Upload batch 20/24: vsfc/mBERT/cls/seed_43 ... vsfc/mBERT/gated_multi_branch/seed_43
Browse files- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_43/all_results.json +15 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_43/eval_results.json +9 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_43/run_command.txt +1 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_43/train_results.json +9 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_43/trainer_state.json +265 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_44/all_results.json +15 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_44/eval_results.json +9 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_44/run_command.txt +1 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_44/train_results.json +9 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_44/trainer_state.json +265 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/gated_multi_branch/seed_42/all_results.json +15 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/gated_multi_branch/seed_42/eval_results.json +9 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/gated_multi_branch/seed_42/run_command.txt +1 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/gated_multi_branch/seed_42/train_results.json +9 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/gated_multi_branch/seed_42/trainer_state.json +265 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/gated_multi_branch/seed_43/run_command.txt +1 -0
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_43/all_results.json
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.9336702463676564,
|
| 4 |
+
"eval_loss": 0.24253493547439575,
|
| 5 |
+
"eval_runtime": 5.964,
|
| 6 |
+
"eval_samples": 1583,
|
| 7 |
+
"eval_samples_per_second": 265.425,
|
| 8 |
+
"eval_steps_per_second": 2.85,
|
| 9 |
+
"total_flos": 1022750124392448.0,
|
| 10 |
+
"train_loss": 0.2038297117837137,
|
| 11 |
+
"train_runtime": 107.6305,
|
| 12 |
+
"train_samples": 11426,
|
| 13 |
+
"train_samples_per_second": 318.479,
|
| 14 |
+
"train_steps_per_second": 6.662
|
| 15 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_43/eval_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.9336702463676564,
|
| 4 |
+
"eval_loss": 0.24253493547439575,
|
| 5 |
+
"eval_runtime": 5.964,
|
| 6 |
+
"eval_samples": 1583,
|
| 7 |
+
"eval_samples_per_second": 265.425,
|
| 8 |
+
"eval_steps_per_second": 2.85
|
| 9 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/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/cls/seed_43 --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_43/train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 1022750124392448.0,
|
| 4 |
+
"train_loss": 0.2038297117837137,
|
| 5 |
+
"train_runtime": 107.6305,
|
| 6 |
+
"train_samples": 11426,
|
| 7 |
+
"train_samples_per_second": 318.479,
|
| 8 |
+
"train_steps_per_second": 6.662
|
| 9 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_43/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": 7.378244876861572,
|
| 14 |
+
"learning_rate": 2.8953974895397488e-05,
|
| 15 |
+
"loss": 0.5073,
|
| 16 |
+
"step": 25
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 0.20920502092050208,
|
| 20 |
+
"grad_norm": 5.249192237854004,
|
| 21 |
+
"learning_rate": 2.7907949790794982e-05,
|
| 22 |
+
"loss": 0.3272,
|
| 23 |
+
"step": 50
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.3138075313807531,
|
| 27 |
+
"grad_norm": 6.105321407318115,
|
| 28 |
+
"learning_rate": 2.686192468619247e-05,
|
| 29 |
+
"loss": 0.31,
|
| 30 |
+
"step": 75
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.41841004184100417,
|
| 34 |
+
"grad_norm": 4.408397674560547,
|
| 35 |
+
"learning_rate": 2.5815899581589957e-05,
|
| 36 |
+
"loss": 0.2785,
|
| 37 |
+
"step": 100
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.5230125523012552,
|
| 41 |
+
"grad_norm": 5.953392505645752,
|
| 42 |
+
"learning_rate": 2.4769874476987448e-05,
|
| 43 |
+
"loss": 0.3148,
|
| 44 |
+
"step": 125
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.6276150627615062,
|
| 48 |
+
"grad_norm": 6.444172382354736,
|
| 49 |
+
"learning_rate": 2.372384937238494e-05,
|
| 50 |
+
"loss": 0.2696,
|
| 51 |
+
"step": 150
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.7322175732217573,
|
| 55 |
+
"grad_norm": 3.9428184032440186,
|
| 56 |
+
"learning_rate": 2.2677824267782426e-05,
|
| 57 |
+
"loss": 0.2604,
|
| 58 |
+
"step": 175
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.8368200836820083,
|
| 62 |
+
"grad_norm": 3.382925033569336,
|
| 63 |
+
"learning_rate": 2.1631799163179917e-05,
|
| 64 |
+
"loss": 0.2277,
|
| 65 |
+
"step": 200
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.9414225941422594,
|
| 69 |
+
"grad_norm": 8.231470108032227,
|
| 70 |
+
"learning_rate": 2.0585774058577408e-05,
|
| 71 |
+
"loss": 0.2447,
|
| 72 |
+
"step": 225
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 1.0,
|
| 76 |
+
"eval_accuracy": 0.9248262792166771,
|
| 77 |
+
"eval_loss": 0.23215016722679138,
|
| 78 |
+
"eval_runtime": 5.1143,
|
| 79 |
+
"eval_samples_per_second": 309.525,
|
| 80 |
+
"eval_steps_per_second": 3.324,
|
| 81 |
+
"step": 239
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"epoch": 1.0460251046025104,
|
| 85 |
+
"grad_norm": 5.831847667694092,
|
| 86 |
+
"learning_rate": 1.9539748953974895e-05,
|
| 87 |
+
"loss": 0.2338,
|
| 88 |
+
"step": 250
|
| 89 |
+
},
|
| 90 |
+
{
|
| 91 |
+
"epoch": 1.1506276150627615,
|
| 92 |
+
"grad_norm": 6.486989974975586,
|
| 93 |
+
"learning_rate": 1.8493723849372386e-05,
|
| 94 |
+
"loss": 0.179,
|
| 95 |
+
"step": 275
|
| 96 |
+
},
|
| 97 |
+
{
|
| 98 |
+
"epoch": 1.2552301255230125,
|
| 99 |
+
"grad_norm": 9.637701034545898,
|
| 100 |
+
"learning_rate": 1.7447698744769877e-05,
|
| 101 |
+
"loss": 0.2111,
|
| 102 |
+
"step": 300
|
| 103 |
+
},
|
| 104 |
+
{
|
| 105 |
+
"epoch": 1.3598326359832635,
|
| 106 |
+
"grad_norm": 6.876131057739258,
|
| 107 |
+
"learning_rate": 1.6401673640167365e-05,
|
| 108 |
+
"loss": 0.1982,
|
| 109 |
+
"step": 325
|
| 110 |
+
},
|
| 111 |
+
{
|
| 112 |
+
"epoch": 1.4644351464435146,
|
| 113 |
+
"grad_norm": 2.6049697399139404,
|
| 114 |
+
"learning_rate": 1.5355648535564852e-05,
|
| 115 |
+
"loss": 0.1653,
|
| 116 |
+
"step": 350
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"epoch": 1.5690376569037658,
|
| 120 |
+
"grad_norm": 4.280735969543457,
|
| 121 |
+
"learning_rate": 1.4309623430962344e-05,
|
| 122 |
+
"loss": 0.189,
|
| 123 |
+
"step": 375
|
| 124 |
+
},
|
| 125 |
+
{
|
| 126 |
+
"epoch": 1.6736401673640167,
|
| 127 |
+
"grad_norm": 5.472908020019531,
|
| 128 |
+
"learning_rate": 1.3263598326359832e-05,
|
| 129 |
+
"loss": 0.1718,
|
| 130 |
+
"step": 400
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 1.778242677824268,
|
| 134 |
+
"grad_norm": 2.487726926803589,
|
| 135 |
+
"learning_rate": 1.2217573221757323e-05,
|
| 136 |
+
"loss": 0.1809,
|
| 137 |
+
"step": 425
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.8828451882845187,
|
| 141 |
+
"grad_norm": 5.079280376434326,
|
| 142 |
+
"learning_rate": 1.1171548117154812e-05,
|
| 143 |
+
"loss": 0.1763,
|
| 144 |
+
"step": 450
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 1.98744769874477,
|
| 148 |
+
"grad_norm": 5.699792385101318,
|
| 149 |
+
"learning_rate": 1.0125523012552301e-05,
|
| 150 |
+
"loss": 0.1696,
|
| 151 |
+
"step": 475
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"epoch": 2.0,
|
| 155 |
+
"eval_accuracy": 0.9349336702463676,
|
| 156 |
+
"eval_loss": 0.21964505314826965,
|
| 157 |
+
"eval_runtime": 6.3297,
|
| 158 |
+
"eval_samples_per_second": 250.092,
|
| 159 |
+
"eval_steps_per_second": 2.686,
|
| 160 |
+
"step": 478
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 2.092050209205021,
|
| 164 |
+
"grad_norm": 7.6017584800720215,
|
| 165 |
+
"learning_rate": 9.079497907949792e-06,
|
| 166 |
+
"loss": 0.1077,
|
| 167 |
+
"step": 500
|
| 168 |
+
},
|
| 169 |
+
{
|
| 170 |
+
"epoch": 2.196652719665272,
|
| 171 |
+
"grad_norm": 10.151995658874512,
|
| 172 |
+
"learning_rate": 8.03347280334728e-06,
|
| 173 |
+
"loss": 0.1318,
|
| 174 |
+
"step": 525
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 2.301255230125523,
|
| 178 |
+
"grad_norm": 0.6403132081031799,
|
| 179 |
+
"learning_rate": 6.98744769874477e-06,
|
| 180 |
+
"loss": 0.1286,
|
| 181 |
+
"step": 550
|
| 182 |
+
},
|
| 183 |
+
{
|
| 184 |
+
"epoch": 2.405857740585774,
|
| 185 |
+
"grad_norm": 3.4294111728668213,
|
| 186 |
+
"learning_rate": 5.941422594142259e-06,
|
| 187 |
+
"loss": 0.1225,
|
| 188 |
+
"step": 575
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"epoch": 2.510460251046025,
|
| 192 |
+
"grad_norm": 2.919726848602295,
|
| 193 |
+
"learning_rate": 4.895397489539749e-06,
|
| 194 |
+
"loss": 0.1355,
|
| 195 |
+
"step": 600
|
| 196 |
+
},
|
| 197 |
+
{
|
| 198 |
+
"epoch": 2.6150627615062763,
|
| 199 |
+
"grad_norm": 1.0730109214782715,
|
| 200 |
+
"learning_rate": 3.849372384937238e-06,
|
| 201 |
+
"loss": 0.1242,
|
| 202 |
+
"step": 625
|
| 203 |
+
},
|
| 204 |
+
{
|
| 205 |
+
"epoch": 2.719665271966527,
|
| 206 |
+
"grad_norm": 0.22635599970817566,
|
| 207 |
+
"learning_rate": 2.803347280334728e-06,
|
| 208 |
+
"loss": 0.122,
|
| 209 |
+
"step": 650
|
| 210 |
+
},
|
| 211 |
+
{
|
| 212 |
+
"epoch": 2.8242677824267783,
|
| 213 |
+
"grad_norm": 5.841732978820801,
|
| 214 |
+
"learning_rate": 1.7573221757322176e-06,
|
| 215 |
+
"loss": 0.1393,
|
| 216 |
+
"step": 675
|
| 217 |
+
},
|
| 218 |
+
{
|
| 219 |
+
"epoch": 2.928870292887029,
|
| 220 |
+
"grad_norm": 3.0765984058380127,
|
| 221 |
+
"learning_rate": 7.112970711297071e-07,
|
| 222 |
+
"loss": 0.1451,
|
| 223 |
+
"step": 700
|
| 224 |
+
},
|
| 225 |
+
{
|
| 226 |
+
"epoch": 3.0,
|
| 227 |
+
"eval_accuracy": 0.9336702463676564,
|
| 228 |
+
"eval_loss": 0.24253493547439575,
|
| 229 |
+
"eval_runtime": 6.0835,
|
| 230 |
+
"eval_samples_per_second": 260.213,
|
| 231 |
+
"eval_steps_per_second": 2.794,
|
| 232 |
+
"step": 717
|
| 233 |
+
},
|
| 234 |
+
{
|
| 235 |
+
"epoch": 3.0,
|
| 236 |
+
"step": 717,
|
| 237 |
+
"total_flos": 1022750124392448.0,
|
| 238 |
+
"train_loss": 0.2038297117837137,
|
| 239 |
+
"train_runtime": 107.6305,
|
| 240 |
+
"train_samples_per_second": 318.479,
|
| 241 |
+
"train_steps_per_second": 6.662
|
| 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": 1022750124392448.0,
|
| 262 |
+
"train_batch_size": 48,
|
| 263 |
+
"trial_name": null,
|
| 264 |
+
"trial_params": null
|
| 265 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_44/all_results.json
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.9368288060644346,
|
| 4 |
+
"eval_loss": 0.22460007667541504,
|
| 5 |
+
"eval_runtime": 6.8591,
|
| 6 |
+
"eval_samples": 1583,
|
| 7 |
+
"eval_samples_per_second": 230.79,
|
| 8 |
+
"eval_steps_per_second": 2.478,
|
| 9 |
+
"total_flos": 1024705701150720.0,
|
| 10 |
+
"train_loss": 0.20489413286020233,
|
| 11 |
+
"train_runtime": 109.4788,
|
| 12 |
+
"train_samples": 11426,
|
| 13 |
+
"train_samples_per_second": 313.102,
|
| 14 |
+
"train_steps_per_second": 6.549
|
| 15 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/seed_44/eval_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.9368288060644346,
|
| 4 |
+
"eval_loss": 0.22460007667541504,
|
| 5 |
+
"eval_runtime": 6.8591,
|
| 6 |
+
"eval_samples": 1583,
|
| 7 |
+
"eval_samples_per_second": 230.79,
|
| 8 |
+
"eval_steps_per_second": 2.478
|
| 9 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/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/cls/seed_44 --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_44/train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 1024705701150720.0,
|
| 4 |
+
"train_loss": 0.20489413286020233,
|
| 5 |
+
"train_runtime": 109.4788,
|
| 6 |
+
"train_samples": 11426,
|
| 7 |
+
"train_samples_per_second": 313.102,
|
| 8 |
+
"train_steps_per_second": 6.549
|
| 9 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/cls/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": 5.839505195617676,
|
| 14 |
+
"learning_rate": 2.8953974895397488e-05,
|
| 15 |
+
"loss": 0.5621,
|
| 16 |
+
"step": 25
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 0.20920502092050208,
|
| 20 |
+
"grad_norm": 8.698034286499023,
|
| 21 |
+
"learning_rate": 2.7907949790794982e-05,
|
| 22 |
+
"loss": 0.3485,
|
| 23 |
+
"step": 50
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.3138075313807531,
|
| 27 |
+
"grad_norm": 4.915554523468018,
|
| 28 |
+
"learning_rate": 2.686192468619247e-05,
|
| 29 |
+
"loss": 0.349,
|
| 30 |
+
"step": 75
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.41841004184100417,
|
| 34 |
+
"grad_norm": 3.205019474029541,
|
| 35 |
+
"learning_rate": 2.5815899581589957e-05,
|
| 36 |
+
"loss": 0.2807,
|
| 37 |
+
"step": 100
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.5230125523012552,
|
| 41 |
+
"grad_norm": 6.627248764038086,
|
| 42 |
+
"learning_rate": 2.4769874476987448e-05,
|
| 43 |
+
"loss": 0.2857,
|
| 44 |
+
"step": 125
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.6276150627615062,
|
| 48 |
+
"grad_norm": 4.360877513885498,
|
| 49 |
+
"learning_rate": 2.372384937238494e-05,
|
| 50 |
+
"loss": 0.2527,
|
| 51 |
+
"step": 150
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.7322175732217573,
|
| 55 |
+
"grad_norm": 6.8538055419921875,
|
| 56 |
+
"learning_rate": 2.2677824267782426e-05,
|
| 57 |
+
"loss": 0.2945,
|
| 58 |
+
"step": 175
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.8368200836820083,
|
| 62 |
+
"grad_norm": 4.769765853881836,
|
| 63 |
+
"learning_rate": 2.1631799163179917e-05,
|
| 64 |
+
"loss": 0.2192,
|
| 65 |
+
"step": 200
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.9414225941422594,
|
| 69 |
+
"grad_norm": 6.304840564727783,
|
| 70 |
+
"learning_rate": 2.0585774058577408e-05,
|
| 71 |
+
"loss": 0.233,
|
| 72 |
+
"step": 225
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 1.0,
|
| 76 |
+
"eval_accuracy": 0.9241945672773215,
|
| 77 |
+
"eval_loss": 0.23117734491825104,
|
| 78 |
+
"eval_runtime": 6.8742,
|
| 79 |
+
"eval_samples_per_second": 230.28,
|
| 80 |
+
"eval_steps_per_second": 2.473,
|
| 81 |
+
"step": 239
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"epoch": 1.0460251046025104,
|
| 85 |
+
"grad_norm": 5.387945652008057,
|
| 86 |
+
"learning_rate": 1.9539748953974895e-05,
|
| 87 |
+
"loss": 0.2108,
|
| 88 |
+
"step": 250
|
| 89 |
+
},
|
| 90 |
+
{
|
| 91 |
+
"epoch": 1.1506276150627615,
|
| 92 |
+
"grad_norm": 8.130422592163086,
|
| 93 |
+
"learning_rate": 1.8493723849372386e-05,
|
| 94 |
+
"loss": 0.1985,
|
| 95 |
+
"step": 275
|
| 96 |
+
},
|
| 97 |
+
{
|
| 98 |
+
"epoch": 1.2552301255230125,
|
| 99 |
+
"grad_norm": 3.605717420578003,
|
| 100 |
+
"learning_rate": 1.7447698744769877e-05,
|
| 101 |
+
"loss": 0.1865,
|
| 102 |
+
"step": 300
|
| 103 |
+
},
|
| 104 |
+
{
|
| 105 |
+
"epoch": 1.3598326359832635,
|
| 106 |
+
"grad_norm": 6.228057861328125,
|
| 107 |
+
"learning_rate": 1.6401673640167365e-05,
|
| 108 |
+
"loss": 0.1918,
|
| 109 |
+
"step": 325
|
| 110 |
+
},
|
| 111 |
+
{
|
| 112 |
+
"epoch": 1.4644351464435146,
|
| 113 |
+
"grad_norm": 5.351726531982422,
|
| 114 |
+
"learning_rate": 1.5355648535564852e-05,
|
| 115 |
+
"loss": 0.1889,
|
| 116 |
+
"step": 350
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"epoch": 1.5690376569037658,
|
| 120 |
+
"grad_norm": 6.882822513580322,
|
| 121 |
+
"learning_rate": 1.4309623430962344e-05,
|
| 122 |
+
"loss": 0.1818,
|
| 123 |
+
"step": 375
|
| 124 |
+
},
|
| 125 |
+
{
|
| 126 |
+
"epoch": 1.6736401673640167,
|
| 127 |
+
"grad_norm": 6.294622898101807,
|
| 128 |
+
"learning_rate": 1.3263598326359832e-05,
|
| 129 |
+
"loss": 0.1878,
|
| 130 |
+
"step": 400
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 1.778242677824268,
|
| 134 |
+
"grad_norm": 4.073336601257324,
|
| 135 |
+
"learning_rate": 1.2217573221757323e-05,
|
| 136 |
+
"loss": 0.1885,
|
| 137 |
+
"step": 425
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.8828451882845187,
|
| 141 |
+
"grad_norm": 7.144247531890869,
|
| 142 |
+
"learning_rate": 1.1171548117154812e-05,
|
| 143 |
+
"loss": 0.2025,
|
| 144 |
+
"step": 450
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 1.98744769874477,
|
| 148 |
+
"grad_norm": 4.857992172241211,
|
| 149 |
+
"learning_rate": 1.0125523012552301e-05,
|
| 150 |
+
"loss": 0.15,
|
| 151 |
+
"step": 475
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"epoch": 2.0,
|
| 155 |
+
"eval_accuracy": 0.9349336702463676,
|
| 156 |
+
"eval_loss": 0.23249971866607666,
|
| 157 |
+
"eval_runtime": 5.897,
|
| 158 |
+
"eval_samples_per_second": 268.44,
|
| 159 |
+
"eval_steps_per_second": 2.883,
|
| 160 |
+
"step": 478
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 2.092050209205021,
|
| 164 |
+
"grad_norm": 5.46315860748291,
|
| 165 |
+
"learning_rate": 9.079497907949792e-06,
|
| 166 |
+
"loss": 0.1299,
|
| 167 |
+
"step": 500
|
| 168 |
+
},
|
| 169 |
+
{
|
| 170 |
+
"epoch": 2.196652719665272,
|
| 171 |
+
"grad_norm": 7.099837779998779,
|
| 172 |
+
"learning_rate": 8.03347280334728e-06,
|
| 173 |
+
"loss": 0.1183,
|
| 174 |
+
"step": 525
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 2.301255230125523,
|
| 178 |
+
"grad_norm": 1.7706037759780884,
|
| 179 |
+
"learning_rate": 6.98744769874477e-06,
|
| 180 |
+
"loss": 0.1248,
|
| 181 |
+
"step": 550
|
| 182 |
+
},
|
| 183 |
+
{
|
| 184 |
+
"epoch": 2.405857740585774,
|
| 185 |
+
"grad_norm": 5.448747634887695,
|
| 186 |
+
"learning_rate": 5.941422594142259e-06,
|
| 187 |
+
"loss": 0.1256,
|
| 188 |
+
"step": 575
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"epoch": 2.510460251046025,
|
| 192 |
+
"grad_norm": 4.138149261474609,
|
| 193 |
+
"learning_rate": 4.895397489539749e-06,
|
| 194 |
+
"loss": 0.1172,
|
| 195 |
+
"step": 600
|
| 196 |
+
},
|
| 197 |
+
{
|
| 198 |
+
"epoch": 2.6150627615062763,
|
| 199 |
+
"grad_norm": 8.39890193939209,
|
| 200 |
+
"learning_rate": 3.849372384937238e-06,
|
| 201 |
+
"loss": 0.1156,
|
| 202 |
+
"step": 625
|
| 203 |
+
},
|
| 204 |
+
{
|
| 205 |
+
"epoch": 2.719665271966527,
|
| 206 |
+
"grad_norm": 6.7020583152771,
|
| 207 |
+
"learning_rate": 2.803347280334728e-06,
|
| 208 |
+
"loss": 0.1359,
|
| 209 |
+
"step": 650
|
| 210 |
+
},
|
| 211 |
+
{
|
| 212 |
+
"epoch": 2.8242677824267783,
|
| 213 |
+
"grad_norm": 4.436653137207031,
|
| 214 |
+
"learning_rate": 1.7573221757322176e-06,
|
| 215 |
+
"loss": 0.1023,
|
| 216 |
+
"step": 675
|
| 217 |
+
},
|
| 218 |
+
{
|
| 219 |
+
"epoch": 2.928870292887029,
|
| 220 |
+
"grad_norm": 4.737896919250488,
|
| 221 |
+
"learning_rate": 7.112970711297071e-07,
|
| 222 |
+
"loss": 0.1212,
|
| 223 |
+
"step": 700
|
| 224 |
+
},
|
| 225 |
+
{
|
| 226 |
+
"epoch": 3.0,
|
| 227 |
+
"eval_accuracy": 0.9368288060644346,
|
| 228 |
+
"eval_loss": 0.22460007667541504,
|
| 229 |
+
"eval_runtime": 6.7744,
|
| 230 |
+
"eval_samples_per_second": 233.675,
|
| 231 |
+
"eval_steps_per_second": 2.509,
|
| 232 |
+
"step": 717
|
| 233 |
+
},
|
| 234 |
+
{
|
| 235 |
+
"epoch": 3.0,
|
| 236 |
+
"step": 717,
|
| 237 |
+
"total_flos": 1024705701150720.0,
|
| 238 |
+
"train_loss": 0.20489413286020233,
|
| 239 |
+
"train_runtime": 109.4788,
|
| 240 |
+
"train_samples_per_second": 313.102,
|
| 241 |
+
"train_steps_per_second": 6.549
|
| 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": 1024705701150720.0,
|
| 262 |
+
"train_batch_size": 48,
|
| 263 |
+
"trial_name": null,
|
| 264 |
+
"trial_params": null
|
| 265 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/gated_multi_branch/seed_42/all_results.json
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.9380922299431459,
|
| 4 |
+
"eval_loss": 0.2133793830871582,
|
| 5 |
+
"eval_runtime": 4.9086,
|
| 6 |
+
"eval_samples": 1583,
|
| 7 |
+
"eval_samples_per_second": 322.498,
|
| 8 |
+
"eval_steps_per_second": 3.463,
|
| 9 |
+
"total_flos": 1056984845046000.0,
|
| 10 |
+
"train_loss": 0.2020011867605659,
|
| 11 |
+
"train_runtime": 76.5559,
|
| 12 |
+
"train_samples": 11426,
|
| 13 |
+
"train_samples_per_second": 447.751,
|
| 14 |
+
"train_steps_per_second": 9.366
|
| 15 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/gated_multi_branch/seed_42/eval_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.9380922299431459,
|
| 4 |
+
"eval_loss": 0.2133793830871582,
|
| 5 |
+
"eval_runtime": 4.9086,
|
| 6 |
+
"eval_samples": 1583,
|
| 7 |
+
"eval_samples_per_second": 322.498,
|
| 8 |
+
"eval_steps_per_second": 3.463
|
| 9 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/gated_multi_branch/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/gated_multi_branch/seed_42 --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/vsfc/mBERT/gated_multi_branch/seed_42/train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 1056984845046000.0,
|
| 4 |
+
"train_loss": 0.2020011867605659,
|
| 5 |
+
"train_runtime": 76.5559,
|
| 6 |
+
"train_samples": 11426,
|
| 7 |
+
"train_samples_per_second": 447.751,
|
| 8 |
+
"train_steps_per_second": 9.366
|
| 9 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/gated_multi_branch/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": 12.183186531066895,
|
| 14 |
+
"learning_rate": 2.8953974895397488e-05,
|
| 15 |
+
"loss": 0.5028,
|
| 16 |
+
"step": 25
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 0.20920502092050208,
|
| 20 |
+
"grad_norm": 22.80581283569336,
|
| 21 |
+
"learning_rate": 2.7907949790794982e-05,
|
| 22 |
+
"loss": 0.3196,
|
| 23 |
+
"step": 50
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.3138075313807531,
|
| 27 |
+
"grad_norm": 4.0907182693481445,
|
| 28 |
+
"learning_rate": 2.686192468619247e-05,
|
| 29 |
+
"loss": 0.3319,
|
| 30 |
+
"step": 75
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.41841004184100417,
|
| 34 |
+
"grad_norm": 8.712798118591309,
|
| 35 |
+
"learning_rate": 2.5815899581589957e-05,
|
| 36 |
+
"loss": 0.2939,
|
| 37 |
+
"step": 100
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.5230125523012552,
|
| 41 |
+
"grad_norm": 3.6525375843048096,
|
| 42 |
+
"learning_rate": 2.4769874476987448e-05,
|
| 43 |
+
"loss": 0.2687,
|
| 44 |
+
"step": 125
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.6276150627615062,
|
| 48 |
+
"grad_norm": 6.947529315948486,
|
| 49 |
+
"learning_rate": 2.372384937238494e-05,
|
| 50 |
+
"loss": 0.2631,
|
| 51 |
+
"step": 150
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.7322175732217573,
|
| 55 |
+
"grad_norm": 6.012436866760254,
|
| 56 |
+
"learning_rate": 2.2677824267782426e-05,
|
| 57 |
+
"loss": 0.2567,
|
| 58 |
+
"step": 175
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.8368200836820083,
|
| 62 |
+
"grad_norm": 8.469512939453125,
|
| 63 |
+
"learning_rate": 2.1631799163179917e-05,
|
| 64 |
+
"loss": 0.2837,
|
| 65 |
+
"step": 200
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.9414225941422594,
|
| 69 |
+
"grad_norm": 4.85753870010376,
|
| 70 |
+
"learning_rate": 2.0585774058577408e-05,
|
| 71 |
+
"loss": 0.2557,
|
| 72 |
+
"step": 225
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 1.0,
|
| 76 |
+
"eval_accuracy": 0.9298799747315224,
|
| 77 |
+
"eval_loss": 0.22936604917049408,
|
| 78 |
+
"eval_runtime": 6.2139,
|
| 79 |
+
"eval_samples_per_second": 254.75,
|
| 80 |
+
"eval_steps_per_second": 2.736,
|
| 81 |
+
"step": 239
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"epoch": 1.0460251046025104,
|
| 85 |
+
"grad_norm": 4.789795398712158,
|
| 86 |
+
"learning_rate": 1.9539748953974895e-05,
|
| 87 |
+
"loss": 0.1914,
|
| 88 |
+
"step": 250
|
| 89 |
+
},
|
| 90 |
+
{
|
| 91 |
+
"epoch": 1.1506276150627615,
|
| 92 |
+
"grad_norm": 4.908311367034912,
|
| 93 |
+
"learning_rate": 1.8493723849372386e-05,
|
| 94 |
+
"loss": 0.1839,
|
| 95 |
+
"step": 275
|
| 96 |
+
},
|
| 97 |
+
{
|
| 98 |
+
"epoch": 1.2552301255230125,
|
| 99 |
+
"grad_norm": 5.252670764923096,
|
| 100 |
+
"learning_rate": 1.7447698744769877e-05,
|
| 101 |
+
"loss": 0.1478,
|
| 102 |
+
"step": 300
|
| 103 |
+
},
|
| 104 |
+
{
|
| 105 |
+
"epoch": 1.3598326359832635,
|
| 106 |
+
"grad_norm": 4.32583475112915,
|
| 107 |
+
"learning_rate": 1.6401673640167365e-05,
|
| 108 |
+
"loss": 0.2132,
|
| 109 |
+
"step": 325
|
| 110 |
+
},
|
| 111 |
+
{
|
| 112 |
+
"epoch": 1.4644351464435146,
|
| 113 |
+
"grad_norm": 2.8848624229431152,
|
| 114 |
+
"learning_rate": 1.5355648535564852e-05,
|
| 115 |
+
"loss": 0.201,
|
| 116 |
+
"step": 350
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"epoch": 1.5690376569037658,
|
| 120 |
+
"grad_norm": 5.385087966918945,
|
| 121 |
+
"learning_rate": 1.4309623430962344e-05,
|
| 122 |
+
"loss": 0.2075,
|
| 123 |
+
"step": 375
|
| 124 |
+
},
|
| 125 |
+
{
|
| 126 |
+
"epoch": 1.6736401673640167,
|
| 127 |
+
"grad_norm": 4.180697441101074,
|
| 128 |
+
"learning_rate": 1.3263598326359832e-05,
|
| 129 |
+
"loss": 0.1798,
|
| 130 |
+
"step": 400
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 1.778242677824268,
|
| 134 |
+
"grad_norm": 5.759138584136963,
|
| 135 |
+
"learning_rate": 1.2217573221757323e-05,
|
| 136 |
+
"loss": 0.163,
|
| 137 |
+
"step": 425
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.8828451882845187,
|
| 141 |
+
"grad_norm": 3.642009735107422,
|
| 142 |
+
"learning_rate": 1.1171548117154812e-05,
|
| 143 |
+
"loss": 0.1473,
|
| 144 |
+
"step": 450
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 1.98744769874477,
|
| 148 |
+
"grad_norm": 4.941315174102783,
|
| 149 |
+
"learning_rate": 1.0125523012552301e-05,
|
| 150 |
+
"loss": 0.188,
|
| 151 |
+
"step": 475
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"epoch": 2.0,
|
| 155 |
+
"eval_accuracy": 0.9387239418825016,
|
| 156 |
+
"eval_loss": 0.19972530007362366,
|
| 157 |
+
"eval_runtime": 5.0651,
|
| 158 |
+
"eval_samples_per_second": 312.529,
|
| 159 |
+
"eval_steps_per_second": 3.356,
|
| 160 |
+
"step": 478
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 2.092050209205021,
|
| 164 |
+
"grad_norm": 4.164829254150391,
|
| 165 |
+
"learning_rate": 9.079497907949792e-06,
|
| 166 |
+
"loss": 0.1226,
|
| 167 |
+
"step": 500
|
| 168 |
+
},
|
| 169 |
+
{
|
| 170 |
+
"epoch": 2.196652719665272,
|
| 171 |
+
"grad_norm": 8.023189544677734,
|
| 172 |
+
"learning_rate": 8.03347280334728e-06,
|
| 173 |
+
"loss": 0.1322,
|
| 174 |
+
"step": 525
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 2.301255230125523,
|
| 178 |
+
"grad_norm": 3.9289586544036865,
|
| 179 |
+
"learning_rate": 6.98744769874477e-06,
|
| 180 |
+
"loss": 0.1437,
|
| 181 |
+
"step": 550
|
| 182 |
+
},
|
| 183 |
+
{
|
| 184 |
+
"epoch": 2.405857740585774,
|
| 185 |
+
"grad_norm": 3.6356136798858643,
|
| 186 |
+
"learning_rate": 5.941422594142259e-06,
|
| 187 |
+
"loss": 0.1346,
|
| 188 |
+
"step": 575
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"epoch": 2.510460251046025,
|
| 192 |
+
"grad_norm": 4.76814079284668,
|
| 193 |
+
"learning_rate": 4.895397489539749e-06,
|
| 194 |
+
"loss": 0.0996,
|
| 195 |
+
"step": 600
|
| 196 |
+
},
|
| 197 |
+
{
|
| 198 |
+
"epoch": 2.6150627615062763,
|
| 199 |
+
"grad_norm": 4.777148723602295,
|
| 200 |
+
"learning_rate": 3.849372384937238e-06,
|
| 201 |
+
"loss": 0.1352,
|
| 202 |
+
"step": 625
|
| 203 |
+
},
|
| 204 |
+
{
|
| 205 |
+
"epoch": 2.719665271966527,
|
| 206 |
+
"grad_norm": 4.211899757385254,
|
| 207 |
+
"learning_rate": 2.803347280334728e-06,
|
| 208 |
+
"loss": 0.1176,
|
| 209 |
+
"step": 650
|
| 210 |
+
},
|
| 211 |
+
{
|
| 212 |
+
"epoch": 2.8242677824267783,
|
| 213 |
+
"grad_norm": 3.3080577850341797,
|
| 214 |
+
"learning_rate": 1.7573221757322176e-06,
|
| 215 |
+
"loss": 0.1185,
|
| 216 |
+
"step": 675
|
| 217 |
+
},
|
| 218 |
+
{
|
| 219 |
+
"epoch": 2.928870292887029,
|
| 220 |
+
"grad_norm": 4.976438045501709,
|
| 221 |
+
"learning_rate": 7.112970711297071e-07,
|
| 222 |
+
"loss": 0.1125,
|
| 223 |
+
"step": 700
|
| 224 |
+
},
|
| 225 |
+
{
|
| 226 |
+
"epoch": 3.0,
|
| 227 |
+
"eval_accuracy": 0.9380922299431459,
|
| 228 |
+
"eval_loss": 0.2133793830871582,
|
| 229 |
+
"eval_runtime": 4.8529,
|
| 230 |
+
"eval_samples_per_second": 326.197,
|
| 231 |
+
"eval_steps_per_second": 3.503,
|
| 232 |
+
"step": 717
|
| 233 |
+
},
|
| 234 |
+
{
|
| 235 |
+
"epoch": 3.0,
|
| 236 |
+
"step": 717,
|
| 237 |
+
"total_flos": 1056984845046000.0,
|
| 238 |
+
"train_loss": 0.2020011867605659,
|
| 239 |
+
"train_runtime": 76.5559,
|
| 240 |
+
"train_samples_per_second": 447.751,
|
| 241 |
+
"train_steps_per_second": 9.366
|
| 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": 1056984845046000.0,
|
| 262 |
+
"train_batch_size": 48,
|
| 263 |
+
"trial_name": null,
|
| 264 |
+
"trial_params": null
|
| 265 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/gated_multi_branch/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/gated_multi_branch/seed_43 --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
|