Hugging Face's logo

PuxAI
/
ablation_study_mha

Model card Files
xet
 Community
PuxAI commited on
Commit
645922a
·
verified ·
1 Parent(s): 9049f32

Upload batch 13/14: vsfc/mBERT/max/seed_42 ... vsfc/mBERT/mha_attention/seed_42

Browse files
Files changed (8) hide show
  1. mbert_paper_metrics/result_ablation_mbert_paper/vsfc/mBERT/max/seed_44/all_results.json +15 -0
  2. mbert_paper_metrics/result_ablation_mbert_paper/vsfc/mBERT/max/seed_44/eval_results.json +9 -0
  3. mbert_paper_metrics/result_ablation_mbert_paper/vsfc/mBERT/max/seed_44/train_results.json +9 -0
  4. mbert_paper_metrics/result_ablation_mbert_paper/vsfc/mBERT/max/seed_44/trainer_state.json +125 -0
  5. mbert_paper_metrics/result_ablation_mbert_paper/vsfc/mBERT/mean/seed_44/all_results.json +15 -0
  6. mbert_paper_metrics/result_ablation_mbert_paper/vsfc/mBERT/mean/seed_44/eval_results.json +9 -0
  7. mbert_paper_metrics/result_ablation_mbert_paper/vsfc/mBERT/mean/seed_44/train_results.json +9 -0
  8. mbert_paper_metrics/result_ablation_mbert_paper/vsfc/mBERT/mean/seed_44/trainer_state.json +125 -0
mbert_paper_metrics/result_ablation_mbert_paper/vsfc/mBERT/max/seed_44/all_results.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "eval_accuracy": 0.9336702463676564,
4
+ "eval_loss": 0.2205778807401657,
5
+ "eval_runtime": 5.7547,
6
+ "eval_samples": 1583,
7
+ "eval_samples_per_second": 275.081,
8
+ "eval_steps_per_second": 1.738,
9
+ "total_flos": 1151671573807104.0,
10
+ "train_loss": 0.23737809152314157,
11
+ "train_runtime": 53.3228,
12
+ "train_samples": 11426,
13
+ "train_samples_per_second": 642.839,
14
+ "train_steps_per_second": 8.045
15
+ }
mbert_paper_metrics/result_ablation_mbert_paper/vsfc/mBERT/max/seed_44/eval_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "eval_accuracy": 0.9336702463676564,
4
+ "eval_loss": 0.2205778807401657,
5
+ "eval_runtime": 5.7547,
6
+ "eval_samples": 1583,
7
+ "eval_samples_per_second": 275.081,
8
+ "eval_steps_per_second": 1.738
9
+ }
mbert_paper_metrics/result_ablation_mbert_paper/vsfc/mBERT/max/seed_44/train_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "total_flos": 1151671573807104.0,
4
+ "train_loss": 0.23737809152314157,
5
+ "train_runtime": 53.3228,
6
+ "train_samples": 11426,
7
+ "train_samples_per_second": 642.839,
8
+ "train_steps_per_second": 8.045
9
+ }
mbert_paper_metrics/result_ablation_mbert_paper/vsfc/mBERT/max/seed_44/trainer_state.json ADDED
@@ -0,0 +1,125 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 3.0,
5
+ "eval_steps": 500,
6
+ "global_step": 429,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.34965034965034963,
13
+ "grad_norm": 5.980754375457764,
14
+ "learning_rate": 1.766899766899767e-05,
15
+ "loss": 0.4927,
16
+ "step": 50
17
+ },
18
+ {
19
+ "epoch": 0.6993006993006993,
20
+ "grad_norm": 6.989233016967773,
21
+ "learning_rate": 1.533799533799534e-05,
22
+ "loss": 0.2897,
23
+ "step": 100
24
+ },
25
+ {
26
+ "epoch": 1.0,
27
+ "eval_accuracy": 0.9229311433986103,
28
+ "eval_loss": 0.22875232994556427,
29
+ "eval_runtime": 6.3806,
30
+ "eval_samples_per_second": 248.096,
31
+ "eval_steps_per_second": 1.567,
32
+ "step": 143
33
+ },
34
+ {
35
+ "epoch": 1.048951048951049,
36
+ "grad_norm": 2.49883770942688,
37
+ "learning_rate": 1.3006993006993008e-05,
38
+ "loss": 0.2407,
39
+ "step": 150
40
+ },
41
+ {
42
+ "epoch": 1.3986013986013985,
43
+ "grad_norm": 2.3097035884857178,
44
+ "learning_rate": 1.0675990675990678e-05,
45
+ "loss": 0.2102,
46
+ "step": 200
47
+ },
48
+ {
49
+ "epoch": 1.7482517482517483,
50
+ "grad_norm": 2.0148630142211914,
51
+ "learning_rate": 8.344988344988347e-06,
52
+ "loss": 0.2032,
53
+ "step": 250
54
+ },
55
+ {
56
+ "epoch": 2.0,
57
+ "eval_accuracy": 0.9330385344283006,
58
+ "eval_loss": 0.20936070382595062,
59
+ "eval_runtime": 6.4865,
60
+ "eval_samples_per_second": 244.045,
61
+ "eval_steps_per_second": 1.542,
62
+ "step": 286
63
+ },
64
+ {
65
+ "epoch": 2.097902097902098,
66
+ "grad_norm": 4.495963096618652,
67
+ "learning_rate": 6.013986013986014e-06,
68
+ "loss": 0.1906,
69
+ "step": 300
70
+ },
71
+ {
72
+ "epoch": 2.4475524475524475,
73
+ "grad_norm": 2.7515621185302734,
74
+ "learning_rate": 3.6829836829836835e-06,
75
+ "loss": 0.1569,
76
+ "step": 350
77
+ },
78
+ {
79
+ "epoch": 2.797202797202797,
80
+ "grad_norm": 3.731598377227783,
81
+ "learning_rate": 1.3519813519813522e-06,
82
+ "loss": 0.1554,
83
+ "step": 400
84
+ },
85
+ {
86
+ "epoch": 3.0,
87
+ "eval_accuracy": 0.9336702463676564,
88
+ "eval_loss": 0.2205778807401657,
89
+ "eval_runtime": 5.5675,
90
+ "eval_samples_per_second": 284.329,
91
+ "eval_steps_per_second": 1.796,
92
+ "step": 429
93
+ },
94
+ {
95
+ "epoch": 3.0,
96
+ "step": 429,
97
+ "total_flos": 1151671573807104.0,
98
+ "train_loss": 0.23737809152314157,
99
+ "train_runtime": 53.3228,
100
+ "train_samples_per_second": 642.839,
101
+ "train_steps_per_second": 8.045
102
+ }
103
+ ],
104
+ "logging_steps": 50,
105
+ "max_steps": 429,
106
+ "num_input_tokens_seen": 0,
107
+ "num_train_epochs": 3,
108
+ "save_steps": 500,
109
+ "stateful_callbacks": {
110
+ "TrainerControl": {
111
+ "args": {
112
+ "should_epoch_stop": false,
113
+ "should_evaluate": false,
114
+ "should_log": false,
115
+ "should_save": false,
116
+ "should_training_stop": false
117
+ },
118
+ "attributes": {}
119
+ }
120
+ },
121
+ "total_flos": 1151671573807104.0,
122
+ "train_batch_size": 80,
123
+ "trial_name": null,
124
+ "trial_params": null
125
+ }
mbert_paper_metrics/result_ablation_mbert_paper/vsfc/mBERT/mean/seed_44/all_results.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "eval_accuracy": 0.932406822488945,
4
+ "eval_loss": 0.21627604961395264,
5
+ "eval_runtime": 5.8552,
6
+ "eval_samples": 1583,
7
+ "eval_samples_per_second": 270.357,
8
+ "eval_steps_per_second": 1.708,
9
+ "total_flos": 1151671573807104.0,
10
+ "train_loss": 0.22658519144658443,
11
+ "train_runtime": 50.1557,
12
+ "train_samples": 11426,
13
+ "train_samples_per_second": 683.431,
14
+ "train_steps_per_second": 8.553
15
+ }
mbert_paper_metrics/result_ablation_mbert_paper/vsfc/mBERT/mean/seed_44/eval_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "eval_accuracy": 0.932406822488945,
4
+ "eval_loss": 0.21627604961395264,
5
+ "eval_runtime": 5.8552,
6
+ "eval_samples": 1583,
7
+ "eval_samples_per_second": 270.357,
8
+ "eval_steps_per_second": 1.708
9
+ }
mbert_paper_metrics/result_ablation_mbert_paper/vsfc/mBERT/mean/seed_44/train_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "total_flos": 1151671573807104.0,
4
+ "train_loss": 0.22658519144658443,
5
+ "train_runtime": 50.1557,
6
+ "train_samples": 11426,
7
+ "train_samples_per_second": 683.431,
8
+ "train_steps_per_second": 8.553
9
+ }
mbert_paper_metrics/result_ablation_mbert_paper/vsfc/mBERT/mean/seed_44/trainer_state.json ADDED
@@ -0,0 +1,125 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 3.0,
5
+ "eval_steps": 500,
6
+ "global_step": 429,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.34965034965034963,
13
+ "grad_norm": 7.715805530548096,
14
+ "learning_rate": 1.766899766899767e-05,
15
+ "loss": 0.4514,
16
+ "step": 50
17
+ },
18
+ {
19
+ "epoch": 0.6993006993006993,
20
+ "grad_norm": 8.426939964294434,
21
+ "learning_rate": 1.533799533799534e-05,
22
+ "loss": 0.2889,
23
+ "step": 100
24
+ },
25
+ {
26
+ "epoch": 1.0,
27
+ "eval_accuracy": 0.9260897030953885,
28
+ "eval_loss": 0.231907919049263,
29
+ "eval_runtime": 6.2563,
30
+ "eval_samples_per_second": 253.023,
31
+ "eval_steps_per_second": 1.598,
32
+ "step": 143
33
+ },
34
+ {
35
+ "epoch": 1.048951048951049,
36
+ "grad_norm": 2.6239259243011475,
37
+ "learning_rate": 1.3006993006993008e-05,
38
+ "loss": 0.2384,
39
+ "step": 150
40
+ },
41
+ {
42
+ "epoch": 1.3986013986013985,
43
+ "grad_norm": 1.9370007514953613,
44
+ "learning_rate": 1.0675990675990678e-05,
45
+ "loss": 0.2026,
46
+ "step": 200
47
+ },
48
+ {
49
+ "epoch": 1.7482517482517483,
50
+ "grad_norm": 2.24284291267395,
51
+ "learning_rate": 8.344988344988347e-06,
52
+ "loss": 0.1964,
53
+ "step": 250
54
+ },
55
+ {
56
+ "epoch": 2.0,
57
+ "eval_accuracy": 0.9311433986102338,
58
+ "eval_loss": 0.20785251259803772,
59
+ "eval_runtime": 6.0629,
60
+ "eval_samples_per_second": 261.098,
61
+ "eval_steps_per_second": 1.649,
62
+ "step": 286
63
+ },
64
+ {
65
+ "epoch": 2.097902097902098,
66
+ "grad_norm": 4.250367641448975,
67
+ "learning_rate": 6.013986013986014e-06,
68
+ "loss": 0.1811,
69
+ "step": 300
70
+ },
71
+ {
72
+ "epoch": 2.4475524475524475,
73
+ "grad_norm": 3.534698724746704,
74
+ "learning_rate": 3.6829836829836835e-06,
75
+ "loss": 0.147,
76
+ "step": 350
77
+ },
78
+ {
79
+ "epoch": 2.797202797202797,
80
+ "grad_norm": 4.129490375518799,
81
+ "learning_rate": 1.3519813519813522e-06,
82
+ "loss": 0.1501,
83
+ "step": 400
84
+ },
85
+ {
86
+ "epoch": 3.0,
87
+ "eval_accuracy": 0.932406822488945,
88
+ "eval_loss": 0.21627604961395264,
89
+ "eval_runtime": 5.6458,
90
+ "eval_samples_per_second": 280.384,
91
+ "eval_steps_per_second": 1.771,
92
+ "step": 429
93
+ },
94
+ {
95
+ "epoch": 3.0,
96
+ "step": 429,
97
+ "total_flos": 1151671573807104.0,
98
+ "train_loss": 0.22658519144658443,
99
+ "train_runtime": 50.1557,
100
+ "train_samples_per_second": 683.431,
101
+ "train_steps_per_second": 8.553
102
+ }
103
+ ],
104
+ "logging_steps": 50,
105
+ "max_steps": 429,
106
+ "num_input_tokens_seen": 0,
107
+ "num_train_epochs": 3,
108
+ "save_steps": 500,
109
+ "stateful_callbacks": {
110
+ "TrainerControl": {
111
+ "args": {
112
+ "should_epoch_stop": false,
113
+ "should_evaluate": false,
114
+ "should_log": false,
115
+ "should_save": false,
116
+ "should_training_stop": false
117
+ },
118
+ "attributes": {}
119
+ }
120
+ },
121
+ "total_flos": 1151671573807104.0,
122
+ "train_batch_size": 80,
123
+ "trial_name": null,
124
+ "trial_params": null
125
+ }