
3547
+ },
3548
+ {
3549
+ "epoch": 1.64,
3550
+ "learning_rate": 9.776119402985074e-07,
3551
+ "loss": 1.8687,
3552
+ "step": 589
3553
+ },
3554
+ {
3555
+ "epoch": 1.64,
3556
+ "learning_rate": 9.701492537313434e-07,
3557
+ "loss": 1.684,
3558
+ "step": 590
3559
+ },
3560
+ {
3561
+ "epoch": 1.64,
3562
+ "learning_rate": 9.626865671641791e-07,
3563
+ "loss": 2.0,
3564
+ "step": 591
3565
+ },
3566
+ {
3567
+ "epoch": 1.64,
3568
+ "learning_rate": 9.55223880597015e-07,
3569
+ "loss": 1.9586,
3570
+ "step": 592
3571
+ },
3572
+ {
3573
+ "epoch": 1.65,
3574
+ "learning_rate": 9.477611940298507e-07,
3575
+ "loss": 1.8259,
3576
+ "step": 593
3577
+ },
3578
+ {
3579
+ "epoch": 1.65,
3580
+ "learning_rate": 9.402985074626867e-07,
3581
+ "loss": 1.9655,
3582
+ "step": 594
3583
+ },
3584
+ {
3585
+ "epoch": 1.65,
3586
+ "learning_rate": 9.328358208955225e-07,
3587
+ "loss": 1.9727,
3588
+ "step": 595
3589
+ },
3590
+ {
3591
+ "epoch": 1.66,
3592
+ "learning_rate": 9.253731343283583e-07,
3593
+ "loss": 2.0243,
3594
+ "step": 596
3595
+ },
3596
+ {
3597
+ "epoch": 1.66,
3598
+ "learning_rate": 9.179104477611941e-07,
3599
+ "loss": 1.8084,
3600
+ "step": 597
3601
+ },
3602
+ {
3603
+ "epoch": 1.66,
3604
+ "learning_rate": 9.104477611940299e-07,
3605
+ "loss": 1.9852,
3606
+ "step": 598
3607
+ },
3608
+ {
3609
+ "epoch": 1.66,
3610
+ "learning_rate": 9.029850746268657e-07,
3611
+ "loss": 1.7237,
3612
+ "step": 599
3613
+ },
3614
+ {
3615
+ "epoch": 1.67,
3616
+ "learning_rate": 8.955223880597015e-07,
3617
+ "loss": 1.8284,
3618
+ "step": 600
3619
+ },
3620
+ {
3621
+ "epoch": 1.67,
3622
+ "learning_rate": 8.880597014925374e-07,
3623
+ "loss": 1.892,
3624
+ "step": 601
3625
+ },
3626
+ {
3627
+ "epoch": 1.67,
3628
+ "learning_rate": 8.805970149253732e-07,
3629
+ "loss": 1.8353,
3630
+ "step": 602
3631
+ },
3632
+ {
3633
+ "epoch": 1.68,
3634
+ "learning_rate": 8.73134328358209e-07,
3635
+ "loss": 1.8581,
3636
+ "step": 603
3637
+ },
3638
+ {
3639
+ "epoch": 1.68,
3640
+ "learning_rate": 8.656716417910449e-07,
3641
+ "loss": 1.6436,
3642
+ "step": 604
3643
+ },
3644
+ {
3645
+ "epoch": 1.68,
3646
+ "learning_rate": 8.582089552238806e-07,
3647
+ "loss": 1.9365,
3648
+ "step": 605
3649
+ },
3650
+ {
3651
+ "epoch": 1.68,
3652
+ "learning_rate": 8.507462686567165e-07,
3653
+ "loss": 1.979,
3654
+ "step": 606
3655
+ },
3656
+ {
3657
+ "epoch": 1.69,
3658
+ "learning_rate": 8.432835820895522e-07,
3659
+ "loss": 1.9144,
3660
+ "step": 607
3661
+ },
3662
+ {
3663
+ "epoch": 1.69,
3664
+ "learning_rate": 8.358208955223882e-07,
3665
+ "loss": 1.6382,
3666
+ "step": 608
3667
+ },
3668
+ {
3669
+ "epoch": 1.69,
3670
+ "learning_rate": 8.283582089552239e-07,
3671
+ "loss": 1.908,
3672
+ "step": 609
3673
+ },
3674
+ {
3675
+ "epoch": 1.69,
3676
+ "learning_rate": 8.208955223880598e-07,
3677
+ "loss": 1.8228,
3678
+ "step": 610
3679
+ },
3680
+ {
3681
+ "epoch": 1.7,
3682
+ "learning_rate": 8.134328358208955e-07,
3683
+ "loss": 1.9059,
3684
+ "step": 611
3685
+ },
3686
+ {
3687
+ "epoch": 1.7,
3688
+ "learning_rate": 8.059701492537314e-07,
3689
+ "loss": 1.8273,
3690
+ "step": 612
3691
+ },
3692
+ {
3693
+ "epoch": 1.7,
3694
+ "learning_rate": 7.985074626865672e-07,
3695
+ "loss": 1.7892,
3696
+ "step": 613
3697
+ },
3698
+ {
3699
+ "epoch": 1.71,
3700
+ "learning_rate": 7.91044776119403e-07,
3701
+ "loss": 1.8526,
3702
+ "step": 614
3703
+ },
3704
+ {
3705
+ "epoch": 1.71,
3706
+ "learning_rate": 7.835820895522389e-07,
3707
+ "loss": 1.7265,
3708
+ "step": 615
3709
+ },
3710
+ {
3711
+ "epoch": 1.71,
3712
+ "learning_rate": 7.761194029850747e-07,
3713
+ "loss": 1.7578,
3714
+ "step": 616
3715
+ },
3716
+ {
3717
+ "epoch": 1.71,
3718
+ "learning_rate": 7.686567164179105e-07,
3719
+ "loss": 1.9066,
3720
+ "step": 617
3721
+ },
3722
+ {
3723
+ "epoch": 1.72,
3724
+ "learning_rate": 7.611940298507463e-07,
3725
+ "loss": 1.9333,
3726
+ "step": 618
3727
+ },
3728
+ {
3729
+ "epoch": 1.72,
3730
+ "learning_rate": 7.537313432835821e-07,
3731
+ "loss": 1.9749,
3732
+ "step": 619
3733
+ },
3734
+ {
3735
+ "epoch": 1.72,
3736
+ "learning_rate": 7.462686567164179e-07,
3737
+ "loss": 1.9252,
3738
+ "step": 620
3739
+ },
3740
+ {
3741
+ "epoch": 1.73,
3742
+ "learning_rate": 7.388059701492537e-07,
3743
+ "loss": 1.6667,
3744
+ "step": 621
3745
+ },
3746
+ {
3747
+ "epoch": 1.73,
3748
+ "learning_rate": 7.313432835820897e-07,
3749
+ "loss": 1.9584,
3750
+ "step": 622
3751
+ },
3752
+ {
3753
+ "epoch": 1.73,
3754
+ "learning_rate": 7.238805970149255e-07,
3755
+ "loss": 1.8053,
3756
+ "step": 623
3757
+ },
3758
+ {
3759
+ "epoch": 1.73,
3760
+ "learning_rate": 7.164179104477613e-07,
3761
+ "loss": 1.9666,
3762
+ "step": 624
3763
+ },
3764
+ {
3765
+ "epoch": 1.74,
3766
+ "learning_rate": 7.08955223880597e-07,
3767
+ "loss": 1.9321,
3768
+ "step": 625
3769
+ },
3770
+ {
3771
+ "epoch": 1.74,
3772
+ "learning_rate": 7.014925373134329e-07,
3773
+ "loss": 1.9918,
3774
+ "step": 626
3775
+ },
3776
+ {
3777
+ "epoch": 1.74,
3778
+ "learning_rate": 6.940298507462687e-07,
3779
+ "loss": 1.9374,
3780
+ "step": 627
3781
+ },
3782
+ {
3783
+ "epoch": 1.74,
3784
+ "learning_rate": 6.865671641791045e-07,
3785
+ "loss": 1.8895,
3786
+ "step": 628
3787
+ },
3788
+ {
3789
+ "epoch": 1.75,
3790
+ "learning_rate": 6.791044776119403e-07,
3791
+ "loss": 2.0706,
3792
+ "step": 629
3793
+ },
3794
+ {
3795
+ "epoch": 1.75,
3796
+ "learning_rate": 6.716417910447762e-07,
3797
+ "loss": 1.6574,
3798
+ "step": 630
3799
+ },
3800
+ {
3801
+ "epoch": 1.75,
3802
+ "learning_rate": 6.641791044776121e-07,
3803
+ "loss": 1.7742,
3804
+ "step": 631
3805
+ },
3806
+ {
3807
+ "epoch": 1.76,
3808
+ "learning_rate": 6.567164179104478e-07,
3809
+ "loss": 1.7373,
3810
+ "step": 632
3811
+ },
3812
+ {
3813
+ "epoch": 1.76,
3814
+ "learning_rate": 6.492537313432837e-07,
3815
+ "loss": 1.9448,
3816
+ "step": 633
3817
+ },
3818
+ {
3819
+ "epoch": 1.76,
3820
+ "learning_rate": 6.417910447761194e-07,
3821
+ "loss": 1.7308,
3822
+ "step": 634
3823
+ },
3824
+ {
3825
+ "epoch": 1.76,
3826
+ "learning_rate": 6.343283582089553e-07,
3827
+ "loss": 2.0017,
3828
+ "step": 635
3829
+ },
3830
+ {
3831
+ "epoch": 1.77,
3832
+ "learning_rate": 6.26865671641791e-07,
3833
+ "loss": 1.891,
3834
+ "step": 636
3835
+ },
3836
+ {
3837
+ "epoch": 1.77,
3838
+ "learning_rate": 6.19402985074627e-07,
3839
+ "loss": 2.1061,
3840
+ "step": 637
3841
+ },
3842
+ {
3843
+ "epoch": 1.77,
3844
+ "learning_rate": 6.119402985074627e-07,
3845
+ "loss": 2.0477,
3846
+ "step": 638
3847
+ },
3848
+ {
3849
+ "epoch": 1.77,
3850
+ "learning_rate": 6.044776119402986e-07,
3851
+ "loss": 1.8572,
3852
+ "step": 639
3853
+ },
3854
+ {
3855
+ "epoch": 1.78,
3856
+ "learning_rate": 5.970149253731343e-07,
3857
+ "loss": 1.6217,
3858
+ "step": 640
3859
+ },
3860
+ {
3861
+ "epoch": 1.78,
3862
+ "learning_rate": 5.895522388059702e-07,
3863
+ "loss": 1.7674,
3864
+ "step": 641
3865
+ },
3866
+ {
3867
+ "epoch": 1.78,
3868
+ "learning_rate": 5.82089552238806e-07,
3869
+ "loss": 1.9851,
3870
+ "step": 642
3871
+ },
3872
+ {
3873
+ "epoch": 1.79,
3874
+ "learning_rate": 5.746268656716418e-07,
3875
+ "loss": 1.911,
3876
+ "step": 643
3877
+ },
3878
+ {
3879
+ "epoch": 1.79,
3880
+ "learning_rate": 5.671641791044777e-07,
3881
+ "loss": 1.9886,
3882
+ "step": 644
3883
+ },
3884
+ {
3885
+ "epoch": 1.79,
3886
+ "learning_rate": 5.597014925373135e-07,
3887
+ "loss": 1.7959,
3888
+ "step": 645
3889
+ },
3890
+ {
3891
+ "epoch": 1.79,
3892
+ "learning_rate": 5.522388059701493e-07,
3893
+ "loss": 1.6594,
3894
+ "step": 646
3895
+ },
3896
+ {
3897
+ "epoch": 1.8,
3898
+ "learning_rate": 5.447761194029851e-07,
3899
+ "loss": 1.8995,
3900
+ "step": 647
3901
+ },
3902
+ {
3903
+ "epoch": 1.8,
3904
+ "learning_rate": 5.373134328358209e-07,
3905
+ "loss": 2.0217,
3906
+ "step": 648
3907
+ },
3908
+ {
3909
+ "epoch": 1.8,
3910
+ "learning_rate": 5.298507462686568e-07,
3911
+ "loss": 1.7008,
3912
+ "step": 649
3913
+ },
3914
+ {
3915
+ "epoch": 1.81,
3916
+ "learning_rate": 5.223880597014925e-07,
3917
+ "loss": 1.9073,
3918
+ "step": 650
3919
+ },
3920
+ {
3921
+ "epoch": 1.81,
3922
+ "learning_rate": 5.149253731343284e-07,
3923
+ "loss": 1.8471,
3924
+ "step": 651
3925
+ },
3926
+ {
3927
+ "epoch": 1.81,
3928
+ "learning_rate": 5.074626865671642e-07,
3929
+ "loss": 1.7691,
3930
+ "step": 652
3931
+ },
3932
+ {
3933
+ "epoch": 1.81,
3934
+ "learning_rate": 5.000000000000001e-07,
3935
+ "loss": 1.8507,
3936
+ "step": 653
3937
+ },
3938
+ {
3939
+ "epoch": 1.82,
3940
+ "learning_rate": 4.925373134328358e-07,
3941
+ "loss": 1.9155,
3942
+ "step": 654
3943
+ },
3944
+ {
3945
+ "epoch": 1.82,
3946
+ "learning_rate": 4.850746268656717e-07,
3947
+ "loss": 1.8051,
3948
+ "step": 655
3949
+ },
3950
+ {
3951
+ "epoch": 1.82,
3952
+ "learning_rate": 4.776119402985075e-07,
3953
+ "loss": 1.9299,
3954
+ "step": 656
3955
+ },
3956
+ {
3957
+ "epoch": 1.82,
3958
+ "learning_rate": 4.7014925373134334e-07,
3959
+ "loss": 1.9093,
3960
+ "step": 657
3961
+ },
3962
+ {
3963
+ "epoch": 1.83,
3964
+ "learning_rate": 4.6268656716417914e-07,
3965
+ "loss": 1.9159,
3966
+ "step": 658
3967
+ },
3968
+ {
3969
+ "epoch": 1.83,
3970
+ "learning_rate": 4.5522388059701494e-07,
3971
+ "loss": 1.9707,
3972
+ "step": 659
3973
+ },
3974
+ {
3975
+ "epoch": 1.83,
3976
+ "learning_rate": 4.4776119402985074e-07,
3977
+ "loss": 1.9074,
3978
+ "step": 660
3979
+ },
3980
+ {
3981
+ "epoch": 1.84,
3982
+ "learning_rate": 4.402985074626866e-07,
3983
+ "loss": 1.8601,
3984
+ "step": 661
3985
+ },
3986
+ {
3987
+ "epoch": 1.84,
3988
+ "learning_rate": 4.3283582089552244e-07,
3989
+ "loss": 1.7746,
3990
+ "step": 662
3991
+ },
3992
+ {
3993
+ "epoch": 1.84,
3994
+ "learning_rate": 4.2537313432835824e-07,
3995
+ "loss": 1.7178,
3996
+ "step": 663
3997
+ },
3998
+ {
3999
+ "epoch": 1.84,
4000
+ "learning_rate": 4.179104477611941e-07,
4001
+ "loss": 1.8743,
4002
+ "step": 664
4003
+ },
4004
+ {
4005
+ "epoch": 1.85,
4006
+ "learning_rate": 4.104477611940299e-07,
4007
+ "loss": 1.7925,
4008
+ "step": 665
4009
+ },
4010
+ {
4011
+ "epoch": 1.85,
4012
+ "learning_rate": 4.029850746268657e-07,
4013
+ "loss": 1.7661,
4014
+ "step": 666
4015
+ },
4016
+ {
4017
+ "epoch": 1.85,
4018
+ "learning_rate": 3.955223880597015e-07,
4019
+ "loss": 1.8683,
4020
+ "step": 667
4021
+ },
4022
+ {
4023
+ "epoch": 1.86,
4024
+ "learning_rate": 3.8805970149253735e-07,
4025
+ "loss": 1.9261,
4026
+ "step": 668
4027
+ },
4028
+ {
4029
+ "epoch": 1.86,
4030
+ "learning_rate": 3.8059701492537315e-07,
4031
+ "loss": 1.8979,
4032
+ "step": 669
4033
+ },
4034
+ {
4035
+ "epoch": 1.86,
4036
+ "learning_rate": 3.7313432835820895e-07,
4037
+ "loss": 1.8414,
4038
+ "step": 670
4039
+ },
4040
+ {
4041
+ "epoch": 1.86,
4042
+ "learning_rate": 3.6567164179104485e-07,
4043
+ "loss": 1.8114,
4044
+ "step": 671
4045
+ },
4046
+ {
4047
+ "epoch": 1.87,
4048
+ "learning_rate": 3.5820895522388065e-07,
4049
+ "loss": 1.7951,
4050
+ "step": 672
4051
+ },
4052
+ {
4053
+ "epoch": 1.87,
4054
+ "learning_rate": 3.5074626865671645e-07,
4055
+ "loss": 1.7864,
4056
+ "step": 673
4057
+ },
4058
+ {
4059
+ "epoch": 1.87,
4060
+ "learning_rate": 3.4328358208955225e-07,
4061
+ "loss": 1.7674,
4062
+ "step": 674
4063
+ },
4064
+ {
4065
+ "epoch": 1.88,
4066
+ "learning_rate": 3.358208955223881e-07,
4067
+ "loss": 1.7364,
4068
+ "step": 675
4069
+ },
4070
+ {
4071
+ "epoch": 1.88,
4072
+ "learning_rate": 3.283582089552239e-07,
4073
+ "loss": 1.8747,
4074
+ "step": 676
4075
+ },
4076
+ {
4077
+ "epoch": 1.88,
4078
+ "learning_rate": 3.208955223880597e-07,
4079
+ "loss": 1.6726,
4080
+ "step": 677
4081
+ },
4082
+ {
4083
+ "epoch": 1.88,
4084
+ "learning_rate": 3.134328358208955e-07,
4085
+ "loss": 1.992,
4086
+ "step": 678
4087
+ },
4088
+ {
4089
+ "epoch": 1.89,
4090
+ "learning_rate": 3.0597014925373136e-07,
4091
+ "loss": 1.7435,
4092
+ "step": 679
4093
+ },
4094
+ {
4095
+ "epoch": 1.89,
4096
+ "learning_rate": 2.9850746268656716e-07,
4097
+ "loss": 1.771,
4098
+ "step": 680
4099
+ },
4100
+ {
4101
+ "epoch": 1.89,
4102
+ "learning_rate": 2.91044776119403e-07,
4103
+ "loss": 1.7188,
4104
+ "step": 681
4105
+ },
4106
+ {
4107
+ "epoch": 1.89,
4108
+ "learning_rate": 2.8358208955223886e-07,
4109
+ "loss": 1.8161,
4110
+ "step": 682
4111
+ },
4112
+ {
4113
+ "epoch": 1.9,
4114
+ "learning_rate": 2.7611940298507466e-07,
4115
+ "loss": 1.8452,
4116
+ "step": 683
4117
+ },
4118
+ {
4119
+ "epoch": 1.9,
4120
+ "learning_rate": 2.6865671641791046e-07,
4121
+ "loss": 1.8403,
4122
+ "step": 684
4123
+ },
4124
+ {
4125
+ "epoch": 1.9,
4126
+ "learning_rate": 2.6119402985074626e-07,
4127
+ "loss": 1.8474,
4128
+ "step": 685
4129
+ },
4130
+ {
4131
+ "epoch": 1.91,
4132
+ "learning_rate": 2.537313432835821e-07,
4133
+ "loss": 1.886,
4134
+ "step": 686
4135
+ },
4136
+ {
4137
+ "epoch": 1.91,
4138
+ "learning_rate": 2.462686567164179e-07,
4139
+ "loss": 1.9012,
4140
+ "step": 687
4141
+ },
4142
+ {
4143
+ "epoch": 1.91,
4144
+ "learning_rate": 2.3880597014925377e-07,
4145
+ "loss": 1.756,
4146
+ "step": 688
4147
+ },
4148
+ {
4149
+ "epoch": 1.91,
4150
+ "learning_rate": 2.3134328358208957e-07,
4151
+ "loss": 1.7697,
4152
+ "step": 689
4153
+ },
4154
+ {
4155
+ "epoch": 1.92,
4156
+ "learning_rate": 2.2388059701492537e-07,
4157
+ "loss": 1.719,
4158
+ "step": 690
4159
+ },
4160
+ {
4161
+ "epoch": 1.92,
4162
+ "learning_rate": 2.1641791044776122e-07,
4163
+ "loss": 1.8717,
4164
+ "step": 691
4165
+ },
4166
+ {
4167
+ "epoch": 1.92,
4168
+ "learning_rate": 2.0895522388059705e-07,
4169
+ "loss": 1.7692,
4170
+ "step": 692
4171
+ },
4172
+ {
4173
+ "epoch": 1.93,
4174
+ "learning_rate": 2.0149253731343285e-07,
4175
+ "loss": 1.7447,
4176
+ "step": 693
4177
+ },
4178
+ {
4179
+ "epoch": 1.93,
4180
+ "learning_rate": 1.9402985074626867e-07,
4181
+ "loss": 1.8626,
4182
+ "step": 694
4183
+ },
4184
+ {
4185
+ "epoch": 1.93,
4186
+ "learning_rate": 1.8656716417910447e-07,
4187
+ "loss": 1.73,
4188
+ "step": 695
4189
+ },
4190
+ {
4191
+ "epoch": 1.93,
4192
+ "learning_rate": 1.7910447761194033e-07,
4193
+ "loss": 1.8676,
4194
+ "step": 696
4195
+ },
4196
+ {
4197
+ "epoch": 1.94,
4198
+ "learning_rate": 1.7164179104477613e-07,
4199
+ "loss": 1.8456,
4200
+ "step": 697
4201
+ },
4202
+ {
4203
+ "epoch": 1.94,
4204
+ "learning_rate": 1.6417910447761195e-07,
4205
+ "loss": 1.828,
4206
+ "step": 698
4207
+ },
4208
+ {
4209
+ "epoch": 1.94,
4210
+ "learning_rate": 1.5671641791044775e-07,
4211
+ "loss": 1.7631,
4212
+ "step": 699
4213
+ },
4214
+ {
4215
+ "epoch": 1.94,
4216
+ "learning_rate": 1.4925373134328358e-07,
4217
+ "loss": 1.8135,
4218
+ "step": 700
4219
+ },
4220
+ {
4221
+ "epoch": 1.95,
4222
+ "learning_rate": 1.4179104477611943e-07,
4223
+ "loss": 1.8722,
4224
+ "step": 701
4225
+ },
4226
+ {
4227
+ "epoch": 1.95,
4228
+ "learning_rate": 1.3432835820895523e-07,
4229
+ "loss": 1.8964,
4230
+ "step": 702
4231
+ },
4232
+ {
4233
+ "epoch": 1.95,
4234
+ "learning_rate": 1.2686567164179106e-07,
4235
+ "loss": 2.0702,
4236
+ "step": 703
4237
+ },
4238
+ {
4239
+ "epoch": 1.96,
4240
+ "learning_rate": 1.1940298507462688e-07,
4241
+ "loss": 1.804,
4242
+ "step": 704
4243
+ },
4244
+ {
4245
+ "epoch": 1.96,
4246
+ "learning_rate": 1.1194029850746268e-07,
4247
+ "loss": 1.8169,
4248
+ "step": 705
4249
+ },
4250
+ {
4251
+ "epoch": 1.96,
4252
+ "learning_rate": 1.0447761194029852e-07,
4253
+ "loss": 1.7613,
4254
+ "step": 706
4255
+ },
4256
+ {
4257
+ "epoch": 1.96,
4258
+ "learning_rate": 9.701492537313434e-08,
4259
+ "loss": 1.9533,
4260
+ "step": 707
4261
+ },
4262
+ {
4263
+ "epoch": 1.97,
4264
+ "learning_rate": 8.955223880597016e-08,
4265
+ "loss": 1.6899,
4266
+ "step": 708
4267
+ },
4268
+ {
4269
+ "epoch": 1.97,
4270
+ "learning_rate": 8.208955223880598e-08,
4271
+ "loss": 1.9187,
4272
+ "step": 709
4273
+ },
4274
+ {
4275
+ "epoch": 1.97,
4276
+ "learning_rate": 7.462686567164179e-08,
4277
+ "loss": 1.7969,
4278
+ "step": 710
4279
+ },
4280
+ {
4281
+ "epoch": 1.98,
4282
+ "learning_rate": 6.716417910447762e-08,
4283
+ "loss": 1.8375,
4284
+ "step": 711
4285
+ },
4286
+ {
4287
+ "epoch": 1.98,
4288
+ "learning_rate": 5.970149253731344e-08,
4289
+ "loss": 1.9321,
4290
+ "step": 712
4291
+ },
4292
+ {
4293
+ "epoch": 1.98,
4294
+ "learning_rate": 5.223880597014926e-08,
4295
+ "loss": 1.7623,
4296
+ "step": 713
4297
+ },
4298
+ {
4299
+ "epoch": 1.98,
4300
+ "learning_rate": 4.477611940298508e-08,
4301
+ "loss": 1.8626,
4302
+ "step": 714
4303
+ },
4304
+ {
4305
+ "epoch": 1.99,
4306
+ "learning_rate": 3.7313432835820895e-08,
4307
+ "loss": 1.6645,
4308
+ "step": 715
4309
+ },
4310
+ {
4311
+ "epoch": 1.99,
4312
+ "learning_rate": 2.985074626865672e-08,
4313
+ "loss": 1.8508,
4314
+ "step": 716
4315
+ },
4316
+ {
4317
+ "epoch": 1.99,
4318
+ "learning_rate": 2.238805970149254e-08,
4319
+ "loss": 1.9498,
4320
+ "step": 717
4321
+ },
4322
+ {
4323
+ "epoch": 1.99,
4324
+ "learning_rate": 1.492537313432836e-08,
4325
+ "loss": 1.7232,
4326
+ "step": 718
4327
+ },
4328
+ {
4329
+ "epoch": 2.0,
4330
+ "learning_rate": 7.46268656716418e-09,
4331
+ "loss": 1.7666,
4332
+ "step": 719
4333
+ },
4334
+ {
4335
+ "epoch": 2.0,
4336
+ "learning_rate": 0.0,
4337
+ "loss": 1.985,
4338
+ "step": 720
4339
+ },
4340
+ {
4341
+ "epoch": 2.0,
4342
+ "eval_accuracy": 0.6111111111111112,
4343
+ "eval_loss": 1.7899529933929443,
4344
+ "eval_roc_auc": 0.934288448435741,
4345
+ "eval_runtime": 14.1739,
4346
+ "eval_samples_per_second": 25.399,
4347
+ "eval_steps_per_second": 25.399,
4348
+ "step": 720
4349
+ }
4350
+ ],
4351
+ "max_steps": 720,
4352
+ "num_train_epochs": 2,
4353
+ "total_flos": 7.844087817216e+16,
4354
+ "trial_name": null,
4355
+ "trial_params": null
4356
+ }
checkpoint-720/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b15460f8960dd772b9155929bdcf211845dce468e9b1843e593fa50f3ff413a7
3
+ size 4024
config.json ADDED
@@ -0,0 +1,131 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "facebook/wav2vec2-base-960h",
3
+ "activation_dropout": 0.1,
4
+ "adapter_kernel_size": 3,
5
+ "adapter_stride": 2,
6
+ "add_adapter": false,
7
+ "apply_spec_augment": true,
8
+ "architectures": [
9
+ "Wav2Vec2ForSequenceClassification"
10
+ ],
11
+ "attention_dropout": 0.1,
12
+ "bos_token_id": 1,
13
+ "classifier_proj_size": 256,
14
+ "codevector_dim": 256,
15
+ "contrastive_logits_temperature": 0.1,
16
+ "conv_bias": false,
17
+ "conv_dim": [
18
+ 512,
19
+ 512,
20
+ 512,
21
+ 512,
22
+ 512,
23
+ 512,
24
+ 512
25
+ ],
26
+ "conv_kernel": [
27
+ 10,
28
+ 3,
29
+ 3,
30
+ 3,
31
+ 3,
32
+ 2,
33
+ 2
34
+ ],
35
+ "conv_stride": [
36
+ 5,
37
+ 2,
38
+ 2,
39
+ 2,
40
+ 2,
41
+ 2,
42
+ 2
43
+ ],
44
+ "ctc_loss_reduction": "sum",
45
+ "ctc_zero_infinity": false,
46
+ "diversity_loss_weight": 0.1,
47
+ "do_stable_layer_norm": false,
48
+ "eos_token_id": 2,
49
+ "feat_extract_activation": "gelu",
50
+ "feat_extract_dropout": 0.0,
51
+ "feat_extract_norm": "group",
52
+ "feat_proj_dropout": 0.1,
53
+ "feat_quantizer_dropout": 0.0,
54
+ "final_dropout": 0.1,
55
+ "gradient_checkpointing": false,
56
+ "hidden_act": "gelu",
57
+ "hidden_dropout": 0.1,
58
+ "hidden_dropout_prob": 0.1,
59
+ "hidden_size": 768,
60
+ "id2label": {
61
+ "0": "Acoustic_Guitar",
62
+ "1": "Bass_Guitar",
63
+ "2": "Drum_set",
64
+ "3": "Electro_Guitar",
65
+ "4": "flute",
66
+ "5": "Hi_Hats",
67
+ "6": "Keyboard",
68
+ "7": "Trumpet",
69
+ "8": "Violin"
70
+ },
71
+ "initializer_range": 0.02,
72
+ "intermediate_size": 3072,
73
+ "label2id": {
74
+ "LABEL_0": 0,
75
+ "LABEL_1": 1,
76
+ "LABEL_2": 2,
77
+ "LABEL_3": 3,
78
+ "LABEL_4": 4,
79
+ "LABEL_5": 5,
80
+ "LABEL_6": 6,
81
+ "LABEL_7": 7,
82
+ "LABEL_8": 8
83
+ },
84
+ "layer_norm_eps": 1e-05,
85
+ "layerdrop": 0.1,
86
+ "mask_feature_length": 10,
87
+ "mask_feature_min_masks": 0,
88
+ "mask_feature_prob": 0.0,
89
+ "mask_time_length": 10,
90
+ "mask_time_min_masks": 2,
91
+ "mask_time_prob": 0.05,
92
+ "model_type": "wav2vec2",
93
+ "num_adapter_layers": 3,
94
+ "num_attention_heads": 12,
95
+ "num_codevector_groups": 2,
96
+ "num_codevectors_per_group": 320,
97
+ "num_conv_pos_embedding_groups": 16,
98
+ "num_conv_pos_embeddings": 128,
99
+ "num_feat_extract_layers": 7,
100
+ "num_hidden_layers": 12,
101
+ "num_negatives": 100,
102
+ "output_hidden_size": 768,
103
+ "pad_token_id": 0,
104
+ "proj_codevector_dim": 256,
105
+ "tdnn_dilation": [
106
+ 1,
107
+ 2,
108
+ 3,
109
+ 1,
110
+ 1
111
+ ],
112
+ "tdnn_dim": [
113
+ 512,
114
+ 512,
115
+ 512,
116
+ 512,
117
+ 1500
118
+ ],
119
+ "tdnn_kernel": [
120
+ 5,
121
+ 3,
122
+ 3,
123
+ 1,
124
+ 1
125
+ ],
126
+ "torch_dtype": "float32",
127
+ "transformers_version": "4.28.1",
128
+ "use_weighted_layer_sum": false,
129
+ "vocab_size": 32,
130
+ "xvector_output_dim": 512
131
+ }
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": 0.0,
7
+ "return_attention_mask": false,
8
+ "sampling_rate": 16000
9
+ }
pytorch_model.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:529b4dcf7630acfa3889e1a4d06f86c9168b131ff064132ec44c1a3222b8ff2e
3
+ size 378358054
training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b15460f8960dd772b9155929bdcf211845dce468e9b1843e593fa50f3ff413a7
3
+ size 4024