Upload batch 19/24: vsfc/mBERT/attention/seed_42 ... vsfc/mBERT/cls/seed_42
Browse files- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/attention/seed_42/run_command.txt +1 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/attention/seed_43/run_command.txt +1 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/attention/seed_44/all_results.json +15 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/attention/seed_44/eval_results.json +9 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/attention/seed_44/run_command.txt +1 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/attention/seed_44/train_results.json +9 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/attention/seed_44/trainer_state.json +265 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_42/all_results.json +15 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_42/eval_results.json +9 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_42/run_command.txt +1 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_42/train_results.json +9 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_42/trainer_state.json +265 -0
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/attention/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 vsfc --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/vsfc/mBERT/attention/seed_42 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save --use_mha true --pooling_strategy attention --n_heads 4 --head_dim 32 --classifier_hidden_dim 512 --head_dropout 0.05
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/attention/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 vsfc --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/vsfc/mBERT/attention/seed_43 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save --use_mha true --pooling_strategy attention --n_heads 4 --head_dim 32 --classifier_hidden_dim 512 --head_dropout 0.05
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/attention/seed_44/all_results.json
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.9330385344283006,
|
| 4 |
+
"eval_loss": 0.2232467234134674,
|
| 5 |
+
"eval_runtime": 6.6813,
|
| 6 |
+
"eval_samples": 1583,
|
| 7 |
+
"eval_samples_per_second": 236.931,
|
| 8 |
+
"eval_steps_per_second": 2.544,
|
| 9 |
+
"total_flos": 1024714859387040.0,
|
| 10 |
+
"train_loss": 0.2006188094034022,
|
| 11 |
+
"train_runtime": 91.8654,
|
| 12 |
+
"train_samples": 11426,
|
| 13 |
+
"train_samples_per_second": 373.133,
|
| 14 |
+
"train_steps_per_second": 7.805
|
| 15 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/attention/seed_44/eval_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.9330385344283006,
|
| 4 |
+
"eval_loss": 0.2232467234134674,
|
| 5 |
+
"eval_runtime": 6.6813,
|
| 6 |
+
"eval_samples": 1583,
|
| 7 |
+
"eval_samples_per_second": 236.931,
|
| 8 |
+
"eval_steps_per_second": 2.544
|
| 9 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/attention/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 vsfc --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/vsfc/mBERT/attention/seed_44 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save --use_mha true --pooling_strategy attention --n_heads 4 --head_dim 32 --classifier_hidden_dim 512 --head_dropout 0.05
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/attention/seed_44/train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 1024714859387040.0,
|
| 4 |
+
"train_loss": 0.2006188094034022,
|
| 5 |
+
"train_runtime": 91.8654,
|
| 6 |
+
"train_samples": 11426,
|
| 7 |
+
"train_samples_per_second": 373.133,
|
| 8 |
+
"train_steps_per_second": 7.805
|
| 9 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/attention/seed_44/trainer_state.json
ADDED
|
@@ -0,0 +1,265 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 3.0,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 717,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.10460251046025104,
|
| 13 |
+
"grad_norm": 5.092621803283691,
|
| 14 |
+
"learning_rate": 2.8953974895397488e-05,
|
| 15 |
+
"loss": 0.5431,
|
| 16 |
+
"step": 25
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 0.20920502092050208,
|
| 20 |
+
"grad_norm": 6.562098503112793,
|
| 21 |
+
"learning_rate": 2.7907949790794982e-05,
|
| 22 |
+
"loss": 0.3397,
|
| 23 |
+
"step": 50
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.3138075313807531,
|
| 27 |
+
"grad_norm": 3.2435336112976074,
|
| 28 |
+
"learning_rate": 2.686192468619247e-05,
|
| 29 |
+
"loss": 0.3356,
|
| 30 |
+
"step": 75
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.41841004184100417,
|
| 34 |
+
"grad_norm": 2.5668106079101562,
|
| 35 |
+
"learning_rate": 2.5815899581589957e-05,
|
| 36 |
+
"loss": 0.2773,
|
| 37 |
+
"step": 100
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.5230125523012552,
|
| 41 |
+
"grad_norm": 7.244137287139893,
|
| 42 |
+
"learning_rate": 2.4769874476987448e-05,
|
| 43 |
+
"loss": 0.2829,
|
| 44 |
+
"step": 125
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.6276150627615062,
|
| 48 |
+
"grad_norm": 4.513981819152832,
|
| 49 |
+
"learning_rate": 2.372384937238494e-05,
|
| 50 |
+
"loss": 0.2524,
|
| 51 |
+
"step": 150
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.7322175732217573,
|
| 55 |
+
"grad_norm": 6.451638221740723,
|
| 56 |
+
"learning_rate": 2.2677824267782426e-05,
|
| 57 |
+
"loss": 0.2787,
|
| 58 |
+
"step": 175
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.8368200836820083,
|
| 62 |
+
"grad_norm": 6.222054958343506,
|
| 63 |
+
"learning_rate": 2.1631799163179917e-05,
|
| 64 |
+
"loss": 0.2148,
|
| 65 |
+
"step": 200
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.9414225941422594,
|
| 69 |
+
"grad_norm": 5.222476482391357,
|
| 70 |
+
"learning_rate": 2.0585774058577408e-05,
|
| 71 |
+
"loss": 0.2219,
|
| 72 |
+
"step": 225
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 1.0,
|
| 76 |
+
"eval_accuracy": 0.9241945672773215,
|
| 77 |
+
"eval_loss": 0.23050804436206818,
|
| 78 |
+
"eval_runtime": 4.714,
|
| 79 |
+
"eval_samples_per_second": 335.808,
|
| 80 |
+
"eval_steps_per_second": 3.606,
|
| 81 |
+
"step": 239
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"epoch": 1.0460251046025104,
|
| 85 |
+
"grad_norm": 6.058961391448975,
|
| 86 |
+
"learning_rate": 1.9539748953974895e-05,
|
| 87 |
+
"loss": 0.215,
|
| 88 |
+
"step": 250
|
| 89 |
+
},
|
| 90 |
+
{
|
| 91 |
+
"epoch": 1.1506276150627615,
|
| 92 |
+
"grad_norm": 7.067775249481201,
|
| 93 |
+
"learning_rate": 1.8493723849372386e-05,
|
| 94 |
+
"loss": 0.1962,
|
| 95 |
+
"step": 275
|
| 96 |
+
},
|
| 97 |
+
{
|
| 98 |
+
"epoch": 1.2552301255230125,
|
| 99 |
+
"grad_norm": 2.362067461013794,
|
| 100 |
+
"learning_rate": 1.7447698744769877e-05,
|
| 101 |
+
"loss": 0.1828,
|
| 102 |
+
"step": 300
|
| 103 |
+
},
|
| 104 |
+
{
|
| 105 |
+
"epoch": 1.3598326359832635,
|
| 106 |
+
"grad_norm": 3.0100741386413574,
|
| 107 |
+
"learning_rate": 1.6401673640167365e-05,
|
| 108 |
+
"loss": 0.1821,
|
| 109 |
+
"step": 325
|
| 110 |
+
},
|
| 111 |
+
{
|
| 112 |
+
"epoch": 1.4644351464435146,
|
| 113 |
+
"grad_norm": 4.510914325714111,
|
| 114 |
+
"learning_rate": 1.5355648535564852e-05,
|
| 115 |
+
"loss": 0.1865,
|
| 116 |
+
"step": 350
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"epoch": 1.5690376569037658,
|
| 120 |
+
"grad_norm": 9.777588844299316,
|
| 121 |
+
"learning_rate": 1.4309623430962344e-05,
|
| 122 |
+
"loss": 0.1709,
|
| 123 |
+
"step": 375
|
| 124 |
+
},
|
| 125 |
+
{
|
| 126 |
+
"epoch": 1.6736401673640167,
|
| 127 |
+
"grad_norm": 5.089670658111572,
|
| 128 |
+
"learning_rate": 1.3263598326359832e-05,
|
| 129 |
+
"loss": 0.1787,
|
| 130 |
+
"step": 400
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 1.778242677824268,
|
| 134 |
+
"grad_norm": 4.752739906311035,
|
| 135 |
+
"learning_rate": 1.2217573221757323e-05,
|
| 136 |
+
"loss": 0.189,
|
| 137 |
+
"step": 425
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.8828451882845187,
|
| 141 |
+
"grad_norm": 4.979706287384033,
|
| 142 |
+
"learning_rate": 1.1171548117154812e-05,
|
| 143 |
+
"loss": 0.1885,
|
| 144 |
+
"step": 450
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 1.98744769874477,
|
| 148 |
+
"grad_norm": 7.023453235626221,
|
| 149 |
+
"learning_rate": 1.0125523012552301e-05,
|
| 150 |
+
"loss": 0.1608,
|
| 151 |
+
"step": 475
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"epoch": 2.0,
|
| 155 |
+
"eval_accuracy": 0.9311433986102338,
|
| 156 |
+
"eval_loss": 0.22642782330513,
|
| 157 |
+
"eval_runtime": 4.9222,
|
| 158 |
+
"eval_samples_per_second": 321.604,
|
| 159 |
+
"eval_steps_per_second": 3.454,
|
| 160 |
+
"step": 478
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 2.092050209205021,
|
| 164 |
+
"grad_norm": 10.432361602783203,
|
| 165 |
+
"learning_rate": 9.079497907949792e-06,
|
| 166 |
+
"loss": 0.1237,
|
| 167 |
+
"step": 500
|
| 168 |
+
},
|
| 169 |
+
{
|
| 170 |
+
"epoch": 2.196652719665272,
|
| 171 |
+
"grad_norm": 3.229215145111084,
|
| 172 |
+
"learning_rate": 8.03347280334728e-06,
|
| 173 |
+
"loss": 0.13,
|
| 174 |
+
"step": 525
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 2.301255230125523,
|
| 178 |
+
"grad_norm": 2.5336787700653076,
|
| 179 |
+
"learning_rate": 6.98744769874477e-06,
|
| 180 |
+
"loss": 0.1319,
|
| 181 |
+
"step": 550
|
| 182 |
+
},
|
| 183 |
+
{
|
| 184 |
+
"epoch": 2.405857740585774,
|
| 185 |
+
"grad_norm": 6.742815017700195,
|
| 186 |
+
"learning_rate": 5.941422594142259e-06,
|
| 187 |
+
"loss": 0.1182,
|
| 188 |
+
"step": 575
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"epoch": 2.510460251046025,
|
| 192 |
+
"grad_norm": 4.245293617248535,
|
| 193 |
+
"learning_rate": 4.895397489539749e-06,
|
| 194 |
+
"loss": 0.1134,
|
| 195 |
+
"step": 600
|
| 196 |
+
},
|
| 197 |
+
{
|
| 198 |
+
"epoch": 2.6150627615062763,
|
| 199 |
+
"grad_norm": 2.052689552307129,
|
| 200 |
+
"learning_rate": 3.849372384937238e-06,
|
| 201 |
+
"loss": 0.114,
|
| 202 |
+
"step": 625
|
| 203 |
+
},
|
| 204 |
+
{
|
| 205 |
+
"epoch": 2.719665271966527,
|
| 206 |
+
"grad_norm": 4.83400821685791,
|
| 207 |
+
"learning_rate": 2.803347280334728e-06,
|
| 208 |
+
"loss": 0.1385,
|
| 209 |
+
"step": 650
|
| 210 |
+
},
|
| 211 |
+
{
|
| 212 |
+
"epoch": 2.8242677824267783,
|
| 213 |
+
"grad_norm": 3.1518092155456543,
|
| 214 |
+
"learning_rate": 1.7573221757322176e-06,
|
| 215 |
+
"loss": 0.0991,
|
| 216 |
+
"step": 675
|
| 217 |
+
},
|
| 218 |
+
{
|
| 219 |
+
"epoch": 2.928870292887029,
|
| 220 |
+
"grad_norm": 5.171344757080078,
|
| 221 |
+
"learning_rate": 7.112970711297071e-07,
|
| 222 |
+
"loss": 0.1123,
|
| 223 |
+
"step": 700
|
| 224 |
+
},
|
| 225 |
+
{
|
| 226 |
+
"epoch": 3.0,
|
| 227 |
+
"eval_accuracy": 0.9330385344283006,
|
| 228 |
+
"eval_loss": 0.2232467234134674,
|
| 229 |
+
"eval_runtime": 6.0708,
|
| 230 |
+
"eval_samples_per_second": 260.758,
|
| 231 |
+
"eval_steps_per_second": 2.8,
|
| 232 |
+
"step": 717
|
| 233 |
+
},
|
| 234 |
+
{
|
| 235 |
+
"epoch": 3.0,
|
| 236 |
+
"step": 717,
|
| 237 |
+
"total_flos": 1024714859387040.0,
|
| 238 |
+
"train_loss": 0.2006188094034022,
|
| 239 |
+
"train_runtime": 91.8654,
|
| 240 |
+
"train_samples_per_second": 373.133,
|
| 241 |
+
"train_steps_per_second": 7.805
|
| 242 |
+
}
|
| 243 |
+
],
|
| 244 |
+
"logging_steps": 25,
|
| 245 |
+
"max_steps": 717,
|
| 246 |
+
"num_input_tokens_seen": 0,
|
| 247 |
+
"num_train_epochs": 3,
|
| 248 |
+
"save_steps": 500,
|
| 249 |
+
"stateful_callbacks": {
|
| 250 |
+
"TrainerControl": {
|
| 251 |
+
"args": {
|
| 252 |
+
"should_epoch_stop": false,
|
| 253 |
+
"should_evaluate": false,
|
| 254 |
+
"should_log": false,
|
| 255 |
+
"should_save": false,
|
| 256 |
+
"should_training_stop": false
|
| 257 |
+
},
|
| 258 |
+
"attributes": {}
|
| 259 |
+
}
|
| 260 |
+
},
|
| 261 |
+
"total_flos": 1024714859387040.0,
|
| 262 |
+
"train_batch_size": 48,
|
| 263 |
+
"trial_name": null,
|
| 264 |
+
"trial_params": null
|
| 265 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_42/all_results.json
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.9412507896399241,
|
| 4 |
+
"eval_loss": 0.21872149407863617,
|
| 5 |
+
"eval_runtime": 6.2737,
|
| 6 |
+
"eval_samples": 1583,
|
| 7 |
+
"eval_samples_per_second": 252.325,
|
| 8 |
+
"eval_steps_per_second": 2.71,
|
| 9 |
+
"total_flos": 1024166730153984.0,
|
| 10 |
+
"train_loss": 0.2038555185166363,
|
| 11 |
+
"train_runtime": 114.9964,
|
| 12 |
+
"train_samples": 11426,
|
| 13 |
+
"train_samples_per_second": 298.079,
|
| 14 |
+
"train_steps_per_second": 6.235
|
| 15 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_42/eval_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.9412507896399241,
|
| 4 |
+
"eval_loss": 0.21872149407863617,
|
| 5 |
+
"eval_runtime": 6.2737,
|
| 6 |
+
"eval_samples": 1583,
|
| 7 |
+
"eval_samples_per_second": 252.325,
|
| 8 |
+
"eval_steps_per_second": 2.71
|
| 9 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/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 vsfc --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/vsfc/mBERT/cls/seed_42 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save --use_mha true --pooling_strategy cls --n_heads 4 --head_dim 32 --classifier_hidden_dim 512 --head_dropout 0.05
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_42/train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 1024166730153984.0,
|
| 4 |
+
"train_loss": 0.2038555185166363,
|
| 5 |
+
"train_runtime": 114.9964,
|
| 6 |
+
"train_samples": 11426,
|
| 7 |
+
"train_samples_per_second": 298.079,
|
| 8 |
+
"train_steps_per_second": 6.235
|
| 9 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_42/trainer_state.json
ADDED
|
@@ -0,0 +1,265 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 3.0,
|
| 5 |
+
"eval_steps": 500,
|
| 6 |
+
"global_step": 717,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.10460251046025104,
|
| 13 |
+
"grad_norm": 16.012611389160156,
|
| 14 |
+
"learning_rate": 2.8953974895397488e-05,
|
| 15 |
+
"loss": 0.5503,
|
| 16 |
+
"step": 25
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 0.20920502092050208,
|
| 20 |
+
"grad_norm": 4.707347869873047,
|
| 21 |
+
"learning_rate": 2.7907949790794982e-05,
|
| 22 |
+
"loss": 0.3373,
|
| 23 |
+
"step": 50
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.3138075313807531,
|
| 27 |
+
"grad_norm": 5.474881649017334,
|
| 28 |
+
"learning_rate": 2.686192468619247e-05,
|
| 29 |
+
"loss": 0.3206,
|
| 30 |
+
"step": 75
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.41841004184100417,
|
| 34 |
+
"grad_norm": 4.987305641174316,
|
| 35 |
+
"learning_rate": 2.5815899581589957e-05,
|
| 36 |
+
"loss": 0.2967,
|
| 37 |
+
"step": 100
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.5230125523012552,
|
| 41 |
+
"grad_norm": 3.684814214706421,
|
| 42 |
+
"learning_rate": 2.4769874476987448e-05,
|
| 43 |
+
"loss": 0.2746,
|
| 44 |
+
"step": 125
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.6276150627615062,
|
| 48 |
+
"grad_norm": 4.312676906585693,
|
| 49 |
+
"learning_rate": 2.372384937238494e-05,
|
| 50 |
+
"loss": 0.2723,
|
| 51 |
+
"step": 150
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.7322175732217573,
|
| 55 |
+
"grad_norm": 5.088875770568848,
|
| 56 |
+
"learning_rate": 2.2677824267782426e-05,
|
| 57 |
+
"loss": 0.2532,
|
| 58 |
+
"step": 175
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.8368200836820083,
|
| 62 |
+
"grad_norm": 4.8535637855529785,
|
| 63 |
+
"learning_rate": 2.1631799163179917e-05,
|
| 64 |
+
"loss": 0.2728,
|
| 65 |
+
"step": 200
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.9414225941422594,
|
| 69 |
+
"grad_norm": 6.244494438171387,
|
| 70 |
+
"learning_rate": 2.0585774058577408e-05,
|
| 71 |
+
"loss": 0.2395,
|
| 72 |
+
"step": 225
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 1.0,
|
| 76 |
+
"eval_accuracy": 0.9355653821857233,
|
| 77 |
+
"eval_loss": 0.22173087298870087,
|
| 78 |
+
"eval_runtime": 5.5217,
|
| 79 |
+
"eval_samples_per_second": 286.687,
|
| 80 |
+
"eval_steps_per_second": 3.079,
|
| 81 |
+
"step": 239
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"epoch": 1.0460251046025104,
|
| 85 |
+
"grad_norm": 6.3032426834106445,
|
| 86 |
+
"learning_rate": 1.9539748953974895e-05,
|
| 87 |
+
"loss": 0.202,
|
| 88 |
+
"step": 250
|
| 89 |
+
},
|
| 90 |
+
{
|
| 91 |
+
"epoch": 1.1506276150627615,
|
| 92 |
+
"grad_norm": 4.0951056480407715,
|
| 93 |
+
"learning_rate": 1.8493723849372386e-05,
|
| 94 |
+
"loss": 0.1822,
|
| 95 |
+
"step": 275
|
| 96 |
+
},
|
| 97 |
+
{
|
| 98 |
+
"epoch": 1.2552301255230125,
|
| 99 |
+
"grad_norm": 3.7123475074768066,
|
| 100 |
+
"learning_rate": 1.7447698744769877e-05,
|
| 101 |
+
"loss": 0.1502,
|
| 102 |
+
"step": 300
|
| 103 |
+
},
|
| 104 |
+
{
|
| 105 |
+
"epoch": 1.3598326359832635,
|
| 106 |
+
"grad_norm": 3.9388022422790527,
|
| 107 |
+
"learning_rate": 1.6401673640167365e-05,
|
| 108 |
+
"loss": 0.2208,
|
| 109 |
+
"step": 325
|
| 110 |
+
},
|
| 111 |
+
{
|
| 112 |
+
"epoch": 1.4644351464435146,
|
| 113 |
+
"grad_norm": 3.688840389251709,
|
| 114 |
+
"learning_rate": 1.5355648535564852e-05,
|
| 115 |
+
"loss": 0.1894,
|
| 116 |
+
"step": 350
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"epoch": 1.5690376569037658,
|
| 120 |
+
"grad_norm": 7.560548305511475,
|
| 121 |
+
"learning_rate": 1.4309623430962344e-05,
|
| 122 |
+
"loss": 0.2218,
|
| 123 |
+
"step": 375
|
| 124 |
+
},
|
| 125 |
+
{
|
| 126 |
+
"epoch": 1.6736401673640167,
|
| 127 |
+
"grad_norm": 3.603820323944092,
|
| 128 |
+
"learning_rate": 1.3263598326359832e-05,
|
| 129 |
+
"loss": 0.1937,
|
| 130 |
+
"step": 400
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 1.778242677824268,
|
| 134 |
+
"grad_norm": 5.51453971862793,
|
| 135 |
+
"learning_rate": 1.2217573221757323e-05,
|
| 136 |
+
"loss": 0.1737,
|
| 137 |
+
"step": 425
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.8828451882845187,
|
| 141 |
+
"grad_norm": 4.128765106201172,
|
| 142 |
+
"learning_rate": 1.1171548117154812e-05,
|
| 143 |
+
"loss": 0.1361,
|
| 144 |
+
"step": 450
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 1.98744769874477,
|
| 148 |
+
"grad_norm": 2.117112636566162,
|
| 149 |
+
"learning_rate": 1.0125523012552301e-05,
|
| 150 |
+
"loss": 0.1872,
|
| 151 |
+
"step": 475
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"epoch": 2.0,
|
| 155 |
+
"eval_accuracy": 0.9374605180037903,
|
| 156 |
+
"eval_loss": 0.2148786038160324,
|
| 157 |
+
"eval_runtime": 5.5551,
|
| 158 |
+
"eval_samples_per_second": 284.965,
|
| 159 |
+
"eval_steps_per_second": 3.06,
|
| 160 |
+
"step": 478
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 2.092050209205021,
|
| 164 |
+
"grad_norm": 2.183661937713623,
|
| 165 |
+
"learning_rate": 9.079497907949792e-06,
|
| 166 |
+
"loss": 0.1156,
|
| 167 |
+
"step": 500
|
| 168 |
+
},
|
| 169 |
+
{
|
| 170 |
+
"epoch": 2.196652719665272,
|
| 171 |
+
"grad_norm": 11.060365676879883,
|
| 172 |
+
"learning_rate": 8.03347280334728e-06,
|
| 173 |
+
"loss": 0.1203,
|
| 174 |
+
"step": 525
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 2.301255230125523,
|
| 178 |
+
"grad_norm": 5.259145259857178,
|
| 179 |
+
"learning_rate": 6.98744769874477e-06,
|
| 180 |
+
"loss": 0.1471,
|
| 181 |
+
"step": 550
|
| 182 |
+
},
|
| 183 |
+
{
|
| 184 |
+
"epoch": 2.405857740585774,
|
| 185 |
+
"grad_norm": 8.719051361083984,
|
| 186 |
+
"learning_rate": 5.941422594142259e-06,
|
| 187 |
+
"loss": 0.1235,
|
| 188 |
+
"step": 575
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"epoch": 2.510460251046025,
|
| 192 |
+
"grad_norm": 2.070744276046753,
|
| 193 |
+
"learning_rate": 4.895397489539749e-06,
|
| 194 |
+
"loss": 0.1058,
|
| 195 |
+
"step": 600
|
| 196 |
+
},
|
| 197 |
+
{
|
| 198 |
+
"epoch": 2.6150627615062763,
|
| 199 |
+
"grad_norm": 3.374753952026367,
|
| 200 |
+
"learning_rate": 3.849372384937238e-06,
|
| 201 |
+
"loss": 0.1431,
|
| 202 |
+
"step": 625
|
| 203 |
+
},
|
| 204 |
+
{
|
| 205 |
+
"epoch": 2.719665271966527,
|
| 206 |
+
"grad_norm": 7.895869731903076,
|
| 207 |
+
"learning_rate": 2.803347280334728e-06,
|
| 208 |
+
"loss": 0.1012,
|
| 209 |
+
"step": 650
|
| 210 |
+
},
|
| 211 |
+
{
|
| 212 |
+
"epoch": 2.8242677824267783,
|
| 213 |
+
"grad_norm": 3.3564658164978027,
|
| 214 |
+
"learning_rate": 1.7573221757322176e-06,
|
| 215 |
+
"loss": 0.1192,
|
| 216 |
+
"step": 675
|
| 217 |
+
},
|
| 218 |
+
{
|
| 219 |
+
"epoch": 2.928870292887029,
|
| 220 |
+
"grad_norm": 3.863130807876587,
|
| 221 |
+
"learning_rate": 7.112970711297071e-07,
|
| 222 |
+
"loss": 0.1118,
|
| 223 |
+
"step": 700
|
| 224 |
+
},
|
| 225 |
+
{
|
| 226 |
+
"epoch": 3.0,
|
| 227 |
+
"eval_accuracy": 0.9412507896399241,
|
| 228 |
+
"eval_loss": 0.21872149407863617,
|
| 229 |
+
"eval_runtime": 6.5671,
|
| 230 |
+
"eval_samples_per_second": 241.051,
|
| 231 |
+
"eval_steps_per_second": 2.589,
|
| 232 |
+
"step": 717
|
| 233 |
+
},
|
| 234 |
+
{
|
| 235 |
+
"epoch": 3.0,
|
| 236 |
+
"step": 717,
|
| 237 |
+
"total_flos": 1024166730153984.0,
|
| 238 |
+
"train_loss": 0.2038555185166363,
|
| 239 |
+
"train_runtime": 114.9964,
|
| 240 |
+
"train_samples_per_second": 298.079,
|
| 241 |
+
"train_steps_per_second": 6.235
|
| 242 |
+
}
|
| 243 |
+
],
|
| 244 |
+
"logging_steps": 25,
|
| 245 |
+
"max_steps": 717,
|
| 246 |
+
"num_input_tokens_seen": 0,
|
| 247 |
+
"num_train_epochs": 3,
|
| 248 |
+
"save_steps": 500,
|
| 249 |
+
"stateful_callbacks": {
|
| 250 |
+
"TrainerControl": {
|
| 251 |
+
"args": {
|
| 252 |
+
"should_epoch_stop": false,
|
| 253 |
+
"should_evaluate": false,
|
| 254 |
+
"should_log": false,
|
| 255 |
+
"should_save": false,
|
| 256 |
+
"should_training_stop": false
|
| 257 |
+
},
|
| 258 |
+
"attributes": {}
|
| 259 |
+
}
|
| 260 |
+
},
|
| 261 |
+
"total_flos": 1024166730153984.0,
|
| 262 |
+
"train_batch_size": 48,
|
| 263 |
+
"trial_name": null,
|
| 264 |
+
"trial_params": null
|
| 265 |
+
}
|