Hugging Face's logo

PuxAI
/
ablation_study_mha

Model card Files
xet
 Community
PuxAI commited on
Commit
0c792bd
·
verified ·
1 Parent(s): 8f68d7c

Upload batch 3/24: cola/mBERT/gated_multi_branch/seed_44 ... cola/mBERT/hf_sequence_classifier/seed_44

Browse files
Files changed (20) hide show
  1. mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/gated_multi_branch/seed_44/all_results.json +15 -0
  2. mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/gated_multi_branch/seed_44/eval_results.json +9 -0
  3. mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/gated_multi_branch/seed_44/run_command.txt +1 -0
  4. mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/gated_multi_branch/seed_44/train_results.json +9 -0
  5. mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/gated_multi_branch/seed_44/trainer_state.json +216 -0
  6. mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_42/all_results.json +15 -0
  7. mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_42/eval_results.json +9 -0
  8. mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_42/run_command.txt +1 -0
  9. mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_42/train_results.json +9 -0
  10. mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_42/trainer_state.json +216 -0
  11. mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_43/all_results.json +15 -0
  12. mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_43/eval_results.json +9 -0
  13. mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_43/run_command.txt +1 -0
  14. mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_43/train_results.json +9 -0
  15. mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_43/trainer_state.json +216 -0
  16. mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_44/all_results.json +15 -0
  17. mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_44/eval_results.json +9 -0
  18. mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_44/run_command.txt +1 -0
  19. mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_44/train_results.json +9 -0
  20. mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_44/trainer_state.json +216 -0
mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/gated_multi_branch/seed_44/all_results.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "eval_loss": 0.30894502997398376,
4
+ "eval_matthews_correlation": 0.780066626120282,
5
+ "eval_runtime": 1.3575,
6
+ "eval_samples": 1043,
7
+ "eval_samples_per_second": 768.336,
8
+ "eval_steps_per_second": 8.103,
9
+ "total_flos": 406258112236284.0,
10
+ "train_loss": 0.22649634483806247,
11
+ "train_runtime": 65.5442,
12
+ "train_samples": 8551,
13
+ "train_samples_per_second": 391.385,
14
+ "train_steps_per_second": 8.193
15
+ }
mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/gated_multi_branch/seed_44/eval_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "eval_loss": 0.30894502997398376,
4
+ "eval_matthews_correlation": 0.780066626120282,
5
+ "eval_runtime": 1.3575,
6
+ "eval_samples": 1043,
7
+ "eval_samples_per_second": 768.336,
8
+ "eval_steps_per_second": 8.103
9
+ }
mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/gated_multi_branch/seed_44/run_command.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ /usr/local/bin/python /workspace/run_glue_MHA_gated.py --model_name_or_path bert-base-multilingual-cased --dataset_name tmnam20/ViGLUE --task_name cola --max_seq_length 128 --do_train --do_eval --evaluation_strategy epoch --save_strategy no --logging_strategy steps --logging_steps 25 --report_to none --num_train_epochs 3 --learning_rate 3e-05 --per_device_train_batch_size 48 --per_device_eval_batch_size 96 --gradient_accumulation_steps 1 --dataloader_num_workers 1 --seed 44 --output_dir result_ablation_mbert/cola/mBERT/gated_multi_branch/seed_44 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save --use_mha true --pooling_strategy gated_multi_branch --n_heads 4 --head_dim 32 --classifier_hidden_dim 512 --head_dropout 0.05
mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/gated_multi_branch/seed_44/train_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "total_flos": 406258112236284.0,
4
+ "train_loss": 0.22649634483806247,
5
+ "train_runtime": 65.5442,
6
+ "train_samples": 8551,
7
+ "train_samples_per_second": 391.385,
8
+ "train_steps_per_second": 8.193
9
+ }
mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/gated_multi_branch/seed_44/trainer_state.json ADDED
@@ -0,0 +1,216 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 3.0,
5
+ "eval_steps": 500,
6
+ "global_step": 537,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.13966480446927373,
13
+ "grad_norm": 7.894700050354004,
14
+ "learning_rate": 2.8603351955307262e-05,
15
+ "loss": 0.5224,
16
+ "step": 25
17
+ },
18
+ {
19
+ "epoch": 0.27932960893854747,
20
+ "grad_norm": 3.6509549617767334,
21
+ "learning_rate": 2.7206703910614527e-05,
22
+ "loss": 0.4222,
23
+ "step": 50
24
+ },
25
+ {
26
+ "epoch": 0.41899441340782123,
27
+ "grad_norm": 4.378454685211182,
28
+ "learning_rate": 2.5810055865921788e-05,
29
+ "loss": 0.3637,
30
+ "step": 75
31
+ },
32
+ {
33
+ "epoch": 0.5586592178770949,
34
+ "grad_norm": 4.795331001281738,
35
+ "learning_rate": 2.441340782122905e-05,
36
+ "loss": 0.3152,
37
+ "step": 100
38
+ },
39
+ {
40
+ "epoch": 0.6983240223463687,
41
+ "grad_norm": 6.246341705322266,
42
+ "learning_rate": 2.3016759776536314e-05,
43
+ "loss": 0.2961,
44
+ "step": 125
45
+ },
46
+ {
47
+ "epoch": 0.8379888268156425,
48
+ "grad_norm": 4.827692031860352,
49
+ "learning_rate": 2.1620111731843575e-05,
50
+ "loss": 0.3178,
51
+ "step": 150
52
+ },
53
+ {
54
+ "epoch": 0.9776536312849162,
55
+ "grad_norm": 7.820989608764648,
56
+ "learning_rate": 2.0223463687150836e-05,
57
+ "loss": 0.3059,
58
+ "step": 175
59
+ },
60
+ {
61
+ "epoch": 1.0,
62
+ "eval_loss": 0.24464663863182068,
63
+ "eval_matthews_correlation": 0.7618656865208808,
64
+ "eval_runtime": 2.6641,
65
+ "eval_samples_per_second": 391.497,
66
+ "eval_steps_per_second": 4.129,
67
+ "step": 179
68
+ },
69
+ {
70
+ "epoch": 1.1173184357541899,
71
+ "grad_norm": 6.193721294403076,
72
+ "learning_rate": 1.88268156424581e-05,
73
+ "loss": 0.2139,
74
+ "step": 200
75
+ },
76
+ {
77
+ "epoch": 1.2569832402234637,
78
+ "grad_norm": 3.6710243225097656,
79
+ "learning_rate": 1.7430167597765365e-05,
80
+ "loss": 0.2027,
81
+ "step": 225
82
+ },
83
+ {
84
+ "epoch": 1.3966480446927374,
85
+ "grad_norm": 8.232946395874023,
86
+ "learning_rate": 1.6033519553072626e-05,
87
+ "loss": 0.216,
88
+ "step": 250
89
+ },
90
+ {
91
+ "epoch": 1.536312849162011,
92
+ "grad_norm": 4.761744499206543,
93
+ "learning_rate": 1.4636871508379887e-05,
94
+ "loss": 0.1966,
95
+ "step": 275
96
+ },
97
+ {
98
+ "epoch": 1.675977653631285,
99
+ "grad_norm": 4.543158531188965,
100
+ "learning_rate": 1.324022346368715e-05,
101
+ "loss": 0.1907,
102
+ "step": 300
103
+ },
104
+ {
105
+ "epoch": 1.8156424581005588,
106
+ "grad_norm": 4.386110782623291,
107
+ "learning_rate": 1.1843575418994415e-05,
108
+ "loss": 0.1912,
109
+ "step": 325
110
+ },
111
+ {
112
+ "epoch": 1.9553072625698324,
113
+ "grad_norm": 7.197762966156006,
114
+ "learning_rate": 1.0446927374301676e-05,
115
+ "loss": 0.2148,
116
+ "step": 350
117
+ },
118
+ {
119
+ "epoch": 2.0,
120
+ "eval_loss": 0.26870936155319214,
121
+ "eval_matthews_correlation": 0.7691368380266224,
122
+ "eval_runtime": 1.7555,
123
+ "eval_samples_per_second": 594.136,
124
+ "eval_steps_per_second": 6.266,
125
+ "step": 358
126
+ },
127
+ {
128
+ "epoch": 2.094972067039106,
129
+ "grad_norm": 9.516730308532715,
130
+ "learning_rate": 9.050279329608939e-06,
131
+ "loss": 0.144,
132
+ "step": 375
133
+ },
134
+ {
135
+ "epoch": 2.2346368715083798,
136
+ "grad_norm": 8.45940113067627,
137
+ "learning_rate": 7.653631284916202e-06,
138
+ "loss": 0.1296,
139
+ "step": 400
140
+ },
141
+ {
142
+ "epoch": 2.3743016759776534,
143
+ "grad_norm": 4.391194820404053,
144
+ "learning_rate": 6.256983240223464e-06,
145
+ "loss": 0.1091,
146
+ "step": 425
147
+ },
148
+ {
149
+ "epoch": 2.5139664804469275,
150
+ "grad_norm": 4.410822868347168,
151
+ "learning_rate": 4.860335195530726e-06,
152
+ "loss": 0.1136,
153
+ "step": 450
154
+ },
155
+ {
156
+ "epoch": 2.653631284916201,
157
+ "grad_norm": 10.327372550964355,
158
+ "learning_rate": 3.463687150837989e-06,
159
+ "loss": 0.1195,
160
+ "step": 475
161
+ },
162
+ {
163
+ "epoch": 2.793296089385475,
164
+ "grad_norm": 1.6035041809082031,
165
+ "learning_rate": 2.0670391061452514e-06,
166
+ "loss": 0.1084,
167
+ "step": 500
168
+ },
169
+ {
170
+ "epoch": 2.9329608938547485,
171
+ "grad_norm": 5.493995666503906,
172
+ "learning_rate": 6.70391061452514e-07,
173
+ "loss": 0.1034,
174
+ "step": 525
175
+ },
176
+ {
177
+ "epoch": 3.0,
178
+ "eval_loss": 0.30894502997398376,
179
+ "eval_matthews_correlation": 0.780066626120282,
180
+ "eval_runtime": 1.2801,
181
+ "eval_samples_per_second": 814.799,
182
+ "eval_steps_per_second": 8.593,
183
+ "step": 537
184
+ },
185
+ {
186
+ "epoch": 3.0,
187
+ "step": 537,
188
+ "total_flos": 406258112236284.0,
189
+ "train_loss": 0.22649634483806247,
190
+ "train_runtime": 65.5442,
191
+ "train_samples_per_second": 391.385,
192
+ "train_steps_per_second": 8.193
193
+ }
194
+ ],
195
+ "logging_steps": 25,
196
+ "max_steps": 537,
197
+ "num_input_tokens_seen": 0,
198
+ "num_train_epochs": 3,
199
+ "save_steps": 500,
200
+ "stateful_callbacks": {
201
+ "TrainerControl": {
202
+ "args": {
203
+ "should_epoch_stop": false,
204
+ "should_evaluate": false,
205
+ "should_log": false,
206
+ "should_save": false,
207
+ "should_training_stop": false
208
+ },
209
+ "attributes": {}
210
+ }
211
+ },
212
+ "total_flos": 406258112236284.0,
213
+ "train_batch_size": 48,
214
+ "trial_name": null,
215
+ "trial_params": null
216
+ }
mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_42/all_results.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "eval_loss": 0.2610536515712738,
4
+ "eval_matthews_correlation": 0.7848625278330646,
5
+ "eval_runtime": 1.4937,
6
+ "eval_samples": 1043,
7
+ "eval_samples_per_second": 698.244,
8
+ "eval_steps_per_second": 7.364,
9
+ "total_flos": 393226666678440.0,
10
+ "train_loss": 0.27871723547994093,
11
+ "train_runtime": 70.8579,
12
+ "train_samples": 8551,
13
+ "train_samples_per_second": 362.034,
14
+ "train_steps_per_second": 7.579
15
+ }
mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_42/eval_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "eval_loss": 0.2610536515712738,
4
+ "eval_matthews_correlation": 0.7848625278330646,
5
+ "eval_runtime": 1.4937,
6
+ "eval_samples": 1043,
7
+ "eval_samples_per_second": 698.244,
8
+ "eval_steps_per_second": 7.364
9
+ }
mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_42/run_command.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ /usr/local/bin/python /workspace/run_glue_MHA_gated.py --model_name_or_path bert-base-multilingual-cased --dataset_name tmnam20/ViGLUE --task_name cola --max_seq_length 128 --do_train --do_eval --evaluation_strategy epoch --save_strategy no --logging_strategy steps --logging_steps 25 --report_to none --num_train_epochs 3 --learning_rate 3e-05 --per_device_train_batch_size 48 --per_device_eval_batch_size 96 --gradient_accumulation_steps 1 --dataloader_num_workers 1 --seed 42 --output_dir result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_42 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save
mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_42/train_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "total_flos": 393226666678440.0,
4
+ "train_loss": 0.27871723547994093,
5
+ "train_runtime": 70.8579,
6
+ "train_samples": 8551,
7
+ "train_samples_per_second": 362.034,
8
+ "train_steps_per_second": 7.579
9
+ }
mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_42/trainer_state.json ADDED
@@ -0,0 +1,216 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 3.0,
5
+ "eval_steps": 500,
6
+ "global_step": 537,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.13966480446927373,
13
+ "grad_norm": 20.55608367919922,
14
+ "learning_rate": 2.8603351955307262e-05,
15
+ "loss": 0.624,
16
+ "step": 25
17
+ },
18
+ {
19
+ "epoch": 0.27932960893854747,
20
+ "grad_norm": 4.743398666381836,
21
+ "learning_rate": 2.7206703910614527e-05,
22
+ "loss": 0.5735,
23
+ "step": 50
24
+ },
25
+ {
26
+ "epoch": 0.41899441340782123,
27
+ "grad_norm": 5.640482425689697,
28
+ "learning_rate": 2.5810055865921788e-05,
29
+ "loss": 0.4837,
30
+ "step": 75
31
+ },
32
+ {
33
+ "epoch": 0.5586592178770949,
34
+ "grad_norm": 5.729703903198242,
35
+ "learning_rate": 2.441340782122905e-05,
36
+ "loss": 0.3789,
37
+ "step": 100
38
+ },
39
+ {
40
+ "epoch": 0.6983240223463687,
41
+ "grad_norm": 5.261131286621094,
42
+ "learning_rate": 2.3016759776536314e-05,
43
+ "loss": 0.3747,
44
+ "step": 125
45
+ },
46
+ {
47
+ "epoch": 0.8379888268156425,
48
+ "grad_norm": 6.10923957824707,
49
+ "learning_rate": 2.1620111731843575e-05,
50
+ "loss": 0.3161,
51
+ "step": 150
52
+ },
53
+ {
54
+ "epoch": 0.9776536312849162,
55
+ "grad_norm": 7.458263397216797,
56
+ "learning_rate": 2.0223463687150836e-05,
57
+ "loss": 0.3266,
58
+ "step": 175
59
+ },
60
+ {
61
+ "epoch": 1.0,
62
+ "eval_loss": 0.26663538813591003,
63
+ "eval_matthews_correlation": 0.7666658377335647,
64
+ "eval_runtime": 1.9975,
65
+ "eval_samples_per_second": 522.147,
66
+ "eval_steps_per_second": 5.507,
67
+ "step": 179
68
+ },
69
+ {
70
+ "epoch": 1.1173184357541899,
71
+ "grad_norm": 5.076351642608643,
72
+ "learning_rate": 1.88268156424581e-05,
73
+ "loss": 0.3197,
74
+ "step": 200
75
+ },
76
+ {
77
+ "epoch": 1.2569832402234637,
78
+ "grad_norm": 6.497809886932373,
79
+ "learning_rate": 1.7430167597765365e-05,
80
+ "loss": 0.2894,
81
+ "step": 225
82
+ },
83
+ {
84
+ "epoch": 1.3966480446927374,
85
+ "grad_norm": 3.972665548324585,
86
+ "learning_rate": 1.6033519553072626e-05,
87
+ "loss": 0.247,
88
+ "step": 250
89
+ },
90
+ {
91
+ "epoch": 1.536312849162011,
92
+ "grad_norm": 8.7925386428833,
93
+ "learning_rate": 1.4636871508379887e-05,
94
+ "loss": 0.2082,
95
+ "step": 275
96
+ },
97
+ {
98
+ "epoch": 1.675977653631285,
99
+ "grad_norm": 7.290075302124023,
100
+ "learning_rate": 1.324022346368715e-05,
101
+ "loss": 0.2508,
102
+ "step": 300
103
+ },
104
+ {
105
+ "epoch": 1.8156424581005588,
106
+ "grad_norm": 6.401366710662842,
107
+ "learning_rate": 1.1843575418994415e-05,
108
+ "loss": 0.2202,
109
+ "step": 325
110
+ },
111
+ {
112
+ "epoch": 1.9553072625698324,
113
+ "grad_norm": 6.282566547393799,
114
+ "learning_rate": 1.0446927374301676e-05,
115
+ "loss": 0.2297,
116
+ "step": 350
117
+ },
118
+ {
119
+ "epoch": 2.0,
120
+ "eval_loss": 0.2509293258190155,
121
+ "eval_matthews_correlation": 0.7573297015336248,
122
+ "eval_runtime": 1.5516,
123
+ "eval_samples_per_second": 672.19,
124
+ "eval_steps_per_second": 7.089,
125
+ "step": 358
126
+ },
127
+ {
128
+ "epoch": 2.094972067039106,
129
+ "grad_norm": 5.3574724197387695,
130
+ "learning_rate": 9.050279329608939e-06,
131
+ "loss": 0.2098,
132
+ "step": 375
133
+ },
134
+ {
135
+ "epoch": 2.2346368715083798,
136
+ "grad_norm": 11.248693466186523,
137
+ "learning_rate": 7.653631284916202e-06,
138
+ "loss": 0.1342,
139
+ "step": 400
140
+ },
141
+ {
142
+ "epoch": 2.3743016759776534,
143
+ "grad_norm": 19.597501754760742,
144
+ "learning_rate": 6.256983240223464e-06,
145
+ "loss": 0.1674,
146
+ "step": 425
147
+ },
148
+ {
149
+ "epoch": 2.5139664804469275,
150
+ "grad_norm": 9.087995529174805,
151
+ "learning_rate": 4.860335195530726e-06,
152
+ "loss": 0.1413,
153
+ "step": 450
154
+ },
155
+ {
156
+ "epoch": 2.653631284916201,
157
+ "grad_norm": 6.359992980957031,
158
+ "learning_rate": 3.463687150837989e-06,
159
+ "loss": 0.1596,
160
+ "step": 475
161
+ },
162
+ {
163
+ "epoch": 2.793296089385475,
164
+ "grad_norm": 7.049188137054443,
165
+ "learning_rate": 2.0670391061452514e-06,
166
+ "loss": 0.1406,
167
+ "step": 500
168
+ },
169
+ {
170
+ "epoch": 2.9329608938547485,
171
+ "grad_norm": 7.78603458404541,
172
+ "learning_rate": 6.70391061452514e-07,
173
+ "loss": 0.1308,
174
+ "step": 525
175
+ },
176
+ {
177
+ "epoch": 3.0,
178
+ "eval_loss": 0.2610536515712738,
179
+ "eval_matthews_correlation": 0.7848625278330646,
180
+ "eval_runtime": 1.5752,
181
+ "eval_samples_per_second": 662.125,
182
+ "eval_steps_per_second": 6.983,
183
+ "step": 537
184
+ },
185
+ {
186
+ "epoch": 3.0,
187
+ "step": 537,
188
+ "total_flos": 393226666678440.0,
189
+ "train_loss": 0.27871723547994093,
190
+ "train_runtime": 70.8579,
191
+ "train_samples_per_second": 362.034,
192
+ "train_steps_per_second": 7.579
193
+ }
194
+ ],
195
+ "logging_steps": 25,
196
+ "max_steps": 537,
197
+ "num_input_tokens_seen": 0,
198
+ "num_train_epochs": 3,
199
+ "save_steps": 500,
200
+ "stateful_callbacks": {
201
+ "TrainerControl": {
202
+ "args": {
203
+ "should_epoch_stop": false,
204
+ "should_evaluate": false,
205
+ "should_log": false,
206
+ "should_save": false,
207
+ "should_training_stop": false
208
+ },
209
+ "attributes": {}
210
+ }
211
+ },
212
+ "total_flos": 393226666678440.0,
213
+ "train_batch_size": 48,
214
+ "trial_name": null,
215
+ "trial_params": null
216
+ }
mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_43/all_results.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "eval_loss": 0.31756097078323364,
4
+ "eval_matthews_correlation": 0.7734460694651355,
5
+ "eval_runtime": 1.4405,
6
+ "eval_samples": 1043,
7
+ "eval_samples_per_second": 724.074,
8
+ "eval_steps_per_second": 7.636,
9
+ "total_flos": 390717347765700.0,
10
+ "train_loss": 0.24823517306556914,
11
+ "train_runtime": 70.5669,
12
+ "train_samples": 8551,
13
+ "train_samples_per_second": 363.527,
14
+ "train_steps_per_second": 7.61
15
+ }
mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_43/eval_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "eval_loss": 0.31756097078323364,
4
+ "eval_matthews_correlation": 0.7734460694651355,
5
+ "eval_runtime": 1.4405,
6
+ "eval_samples": 1043,
7
+ "eval_samples_per_second": 724.074,
8
+ "eval_steps_per_second": 7.636
9
+ }
mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_43/run_command.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ /usr/local/bin/python /workspace/run_glue_MHA_gated.py --model_name_or_path bert-base-multilingual-cased --dataset_name tmnam20/ViGLUE --task_name cola --max_seq_length 128 --do_train --do_eval --evaluation_strategy epoch --save_strategy no --logging_strategy steps --logging_steps 25 --report_to none --num_train_epochs 3 --learning_rate 3e-05 --per_device_train_batch_size 48 --per_device_eval_batch_size 96 --gradient_accumulation_steps 1 --dataloader_num_workers 1 --seed 43 --output_dir result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_43 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save
mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_43/train_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "total_flos": 390717347765700.0,
4
+ "train_loss": 0.24823517306556914,
5
+ "train_runtime": 70.5669,
6
+ "train_samples": 8551,
7
+ "train_samples_per_second": 363.527,
8
+ "train_steps_per_second": 7.61
9
+ }
mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_43/trainer_state.json ADDED
@@ -0,0 +1,216 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 3.0,
5
+ "eval_steps": 500,
6
+ "global_step": 537,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.13966480446927373,
13
+ "grad_norm": 6.581408977508545,
14
+ "learning_rate": 2.8603351955307262e-05,
15
+ "loss": 0.5636,
16
+ "step": 25
17
+ },
18
+ {
19
+ "epoch": 0.27932960893854747,
20
+ "grad_norm": 11.804716110229492,
21
+ "learning_rate": 2.7206703910614527e-05,
22
+ "loss": 0.4382,
23
+ "step": 50
24
+ },
25
+ {
26
+ "epoch": 0.41899441340782123,
27
+ "grad_norm": 4.38754940032959,
28
+ "learning_rate": 2.5810055865921788e-05,
29
+ "loss": 0.3702,
30
+ "step": 75
31
+ },
32
+ {
33
+ "epoch": 0.5586592178770949,
34
+ "grad_norm": 5.732598781585693,
35
+ "learning_rate": 2.441340782122905e-05,
36
+ "loss": 0.3414,
37
+ "step": 100
38
+ },
39
+ {
40
+ "epoch": 0.6983240223463687,
41
+ "grad_norm": 3.4510881900787354,
42
+ "learning_rate": 2.3016759776536314e-05,
43
+ "loss": 0.3388,
44
+ "step": 125
45
+ },
46
+ {
47
+ "epoch": 0.8379888268156425,
48
+ "grad_norm": 6.2728729248046875,
49
+ "learning_rate": 2.1620111731843575e-05,
50
+ "loss": 0.3279,
51
+ "step": 150
52
+ },
53
+ {
54
+ "epoch": 0.9776536312849162,
55
+ "grad_norm": 4.220879077911377,
56
+ "learning_rate": 2.0223463687150836e-05,
57
+ "loss": 0.3057,
58
+ "step": 175
59
+ },
60
+ {
61
+ "epoch": 1.0,
62
+ "eval_loss": 0.2580081820487976,
63
+ "eval_matthews_correlation": 0.7712806700875474,
64
+ "eval_runtime": 1.9021,
65
+ "eval_samples_per_second": 548.336,
66
+ "eval_steps_per_second": 5.783,
67
+ "step": 179
68
+ },
69
+ {
70
+ "epoch": 1.1173184357541899,
71
+ "grad_norm": 4.043488025665283,
72
+ "learning_rate": 1.88268156424581e-05,
73
+ "loss": 0.2786,
74
+ "step": 200
75
+ },
76
+ {
77
+ "epoch": 1.2569832402234637,
78
+ "grad_norm": 5.739426612854004,
79
+ "learning_rate": 1.7430167597765365e-05,
80
+ "loss": 0.2455,
81
+ "step": 225
82
+ },
83
+ {
84
+ "epoch": 1.3966480446927374,
85
+ "grad_norm": 7.959273815155029,
86
+ "learning_rate": 1.6033519553072626e-05,
87
+ "loss": 0.2405,
88
+ "step": 250
89
+ },
90
+ {
91
+ "epoch": 1.536312849162011,
92
+ "grad_norm": 7.122917652130127,
93
+ "learning_rate": 1.4636871508379887e-05,
94
+ "loss": 0.2357,
95
+ "step": 275
96
+ },
97
+ {
98
+ "epoch": 1.675977653631285,
99
+ "grad_norm": 7.435409069061279,
100
+ "learning_rate": 1.324022346368715e-05,
101
+ "loss": 0.2042,
102
+ "step": 300
103
+ },
104
+ {
105
+ "epoch": 1.8156424581005588,
106
+ "grad_norm": 9.217028617858887,
107
+ "learning_rate": 1.1843575418994415e-05,
108
+ "loss": 0.1912,
109
+ "step": 325
110
+ },
111
+ {
112
+ "epoch": 1.9553072625698324,
113
+ "grad_norm": 7.499634265899658,
114
+ "learning_rate": 1.0446927374301676e-05,
115
+ "loss": 0.2048,
116
+ "step": 350
117
+ },
118
+ {
119
+ "epoch": 2.0,
120
+ "eval_loss": 0.24817638099193573,
121
+ "eval_matthews_correlation": 0.7707964764354,
122
+ "eval_runtime": 1.4532,
123
+ "eval_samples_per_second": 717.75,
124
+ "eval_steps_per_second": 7.57,
125
+ "step": 358
126
+ },
127
+ {
128
+ "epoch": 2.094972067039106,
129
+ "grad_norm": 3.717542886734009,
130
+ "learning_rate": 9.050279329608939e-06,
131
+ "loss": 0.1749,
132
+ "step": 375
133
+ },
134
+ {
135
+ "epoch": 2.2346368715083798,
136
+ "grad_norm": 2.3178093433380127,
137
+ "learning_rate": 7.653631284916202e-06,
138
+ "loss": 0.1468,
139
+ "step": 400
140
+ },
141
+ {
142
+ "epoch": 2.3743016759776534,
143
+ "grad_norm": 7.624992370605469,
144
+ "learning_rate": 6.256983240223464e-06,
145
+ "loss": 0.1287,
146
+ "step": 425
147
+ },
148
+ {
149
+ "epoch": 2.5139664804469275,
150
+ "grad_norm": 9.634353637695312,
151
+ "learning_rate": 4.860335195530726e-06,
152
+ "loss": 0.1377,
153
+ "step": 450
154
+ },
155
+ {
156
+ "epoch": 2.653631284916201,
157
+ "grad_norm": 5.767144680023193,
158
+ "learning_rate": 3.463687150837989e-06,
159
+ "loss": 0.1396,
160
+ "step": 475
161
+ },
162
+ {
163
+ "epoch": 2.793296089385475,
164
+ "grad_norm": 6.6265459060668945,
165
+ "learning_rate": 2.0670391061452514e-06,
166
+ "loss": 0.1309,
167
+ "step": 500
168
+ },
169
+ {
170
+ "epoch": 2.9329608938547485,
171
+ "grad_norm": 8.028974533081055,
172
+ "learning_rate": 6.70391061452514e-07,
173
+ "loss": 0.1325,
174
+ "step": 525
175
+ },
176
+ {
177
+ "epoch": 3.0,
178
+ "eval_loss": 0.31756097078323364,
179
+ "eval_matthews_correlation": 0.7734460694651355,
180
+ "eval_runtime": 1.4279,
181
+ "eval_samples_per_second": 730.454,
182
+ "eval_steps_per_second": 7.704,
183
+ "step": 537
184
+ },
185
+ {
186
+ "epoch": 3.0,
187
+ "step": 537,
188
+ "total_flos": 390717347765700.0,
189
+ "train_loss": 0.24823517306556914,
190
+ "train_runtime": 70.5669,
191
+ "train_samples_per_second": 363.527,
192
+ "train_steps_per_second": 7.61
193
+ }
194
+ ],
195
+ "logging_steps": 25,
196
+ "max_steps": 537,
197
+ "num_input_tokens_seen": 0,
198
+ "num_train_epochs": 3,
199
+ "save_steps": 500,
200
+ "stateful_callbacks": {
201
+ "TrainerControl": {
202
+ "args": {
203
+ "should_epoch_stop": false,
204
+ "should_evaluate": false,
205
+ "should_log": false,
206
+ "should_save": false,
207
+ "should_training_stop": false
208
+ },
209
+ "attributes": {}
210
+ }
211
+ },
212
+ "total_flos": 390717347765700.0,
213
+ "train_batch_size": 48,
214
+ "trial_name": null,
215
+ "trial_params": null
216
+ }
mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_44/all_results.json ADDED
@@ -0,0 +1,15 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "eval_loss": 0.2951465845108032,
4
+ "eval_matthews_correlation": 0.7800284888427224,
5
+ "eval_runtime": 1.4836,
6
+ "eval_samples": 1043,
7
+ "eval_samples_per_second": 703.013,
8
+ "eval_steps_per_second": 7.414,
9
+ "total_flos": 391841736416340.0,
10
+ "train_loss": 0.23531772815093427,
11
+ "train_runtime": 70.7996,
12
+ "train_samples": 8551,
13
+ "train_samples_per_second": 362.333,
14
+ "train_steps_per_second": 7.585
15
+ }
mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_44/eval_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "eval_loss": 0.2951465845108032,
4
+ "eval_matthews_correlation": 0.7800284888427224,
5
+ "eval_runtime": 1.4836,
6
+ "eval_samples": 1043,
7
+ "eval_samples_per_second": 703.013,
8
+ "eval_steps_per_second": 7.414
9
+ }
mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_44/run_command.txt ADDED
@@ -0,0 +1 @@
 
 
1
+ /usr/local/bin/python /workspace/run_glue_MHA_gated.py --model_name_or_path bert-base-multilingual-cased --dataset_name tmnam20/ViGLUE --task_name cola --max_seq_length 128 --do_train --do_eval --evaluation_strategy epoch --save_strategy no --logging_strategy steps --logging_steps 25 --report_to none --num_train_epochs 3 --learning_rate 3e-05 --per_device_train_batch_size 48 --per_device_eval_batch_size 96 --gradient_accumulation_steps 1 --dataloader_num_workers 1 --seed 44 --output_dir result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_44 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save
mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_44/train_results.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "epoch": 3.0,
3
+ "total_flos": 391841736416340.0,
4
+ "train_loss": 0.23531772815093427,
5
+ "train_runtime": 70.7996,
6
+ "train_samples": 8551,
7
+ "train_samples_per_second": 362.333,
8
+ "train_steps_per_second": 7.585
9
+ }
mbert_rtx6000_metrics/result_ablation_mbert/cola/mBERT/hf_sequence_classifier/seed_44/trainer_state.json ADDED
@@ -0,0 +1,216 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "best_metric": null,
3
+ "best_model_checkpoint": null,
4
+ "epoch": 3.0,
5
+ "eval_steps": 500,
6
+ "global_step": 537,
7
+ "is_hyper_param_search": false,
8
+ "is_local_process_zero": true,
9
+ "is_world_process_zero": true,
10
+ "log_history": [
11
+ {
12
+ "epoch": 0.13966480446927373,
13
+ "grad_norm": 6.386740684509277,
14
+ "learning_rate": 2.8603351955307262e-05,
15
+ "loss": 0.5258,
16
+ "step": 25
17
+ },
18
+ {
19
+ "epoch": 0.27932960893854747,
20
+ "grad_norm": 3.3861515522003174,
21
+ "learning_rate": 2.7206703910614527e-05,
22
+ "loss": 0.4253,
23
+ "step": 50
24
+ },
25
+ {
26
+ "epoch": 0.41899441340782123,
27
+ "grad_norm": 3.794750213623047,
28
+ "learning_rate": 2.5810055865921788e-05,
29
+ "loss": 0.3852,
30
+ "step": 75
31
+ },
32
+ {
33
+ "epoch": 0.5586592178770949,
34
+ "grad_norm": 3.815369129180908,
35
+ "learning_rate": 2.441340782122905e-05,
36
+ "loss": 0.336,
37
+ "step": 100
38
+ },
39
+ {
40
+ "epoch": 0.6983240223463687,
41
+ "grad_norm": 7.895872592926025,
42
+ "learning_rate": 2.3016759776536314e-05,
43
+ "loss": 0.3274,
44
+ "step": 125
45
+ },
46
+ {
47
+ "epoch": 0.8379888268156425,
48
+ "grad_norm": 7.099031448364258,
49
+ "learning_rate": 2.1620111731843575e-05,
50
+ "loss": 0.3144,
51
+ "step": 150
52
+ },
53
+ {
54
+ "epoch": 0.9776536312849162,
55
+ "grad_norm": 6.187333583831787,
56
+ "learning_rate": 2.0223463687150836e-05,
57
+ "loss": 0.3006,
58
+ "step": 175
59
+ },
60
+ {
61
+ "epoch": 1.0,
62
+ "eval_loss": 0.23891180753707886,
63
+ "eval_matthews_correlation": 0.773663670490989,
64
+ "eval_runtime": 1.9407,
65
+ "eval_samples_per_second": 537.447,
66
+ "eval_steps_per_second": 5.668,
67
+ "step": 179
68
+ },
69
+ {
70
+ "epoch": 1.1173184357541899,
71
+ "grad_norm": 7.869960784912109,
72
+ "learning_rate": 1.88268156424581e-05,
73
+ "loss": 0.211,
74
+ "step": 200
75
+ },
76
+ {
77
+ "epoch": 1.2569832402234637,
78
+ "grad_norm": 8.176018714904785,
79
+ "learning_rate": 1.7430167597765365e-05,
80
+ "loss": 0.2436,
81
+ "step": 225
82
+ },
83
+ {
84
+ "epoch": 1.3966480446927374,
85
+ "grad_norm": 7.732117176055908,
86
+ "learning_rate": 1.6033519553072626e-05,
87
+ "loss": 0.2308,
88
+ "step": 250
89
+ },
90
+ {
91
+ "epoch": 1.536312849162011,
92
+ "grad_norm": 6.40506649017334,
93
+ "learning_rate": 1.4636871508379887e-05,
94
+ "loss": 0.2067,
95
+ "step": 275
96
+ },
97
+ {
98
+ "epoch": 1.675977653631285,
99
+ "grad_norm": 6.098687648773193,
100
+ "learning_rate": 1.324022346368715e-05,
101
+ "loss": 0.1937,
102
+ "step": 300
103
+ },
104
+ {
105
+ "epoch": 1.8156424581005588,
106
+ "grad_norm": 7.470464706420898,
107
+ "learning_rate": 1.1843575418994415e-05,
108
+ "loss": 0.203,
109
+ "step": 325
110
+ },
111
+ {
112
+ "epoch": 1.9553072625698324,
113
+ "grad_norm": 5.939926624298096,
114
+ "learning_rate": 1.0446927374301676e-05,
115
+ "loss": 0.2056,
116
+ "step": 350
117
+ },
118
+ {
119
+ "epoch": 2.0,
120
+ "eval_loss": 0.2881302237510681,
121
+ "eval_matthews_correlation": 0.7711205921878477,
122
+ "eval_runtime": 1.496,
123
+ "eval_samples_per_second": 697.192,
124
+ "eval_steps_per_second": 7.353,
125
+ "step": 358
126
+ },
127
+ {
128
+ "epoch": 2.094972067039106,
129
+ "grad_norm": 7.537599563598633,
130
+ "learning_rate": 9.050279329608939e-06,
131
+ "loss": 0.1373,
132
+ "step": 375
133
+ },
134
+ {
135
+ "epoch": 2.2346368715083798,
136
+ "grad_norm": 13.339438438415527,
137
+ "learning_rate": 7.653631284916202e-06,
138
+ "loss": 0.122,
139
+ "step": 400
140
+ },
141
+ {
142
+ "epoch": 2.3743016759776534,
143
+ "grad_norm": 7.579359531402588,
144
+ "learning_rate": 6.256983240223464e-06,
145
+ "loss": 0.1361,
146
+ "step": 425
147
+ },
148
+ {
149
+ "epoch": 2.5139664804469275,
150
+ "grad_norm": 8.117977142333984,
151
+ "learning_rate": 4.860335195530726e-06,
152
+ "loss": 0.1345,
153
+ "step": 450
154
+ },
155
+ {
156
+ "epoch": 2.653631284916201,
157
+ "grad_norm": 7.523886203765869,
158
+ "learning_rate": 3.463687150837989e-06,
159
+ "loss": 0.1269,
160
+ "step": 475
161
+ },
162
+ {
163
+ "epoch": 2.793296089385475,
164
+ "grad_norm": 6.03088903427124,
165
+ "learning_rate": 2.0670391061452514e-06,
166
+ "loss": 0.1163,
167
+ "step": 500
168
+ },
169
+ {
170
+ "epoch": 2.9329608938547485,
171
+ "grad_norm": 11.261451721191406,
172
+ "learning_rate": 6.70391061452514e-07,
173
+ "loss": 0.1222,
174
+ "step": 525
175
+ },
176
+ {
177
+ "epoch": 3.0,
178
+ "eval_loss": 0.2951465845108032,
179
+ "eval_matthews_correlation": 0.7800284888427224,
180
+ "eval_runtime": 1.5256,
181
+ "eval_samples_per_second": 683.659,
182
+ "eval_steps_per_second": 7.21,
183
+ "step": 537
184
+ },
185
+ {
186
+ "epoch": 3.0,
187
+ "step": 537,
188
+ "total_flos": 391841736416340.0,
189
+ "train_loss": 0.23531772815093427,
190
+ "train_runtime": 70.7996,
191
+ "train_samples_per_second": 362.333,
192
+ "train_steps_per_second": 7.585
193
+ }
194
+ ],
195
+ "logging_steps": 25,
196
+ "max_steps": 537,
197
+ "num_input_tokens_seen": 0,
198
+ "num_train_epochs": 3,
199
+ "save_steps": 500,
200
+ "stateful_callbacks": {
201
+ "TrainerControl": {
202
+ "args": {
203
+ "should_epoch_stop": false,
204
+ "should_evaluate": false,
205
+ "should_log": false,
206
+ "should_save": false,
207
+ "should_training_stop": false
208
+ },
209
+ "attributes": {}
210
+ }
211
+ },
212
+ "total_flos": 391841736416340.0,
213
+ "train_batch_size": 48,
214
+ "trial_name": null,
215
+ "trial_params": null
216
+ }