Upload batch 21/24: vsfc/mBERT/gated_multi_branch/seed_44 ... vsfc/mBERT/hf_sequence_classifier/seed_44
Browse files- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/gated_multi_branch/seed_44/all_results.json +15 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/gated_multi_branch/seed_44/eval_results.json +9 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/gated_multi_branch/seed_44/run_command.txt +1 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/gated_multi_branch/seed_44/train_results.json +9 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/gated_multi_branch/seed_44/trainer_state.json +265 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_42/all_results.json +15 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_42/eval_results.json +9 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_42/run_command.txt +1 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_42/train_results.json +9 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_42/trainer_state.json +265 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_43/all_results.json +15 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_43/eval_results.json +9 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_43/run_command.txt +1 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_43/train_results.json +9 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_43/trainer_state.json +265 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_44/all_results.json +15 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_44/eval_results.json +9 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_44/run_command.txt +1 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_44/train_results.json +9 -0
- mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_44/trainer_state.json +265 -0
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/gated_multi_branch/seed_44/all_results.json
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.9387239418825016,
|
| 4 |
+
"eval_loss": 0.22616708278656006,
|
| 5 |
+
"eval_runtime": 4.4746,
|
| 6 |
+
"eval_samples": 1583,
|
| 7 |
+
"eval_samples_per_second": 353.774,
|
| 8 |
+
"eval_steps_per_second": 3.799,
|
| 9 |
+
"total_flos": 1057541086680000.0,
|
| 10 |
+
"train_loss": 0.20627780971500498,
|
| 11 |
+
"train_runtime": 63.2564,
|
| 12 |
+
"train_samples": 11426,
|
| 13 |
+
"train_samples_per_second": 541.89,
|
| 14 |
+
"train_steps_per_second": 11.335
|
| 15 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/gated_multi_branch/seed_44/eval_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.9387239418825016,
|
| 4 |
+
"eval_loss": 0.22616708278656006,
|
| 5 |
+
"eval_runtime": 4.4746,
|
| 6 |
+
"eval_samples": 1583,
|
| 7 |
+
"eval_samples_per_second": 353.774,
|
| 8 |
+
"eval_steps_per_second": 3.799
|
| 9 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/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 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/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/vsfc/mBERT/gated_multi_branch/seed_44/train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 1057541086680000.0,
|
| 4 |
+
"train_loss": 0.20627780971500498,
|
| 5 |
+
"train_runtime": 63.2564,
|
| 6 |
+
"train_samples": 11426,
|
| 7 |
+
"train_samples_per_second": 541.89,
|
| 8 |
+
"train_steps_per_second": 11.335
|
| 9 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/gated_multi_branch/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": 6.1553754806518555,
|
| 14 |
+
"learning_rate": 2.8953974895397488e-05,
|
| 15 |
+
"loss": 0.5197,
|
| 16 |
+
"step": 25
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 0.20920502092050208,
|
| 20 |
+
"grad_norm": 2.645526647567749,
|
| 21 |
+
"learning_rate": 2.7907949790794982e-05,
|
| 22 |
+
"loss": 0.365,
|
| 23 |
+
"step": 50
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.3138075313807531,
|
| 27 |
+
"grad_norm": 4.37541389465332,
|
| 28 |
+
"learning_rate": 2.686192468619247e-05,
|
| 29 |
+
"loss": 0.3619,
|
| 30 |
+
"step": 75
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.41841004184100417,
|
| 34 |
+
"grad_norm": 8.045869827270508,
|
| 35 |
+
"learning_rate": 2.5815899581589957e-05,
|
| 36 |
+
"loss": 0.2778,
|
| 37 |
+
"step": 100
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.5230125523012552,
|
| 41 |
+
"grad_norm": 5.381649971008301,
|
| 42 |
+
"learning_rate": 2.4769874476987448e-05,
|
| 43 |
+
"loss": 0.277,
|
| 44 |
+
"step": 125
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.6276150627615062,
|
| 48 |
+
"grad_norm": 6.045936107635498,
|
| 49 |
+
"learning_rate": 2.372384937238494e-05,
|
| 50 |
+
"loss": 0.2657,
|
| 51 |
+
"step": 150
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.7322175732217573,
|
| 55 |
+
"grad_norm": 5.311103343963623,
|
| 56 |
+
"learning_rate": 2.2677824267782426e-05,
|
| 57 |
+
"loss": 0.3023,
|
| 58 |
+
"step": 175
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.8368200836820083,
|
| 62 |
+
"grad_norm": 3.3695340156555176,
|
| 63 |
+
"learning_rate": 2.1631799163179917e-05,
|
| 64 |
+
"loss": 0.2205,
|
| 65 |
+
"step": 200
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.9414225941422594,
|
| 69 |
+
"grad_norm": 5.4858198165893555,
|
| 70 |
+
"learning_rate": 2.0585774058577408e-05,
|
| 71 |
+
"loss": 0.2353,
|
| 72 |
+
"step": 225
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 1.0,
|
| 76 |
+
"eval_accuracy": 0.9235628553379659,
|
| 77 |
+
"eval_loss": 0.22577011585235596,
|
| 78 |
+
"eval_runtime": 5.2608,
|
| 79 |
+
"eval_samples_per_second": 300.905,
|
| 80 |
+
"eval_steps_per_second": 3.231,
|
| 81 |
+
"step": 239
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"epoch": 1.0460251046025104,
|
| 85 |
+
"grad_norm": 6.068897724151611,
|
| 86 |
+
"learning_rate": 1.9539748953974895e-05,
|
| 87 |
+
"loss": 0.2112,
|
| 88 |
+
"step": 250
|
| 89 |
+
},
|
| 90 |
+
{
|
| 91 |
+
"epoch": 1.1506276150627615,
|
| 92 |
+
"grad_norm": 6.946470737457275,
|
| 93 |
+
"learning_rate": 1.8493723849372386e-05,
|
| 94 |
+
"loss": 0.2006,
|
| 95 |
+
"step": 275
|
| 96 |
+
},
|
| 97 |
+
{
|
| 98 |
+
"epoch": 1.2552301255230125,
|
| 99 |
+
"grad_norm": 4.09370756149292,
|
| 100 |
+
"learning_rate": 1.7447698744769877e-05,
|
| 101 |
+
"loss": 0.1745,
|
| 102 |
+
"step": 300
|
| 103 |
+
},
|
| 104 |
+
{
|
| 105 |
+
"epoch": 1.3598326359832635,
|
| 106 |
+
"grad_norm": 1.5913387537002563,
|
| 107 |
+
"learning_rate": 1.6401673640167365e-05,
|
| 108 |
+
"loss": 0.2029,
|
| 109 |
+
"step": 325
|
| 110 |
+
},
|
| 111 |
+
{
|
| 112 |
+
"epoch": 1.4644351464435146,
|
| 113 |
+
"grad_norm": 3.317037343978882,
|
| 114 |
+
"learning_rate": 1.5355648535564852e-05,
|
| 115 |
+
"loss": 0.1949,
|
| 116 |
+
"step": 350
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"epoch": 1.5690376569037658,
|
| 120 |
+
"grad_norm": 6.61057710647583,
|
| 121 |
+
"learning_rate": 1.4309623430962344e-05,
|
| 122 |
+
"loss": 0.186,
|
| 123 |
+
"step": 375
|
| 124 |
+
},
|
| 125 |
+
{
|
| 126 |
+
"epoch": 1.6736401673640167,
|
| 127 |
+
"grad_norm": 4.376493453979492,
|
| 128 |
+
"learning_rate": 1.3263598326359832e-05,
|
| 129 |
+
"loss": 0.2018,
|
| 130 |
+
"step": 400
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 1.778242677824268,
|
| 134 |
+
"grad_norm": 4.861920356750488,
|
| 135 |
+
"learning_rate": 1.2217573221757323e-05,
|
| 136 |
+
"loss": 0.1894,
|
| 137 |
+
"step": 425
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.8828451882845187,
|
| 141 |
+
"grad_norm": 7.5122904777526855,
|
| 142 |
+
"learning_rate": 1.1171548117154812e-05,
|
| 143 |
+
"loss": 0.1846,
|
| 144 |
+
"step": 450
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 1.98744769874477,
|
| 148 |
+
"grad_norm": 7.324886798858643,
|
| 149 |
+
"learning_rate": 1.0125523012552301e-05,
|
| 150 |
+
"loss": 0.1637,
|
| 151 |
+
"step": 475
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"epoch": 2.0,
|
| 155 |
+
"eval_accuracy": 0.934301958307012,
|
| 156 |
+
"eval_loss": 0.22446905076503754,
|
| 157 |
+
"eval_runtime": 4.8614,
|
| 158 |
+
"eval_samples_per_second": 325.625,
|
| 159 |
+
"eval_steps_per_second": 3.497,
|
| 160 |
+
"step": 478
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 2.092050209205021,
|
| 164 |
+
"grad_norm": 6.352123737335205,
|
| 165 |
+
"learning_rate": 9.079497907949792e-06,
|
| 166 |
+
"loss": 0.1336,
|
| 167 |
+
"step": 500
|
| 168 |
+
},
|
| 169 |
+
{
|
| 170 |
+
"epoch": 2.196652719665272,
|
| 171 |
+
"grad_norm": 3.2229907512664795,
|
| 172 |
+
"learning_rate": 8.03347280334728e-06,
|
| 173 |
+
"loss": 0.1199,
|
| 174 |
+
"step": 525
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 2.301255230125523,
|
| 178 |
+
"grad_norm": 1.5446957349777222,
|
| 179 |
+
"learning_rate": 6.98744769874477e-06,
|
| 180 |
+
"loss": 0.1315,
|
| 181 |
+
"step": 550
|
| 182 |
+
},
|
| 183 |
+
{
|
| 184 |
+
"epoch": 2.405857740585774,
|
| 185 |
+
"grad_norm": 6.907514572143555,
|
| 186 |
+
"learning_rate": 5.941422594142259e-06,
|
| 187 |
+
"loss": 0.1125,
|
| 188 |
+
"step": 575
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"epoch": 2.510460251046025,
|
| 192 |
+
"grad_norm": 9.05813217163086,
|
| 193 |
+
"learning_rate": 4.895397489539749e-06,
|
| 194 |
+
"loss": 0.1184,
|
| 195 |
+
"step": 600
|
| 196 |
+
},
|
| 197 |
+
{
|
| 198 |
+
"epoch": 2.6150627615062763,
|
| 199 |
+
"grad_norm": 5.002927303314209,
|
| 200 |
+
"learning_rate": 3.849372384937238e-06,
|
| 201 |
+
"loss": 0.1086,
|
| 202 |
+
"step": 625
|
| 203 |
+
},
|
| 204 |
+
{
|
| 205 |
+
"epoch": 2.719665271966527,
|
| 206 |
+
"grad_norm": 5.513826847076416,
|
| 207 |
+
"learning_rate": 2.803347280334728e-06,
|
| 208 |
+
"loss": 0.146,
|
| 209 |
+
"step": 650
|
| 210 |
+
},
|
| 211 |
+
{
|
| 212 |
+
"epoch": 2.8242677824267783,
|
| 213 |
+
"grad_norm": 5.641959190368652,
|
| 214 |
+
"learning_rate": 1.7573221757322176e-06,
|
| 215 |
+
"loss": 0.1078,
|
| 216 |
+
"step": 675
|
| 217 |
+
},
|
| 218 |
+
{
|
| 219 |
+
"epoch": 2.928870292887029,
|
| 220 |
+
"grad_norm": 5.182094097137451,
|
| 221 |
+
"learning_rate": 7.112970711297071e-07,
|
| 222 |
+
"loss": 0.12,
|
| 223 |
+
"step": 700
|
| 224 |
+
},
|
| 225 |
+
{
|
| 226 |
+
"epoch": 3.0,
|
| 227 |
+
"eval_accuracy": 0.9387239418825016,
|
| 228 |
+
"eval_loss": 0.22616708278656006,
|
| 229 |
+
"eval_runtime": 4.2732,
|
| 230 |
+
"eval_samples_per_second": 370.45,
|
| 231 |
+
"eval_steps_per_second": 3.978,
|
| 232 |
+
"step": 717
|
| 233 |
+
},
|
| 234 |
+
{
|
| 235 |
+
"epoch": 3.0,
|
| 236 |
+
"step": 717,
|
| 237 |
+
"total_flos": 1057541086680000.0,
|
| 238 |
+
"train_loss": 0.20627780971500498,
|
| 239 |
+
"train_runtime": 63.2564,
|
| 240 |
+
"train_samples_per_second": 541.89,
|
| 241 |
+
"train_steps_per_second": 11.335
|
| 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": 1057541086680000.0,
|
| 262 |
+
"train_batch_size": 48,
|
| 263 |
+
"trial_name": null,
|
| 264 |
+
"trial_params": null
|
| 265 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_42/all_results.json
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.9393556538218573,
|
| 4 |
+
"eval_loss": 0.22358696162700653,
|
| 5 |
+
"eval_runtime": 6.5609,
|
| 6 |
+
"eval_samples": 1583,
|
| 7 |
+
"eval_samples_per_second": 241.279,
|
| 8 |
+
"eval_steps_per_second": 2.591,
|
| 9 |
+
"total_flos": 1019480215179384.0,
|
| 10 |
+
"train_loss": 0.20489713470639875,
|
| 11 |
+
"train_runtime": 125.6498,
|
| 12 |
+
"train_samples": 11426,
|
| 13 |
+
"train_samples_per_second": 272.806,
|
| 14 |
+
"train_steps_per_second": 5.706
|
| 15 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_42/eval_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.9393556538218573,
|
| 4 |
+
"eval_loss": 0.22358696162700653,
|
| 5 |
+
"eval_runtime": 6.5609,
|
| 6 |
+
"eval_samples": 1583,
|
| 7 |
+
"eval_samples_per_second": 241.279,
|
| 8 |
+
"eval_steps_per_second": 2.591
|
| 9 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/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 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/hf_sequence_classifier/seed_42 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_42/train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 1019480215179384.0,
|
| 4 |
+
"train_loss": 0.20489713470639875,
|
| 5 |
+
"train_runtime": 125.6498,
|
| 6 |
+
"train_samples": 11426,
|
| 7 |
+
"train_samples_per_second": 272.806,
|
| 8 |
+
"train_steps_per_second": 5.706
|
| 9 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/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": 7.640495777130127,
|
| 14 |
+
"learning_rate": 2.8953974895397488e-05,
|
| 15 |
+
"loss": 0.5749,
|
| 16 |
+
"step": 25
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 0.20920502092050208,
|
| 20 |
+
"grad_norm": 4.028187274932861,
|
| 21 |
+
"learning_rate": 2.7907949790794982e-05,
|
| 22 |
+
"loss": 0.3413,
|
| 23 |
+
"step": 50
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.3138075313807531,
|
| 27 |
+
"grad_norm": 7.7388458251953125,
|
| 28 |
+
"learning_rate": 2.686192468619247e-05,
|
| 29 |
+
"loss": 0.3232,
|
| 30 |
+
"step": 75
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.41841004184100417,
|
| 34 |
+
"grad_norm": 6.740450382232666,
|
| 35 |
+
"learning_rate": 2.5815899581589957e-05,
|
| 36 |
+
"loss": 0.3017,
|
| 37 |
+
"step": 100
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.5230125523012552,
|
| 41 |
+
"grad_norm": 2.691441535949707,
|
| 42 |
+
"learning_rate": 2.4769874476987448e-05,
|
| 43 |
+
"loss": 0.2741,
|
| 44 |
+
"step": 125
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.6276150627615062,
|
| 48 |
+
"grad_norm": 5.469559192657471,
|
| 49 |
+
"learning_rate": 2.372384937238494e-05,
|
| 50 |
+
"loss": 0.2664,
|
| 51 |
+
"step": 150
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.7322175732217573,
|
| 55 |
+
"grad_norm": 4.212676525115967,
|
| 56 |
+
"learning_rate": 2.2677824267782426e-05,
|
| 57 |
+
"loss": 0.2595,
|
| 58 |
+
"step": 175
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.8368200836820083,
|
| 62 |
+
"grad_norm": 7.397810459136963,
|
| 63 |
+
"learning_rate": 2.1631799163179917e-05,
|
| 64 |
+
"loss": 0.2817,
|
| 65 |
+
"step": 200
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.9414225941422594,
|
| 69 |
+
"grad_norm": 7.419157028198242,
|
| 70 |
+
"learning_rate": 2.0585774058577408e-05,
|
| 71 |
+
"loss": 0.2537,
|
| 72 |
+
"step": 225
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 1.0,
|
| 76 |
+
"eval_accuracy": 0.9330385344283006,
|
| 77 |
+
"eval_loss": 0.2244487702846527,
|
| 78 |
+
"eval_runtime": 6.4795,
|
| 79 |
+
"eval_samples_per_second": 244.308,
|
| 80 |
+
"eval_steps_per_second": 2.624,
|
| 81 |
+
"step": 239
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"epoch": 1.0460251046025104,
|
| 85 |
+
"grad_norm": 5.6753315925598145,
|
| 86 |
+
"learning_rate": 1.9539748953974895e-05,
|
| 87 |
+
"loss": 0.1949,
|
| 88 |
+
"step": 250
|
| 89 |
+
},
|
| 90 |
+
{
|
| 91 |
+
"epoch": 1.1506276150627615,
|
| 92 |
+
"grad_norm": 4.945658206939697,
|
| 93 |
+
"learning_rate": 1.8493723849372386e-05,
|
| 94 |
+
"loss": 0.177,
|
| 95 |
+
"step": 275
|
| 96 |
+
},
|
| 97 |
+
{
|
| 98 |
+
"epoch": 1.2552301255230125,
|
| 99 |
+
"grad_norm": 2.2462313175201416,
|
| 100 |
+
"learning_rate": 1.7447698744769877e-05,
|
| 101 |
+
"loss": 0.1519,
|
| 102 |
+
"step": 300
|
| 103 |
+
},
|
| 104 |
+
{
|
| 105 |
+
"epoch": 1.3598326359832635,
|
| 106 |
+
"grad_norm": 3.609130620956421,
|
| 107 |
+
"learning_rate": 1.6401673640167365e-05,
|
| 108 |
+
"loss": 0.2185,
|
| 109 |
+
"step": 325
|
| 110 |
+
},
|
| 111 |
+
{
|
| 112 |
+
"epoch": 1.4644351464435146,
|
| 113 |
+
"grad_norm": 3.552107095718384,
|
| 114 |
+
"learning_rate": 1.5355648535564852e-05,
|
| 115 |
+
"loss": 0.1841,
|
| 116 |
+
"step": 350
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"epoch": 1.5690376569037658,
|
| 120 |
+
"grad_norm": 5.812036514282227,
|
| 121 |
+
"learning_rate": 1.4309623430962344e-05,
|
| 122 |
+
"loss": 0.2129,
|
| 123 |
+
"step": 375
|
| 124 |
+
},
|
| 125 |
+
{
|
| 126 |
+
"epoch": 1.6736401673640167,
|
| 127 |
+
"grad_norm": 2.933537244796753,
|
| 128 |
+
"learning_rate": 1.3263598326359832e-05,
|
| 129 |
+
"loss": 0.1812,
|
| 130 |
+
"step": 400
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 1.778242677824268,
|
| 134 |
+
"grad_norm": 4.838982582092285,
|
| 135 |
+
"learning_rate": 1.2217573221757323e-05,
|
| 136 |
+
"loss": 0.1646,
|
| 137 |
+
"step": 425
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.8828451882845187,
|
| 141 |
+
"grad_norm": 6.99026346206665,
|
| 142 |
+
"learning_rate": 1.1171548117154812e-05,
|
| 143 |
+
"loss": 0.1463,
|
| 144 |
+
"step": 450
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 1.98744769874477,
|
| 148 |
+
"grad_norm": 3.795510768890381,
|
| 149 |
+
"learning_rate": 1.0125523012552301e-05,
|
| 150 |
+
"loss": 0.1846,
|
| 151 |
+
"step": 475
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"epoch": 2.0,
|
| 155 |
+
"eval_accuracy": 0.9374605180037903,
|
| 156 |
+
"eval_loss": 0.20885473489761353,
|
| 157 |
+
"eval_runtime": 6.4005,
|
| 158 |
+
"eval_samples_per_second": 247.323,
|
| 159 |
+
"eval_steps_per_second": 2.656,
|
| 160 |
+
"step": 478
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 2.092050209205021,
|
| 164 |
+
"grad_norm": 2.2493112087249756,
|
| 165 |
+
"learning_rate": 9.079497907949792e-06,
|
| 166 |
+
"loss": 0.1173,
|
| 167 |
+
"step": 500
|
| 168 |
+
},
|
| 169 |
+
{
|
| 170 |
+
"epoch": 2.196652719665272,
|
| 171 |
+
"grad_norm": 8.88083267211914,
|
| 172 |
+
"learning_rate": 8.03347280334728e-06,
|
| 173 |
+
"loss": 0.1266,
|
| 174 |
+
"step": 525
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 2.301255230125523,
|
| 178 |
+
"grad_norm": 7.1378655433654785,
|
| 179 |
+
"learning_rate": 6.98744769874477e-06,
|
| 180 |
+
"loss": 0.1507,
|
| 181 |
+
"step": 550
|
| 182 |
+
},
|
| 183 |
+
{
|
| 184 |
+
"epoch": 2.405857740585774,
|
| 185 |
+
"grad_norm": 4.881401538848877,
|
| 186 |
+
"learning_rate": 5.941422594142259e-06,
|
| 187 |
+
"loss": 0.145,
|
| 188 |
+
"step": 575
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"epoch": 2.510460251046025,
|
| 192 |
+
"grad_norm": 3.265550136566162,
|
| 193 |
+
"learning_rate": 4.895397489539749e-06,
|
| 194 |
+
"loss": 0.0905,
|
| 195 |
+
"step": 600
|
| 196 |
+
},
|
| 197 |
+
{
|
| 198 |
+
"epoch": 2.6150627615062763,
|
| 199 |
+
"grad_norm": 3.416532039642334,
|
| 200 |
+
"learning_rate": 3.849372384937238e-06,
|
| 201 |
+
"loss": 0.1339,
|
| 202 |
+
"step": 625
|
| 203 |
+
},
|
| 204 |
+
{
|
| 205 |
+
"epoch": 2.719665271966527,
|
| 206 |
+
"grad_norm": 10.348471641540527,
|
| 207 |
+
"learning_rate": 2.803347280334728e-06,
|
| 208 |
+
"loss": 0.1135,
|
| 209 |
+
"step": 650
|
| 210 |
+
},
|
| 211 |
+
{
|
| 212 |
+
"epoch": 2.8242677824267783,
|
| 213 |
+
"grad_norm": 3.085171699523926,
|
| 214 |
+
"learning_rate": 1.7573221757322176e-06,
|
| 215 |
+
"loss": 0.1151,
|
| 216 |
+
"step": 675
|
| 217 |
+
},
|
| 218 |
+
{
|
| 219 |
+
"epoch": 2.928870292887029,
|
| 220 |
+
"grad_norm": 1.947435736656189,
|
| 221 |
+
"learning_rate": 7.112970711297071e-07,
|
| 222 |
+
"loss": 0.1082,
|
| 223 |
+
"step": 700
|
| 224 |
+
},
|
| 225 |
+
{
|
| 226 |
+
"epoch": 3.0,
|
| 227 |
+
"eval_accuracy": 0.9393556538218573,
|
| 228 |
+
"eval_loss": 0.22358696162700653,
|
| 229 |
+
"eval_runtime": 6.5577,
|
| 230 |
+
"eval_samples_per_second": 241.396,
|
| 231 |
+
"eval_steps_per_second": 2.592,
|
| 232 |
+
"step": 717
|
| 233 |
+
},
|
| 234 |
+
{
|
| 235 |
+
"epoch": 3.0,
|
| 236 |
+
"step": 717,
|
| 237 |
+
"total_flos": 1019480215179384.0,
|
| 238 |
+
"train_loss": 0.20489713470639875,
|
| 239 |
+
"train_runtime": 125.6498,
|
| 240 |
+
"train_samples_per_second": 272.806,
|
| 241 |
+
"train_steps_per_second": 5.706
|
| 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": 1019480215179384.0,
|
| 262 |
+
"train_batch_size": 48,
|
| 263 |
+
"trial_name": null,
|
| 264 |
+
"trial_params": null
|
| 265 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_43/all_results.json
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.9374605180037903,
|
| 4 |
+
"eval_loss": 0.22951717674732208,
|
| 5 |
+
"eval_runtime": 5.1616,
|
| 6 |
+
"eval_samples": 1583,
|
| 7 |
+
"eval_samples_per_second": 306.686,
|
| 8 |
+
"eval_steps_per_second": 3.294,
|
| 9 |
+
"total_flos": 1018070091706248.0,
|
| 10 |
+
"train_loss": 0.21280724117779168,
|
| 11 |
+
"train_runtime": 117.1565,
|
| 12 |
+
"train_samples": 11426,
|
| 13 |
+
"train_samples_per_second": 292.583,
|
| 14 |
+
"train_steps_per_second": 6.12
|
| 15 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_43/eval_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.9374605180037903,
|
| 4 |
+
"eval_loss": 0.22951717674732208,
|
| 5 |
+
"eval_runtime": 5.1616,
|
| 6 |
+
"eval_samples": 1583,
|
| 7 |
+
"eval_samples_per_second": 306.686,
|
| 8 |
+
"eval_steps_per_second": 3.294
|
| 9 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/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 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/hf_sequence_classifier/seed_43 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_43/train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 1018070091706248.0,
|
| 4 |
+
"train_loss": 0.21280724117779168,
|
| 5 |
+
"train_runtime": 117.1565,
|
| 6 |
+
"train_samples": 11426,
|
| 7 |
+
"train_samples_per_second": 292.583,
|
| 8 |
+
"train_steps_per_second": 6.12
|
| 9 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/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": 4.272794246673584,
|
| 14 |
+
"learning_rate": 2.8953974895397488e-05,
|
| 15 |
+
"loss": 0.548,
|
| 16 |
+
"step": 25
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 0.20920502092050208,
|
| 20 |
+
"grad_norm": 4.229963779449463,
|
| 21 |
+
"learning_rate": 2.7907949790794982e-05,
|
| 22 |
+
"loss": 0.342,
|
| 23 |
+
"step": 50
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.3138075313807531,
|
| 27 |
+
"grad_norm": 10.101731300354004,
|
| 28 |
+
"learning_rate": 2.686192468619247e-05,
|
| 29 |
+
"loss": 0.3111,
|
| 30 |
+
"step": 75
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.41841004184100417,
|
| 34 |
+
"grad_norm": 5.7693095207214355,
|
| 35 |
+
"learning_rate": 2.5815899581589957e-05,
|
| 36 |
+
"loss": 0.3185,
|
| 37 |
+
"step": 100
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.5230125523012552,
|
| 41 |
+
"grad_norm": 5.563674449920654,
|
| 42 |
+
"learning_rate": 2.4769874476987448e-05,
|
| 43 |
+
"loss": 0.3515,
|
| 44 |
+
"step": 125
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.6276150627615062,
|
| 48 |
+
"grad_norm": 4.599033832550049,
|
| 49 |
+
"learning_rate": 2.372384937238494e-05,
|
| 50 |
+
"loss": 0.2799,
|
| 51 |
+
"step": 150
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.7322175732217573,
|
| 55 |
+
"grad_norm": 4.112405776977539,
|
| 56 |
+
"learning_rate": 2.2677824267782426e-05,
|
| 57 |
+
"loss": 0.2672,
|
| 58 |
+
"step": 175
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.8368200836820083,
|
| 62 |
+
"grad_norm": 9.992273330688477,
|
| 63 |
+
"learning_rate": 2.1631799163179917e-05,
|
| 64 |
+
"loss": 0.2307,
|
| 65 |
+
"step": 200
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.9414225941422594,
|
| 69 |
+
"grad_norm": 8.105793952941895,
|
| 70 |
+
"learning_rate": 2.0585774058577408e-05,
|
| 71 |
+
"loss": 0.2542,
|
| 72 |
+
"step": 225
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 1.0,
|
| 76 |
+
"eval_accuracy": 0.9241945672773215,
|
| 77 |
+
"eval_loss": 0.23398654162883759,
|
| 78 |
+
"eval_runtime": 5.177,
|
| 79 |
+
"eval_samples_per_second": 305.775,
|
| 80 |
+
"eval_steps_per_second": 3.284,
|
| 81 |
+
"step": 239
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"epoch": 1.0460251046025104,
|
| 85 |
+
"grad_norm": 3.9256441593170166,
|
| 86 |
+
"learning_rate": 1.9539748953974895e-05,
|
| 87 |
+
"loss": 0.2378,
|
| 88 |
+
"step": 250
|
| 89 |
+
},
|
| 90 |
+
{
|
| 91 |
+
"epoch": 1.1506276150627615,
|
| 92 |
+
"grad_norm": 5.58455753326416,
|
| 93 |
+
"learning_rate": 1.8493723849372386e-05,
|
| 94 |
+
"loss": 0.1876,
|
| 95 |
+
"step": 275
|
| 96 |
+
},
|
| 97 |
+
{
|
| 98 |
+
"epoch": 1.2552301255230125,
|
| 99 |
+
"grad_norm": 9.820943832397461,
|
| 100 |
+
"learning_rate": 1.7447698744769877e-05,
|
| 101 |
+
"loss": 0.2175,
|
| 102 |
+
"step": 300
|
| 103 |
+
},
|
| 104 |
+
{
|
| 105 |
+
"epoch": 1.3598326359832635,
|
| 106 |
+
"grad_norm": 8.81803035736084,
|
| 107 |
+
"learning_rate": 1.6401673640167365e-05,
|
| 108 |
+
"loss": 0.1995,
|
| 109 |
+
"step": 325
|
| 110 |
+
},
|
| 111 |
+
{
|
| 112 |
+
"epoch": 1.4644351464435146,
|
| 113 |
+
"grad_norm": 5.75858211517334,
|
| 114 |
+
"learning_rate": 1.5355648535564852e-05,
|
| 115 |
+
"loss": 0.1625,
|
| 116 |
+
"step": 350
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"epoch": 1.5690376569037658,
|
| 120 |
+
"grad_norm": 4.1745686531066895,
|
| 121 |
+
"learning_rate": 1.4309623430962344e-05,
|
| 122 |
+
"loss": 0.2012,
|
| 123 |
+
"step": 375
|
| 124 |
+
},
|
| 125 |
+
{
|
| 126 |
+
"epoch": 1.6736401673640167,
|
| 127 |
+
"grad_norm": 5.042236804962158,
|
| 128 |
+
"learning_rate": 1.3263598326359832e-05,
|
| 129 |
+
"loss": 0.1873,
|
| 130 |
+
"step": 400
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 1.778242677824268,
|
| 134 |
+
"grad_norm": 6.48967981338501,
|
| 135 |
+
"learning_rate": 1.2217573221757323e-05,
|
| 136 |
+
"loss": 0.1917,
|
| 137 |
+
"step": 425
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.8828451882845187,
|
| 141 |
+
"grad_norm": 3.13720965385437,
|
| 142 |
+
"learning_rate": 1.1171548117154812e-05,
|
| 143 |
+
"loss": 0.1739,
|
| 144 |
+
"step": 450
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 1.98744769874477,
|
| 148 |
+
"grad_norm": 4.325597286224365,
|
| 149 |
+
"learning_rate": 1.0125523012552301e-05,
|
| 150 |
+
"loss": 0.1832,
|
| 151 |
+
"step": 475
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"epoch": 2.0,
|
| 155 |
+
"eval_accuracy": 0.9311433986102338,
|
| 156 |
+
"eval_loss": 0.21307209134101868,
|
| 157 |
+
"eval_runtime": 6.9381,
|
| 158 |
+
"eval_samples_per_second": 228.161,
|
| 159 |
+
"eval_steps_per_second": 2.45,
|
| 160 |
+
"step": 478
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 2.092050209205021,
|
| 164 |
+
"grad_norm": 5.890207767486572,
|
| 165 |
+
"learning_rate": 9.079497907949792e-06,
|
| 166 |
+
"loss": 0.1118,
|
| 167 |
+
"step": 500
|
| 168 |
+
},
|
| 169 |
+
{
|
| 170 |
+
"epoch": 2.196652719665272,
|
| 171 |
+
"grad_norm": 4.331007957458496,
|
| 172 |
+
"learning_rate": 8.03347280334728e-06,
|
| 173 |
+
"loss": 0.1281,
|
| 174 |
+
"step": 525
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 2.301255230125523,
|
| 178 |
+
"grad_norm": 1.158936858177185,
|
| 179 |
+
"learning_rate": 6.98744769874477e-06,
|
| 180 |
+
"loss": 0.1279,
|
| 181 |
+
"step": 550
|
| 182 |
+
},
|
| 183 |
+
{
|
| 184 |
+
"epoch": 2.405857740585774,
|
| 185 |
+
"grad_norm": 3.5535171031951904,
|
| 186 |
+
"learning_rate": 5.941422594142259e-06,
|
| 187 |
+
"loss": 0.1404,
|
| 188 |
+
"step": 575
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"epoch": 2.510460251046025,
|
| 192 |
+
"grad_norm": 3.112006902694702,
|
| 193 |
+
"learning_rate": 4.895397489539749e-06,
|
| 194 |
+
"loss": 0.1367,
|
| 195 |
+
"step": 600
|
| 196 |
+
},
|
| 197 |
+
{
|
| 198 |
+
"epoch": 2.6150627615062763,
|
| 199 |
+
"grad_norm": 5.817686557769775,
|
| 200 |
+
"learning_rate": 3.849372384937238e-06,
|
| 201 |
+
"loss": 0.1293,
|
| 202 |
+
"step": 625
|
| 203 |
+
},
|
| 204 |
+
{
|
| 205 |
+
"epoch": 2.719665271966527,
|
| 206 |
+
"grad_norm": 0.4306275546550751,
|
| 207 |
+
"learning_rate": 2.803347280334728e-06,
|
| 208 |
+
"loss": 0.1315,
|
| 209 |
+
"step": 650
|
| 210 |
+
},
|
| 211 |
+
{
|
| 212 |
+
"epoch": 2.8242677824267783,
|
| 213 |
+
"grad_norm": 3.946089506149292,
|
| 214 |
+
"learning_rate": 1.7573221757322176e-06,
|
| 215 |
+
"loss": 0.136,
|
| 216 |
+
"step": 675
|
| 217 |
+
},
|
| 218 |
+
{
|
| 219 |
+
"epoch": 2.928870292887029,
|
| 220 |
+
"grad_norm": 4.047253608703613,
|
| 221 |
+
"learning_rate": 7.112970711297071e-07,
|
| 222 |
+
"loss": 0.1436,
|
| 223 |
+
"step": 700
|
| 224 |
+
},
|
| 225 |
+
{
|
| 226 |
+
"epoch": 3.0,
|
| 227 |
+
"eval_accuracy": 0.9374605180037903,
|
| 228 |
+
"eval_loss": 0.22951717674732208,
|
| 229 |
+
"eval_runtime": 6.894,
|
| 230 |
+
"eval_samples_per_second": 229.62,
|
| 231 |
+
"eval_steps_per_second": 2.466,
|
| 232 |
+
"step": 717
|
| 233 |
+
},
|
| 234 |
+
{
|
| 235 |
+
"epoch": 3.0,
|
| 236 |
+
"step": 717,
|
| 237 |
+
"total_flos": 1018070091706248.0,
|
| 238 |
+
"train_loss": 0.21280724117779168,
|
| 239 |
+
"train_runtime": 117.1565,
|
| 240 |
+
"train_samples_per_second": 292.583,
|
| 241 |
+
"train_steps_per_second": 6.12
|
| 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": 1018070091706248.0,
|
| 262 |
+
"train_batch_size": 48,
|
| 263 |
+
"trial_name": null,
|
| 264 |
+
"trial_params": null
|
| 265 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_44/all_results.json
ADDED
|
@@ -0,0 +1,15 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.932406822488945,
|
| 4 |
+
"eval_loss": 0.23924019932746887,
|
| 5 |
+
"eval_runtime": 6.3759,
|
| 6 |
+
"eval_samples": 1583,
|
| 7 |
+
"eval_samples_per_second": 248.279,
|
| 8 |
+
"eval_steps_per_second": 2.666,
|
| 9 |
+
"total_flos": 1020016719882720.0,
|
| 10 |
+
"train_loss": 0.21428075062846205,
|
| 11 |
+
"train_runtime": 119.0204,
|
| 12 |
+
"train_samples": 11426,
|
| 13 |
+
"train_samples_per_second": 288.001,
|
| 14 |
+
"train_steps_per_second": 6.024
|
| 15 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_44/eval_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"eval_accuracy": 0.932406822488945,
|
| 4 |
+
"eval_loss": 0.23924019932746887,
|
| 5 |
+
"eval_runtime": 6.3759,
|
| 6 |
+
"eval_samples": 1583,
|
| 7 |
+
"eval_samples_per_second": 248.279,
|
| 8 |
+
"eval_steps_per_second": 2.666
|
| 9 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/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 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/hf_sequence_classifier/seed_44 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/seed_44/train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3.0,
|
| 3 |
+
"total_flos": 1020016719882720.0,
|
| 4 |
+
"train_loss": 0.21428075062846205,
|
| 5 |
+
"train_runtime": 119.0204,
|
| 6 |
+
"train_samples": 11426,
|
| 7 |
+
"train_samples_per_second": 288.001,
|
| 8 |
+
"train_steps_per_second": 6.024
|
| 9 |
+
}
|
mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/hf_sequence_classifier/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": 5.493389129638672,
|
| 14 |
+
"learning_rate": 2.8953974895397488e-05,
|
| 15 |
+
"loss": 0.583,
|
| 16 |
+
"step": 25
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 0.20920502092050208,
|
| 20 |
+
"grad_norm": 5.90028190612793,
|
| 21 |
+
"learning_rate": 2.7907949790794982e-05,
|
| 22 |
+
"loss": 0.3618,
|
| 23 |
+
"step": 50
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.3138075313807531,
|
| 27 |
+
"grad_norm": 4.627682685852051,
|
| 28 |
+
"learning_rate": 2.686192468619247e-05,
|
| 29 |
+
"loss": 0.3506,
|
| 30 |
+
"step": 75
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.41841004184100417,
|
| 34 |
+
"grad_norm": 3.2980947494506836,
|
| 35 |
+
"learning_rate": 2.5815899581589957e-05,
|
| 36 |
+
"loss": 0.3232,
|
| 37 |
+
"step": 100
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.5230125523012552,
|
| 41 |
+
"grad_norm": 5.769346237182617,
|
| 42 |
+
"learning_rate": 2.4769874476987448e-05,
|
| 43 |
+
"loss": 0.2833,
|
| 44 |
+
"step": 125
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.6276150627615062,
|
| 48 |
+
"grad_norm": 5.368684768676758,
|
| 49 |
+
"learning_rate": 2.372384937238494e-05,
|
| 50 |
+
"loss": 0.2583,
|
| 51 |
+
"step": 150
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.7322175732217573,
|
| 55 |
+
"grad_norm": 6.5861053466796875,
|
| 56 |
+
"learning_rate": 2.2677824267782426e-05,
|
| 57 |
+
"loss": 0.296,
|
| 58 |
+
"step": 175
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.8368200836820083,
|
| 62 |
+
"grad_norm": 3.3918707370758057,
|
| 63 |
+
"learning_rate": 2.1631799163179917e-05,
|
| 64 |
+
"loss": 0.2219,
|
| 65 |
+
"step": 200
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.9414225941422594,
|
| 69 |
+
"grad_norm": 7.96839714050293,
|
| 70 |
+
"learning_rate": 2.0585774058577408e-05,
|
| 71 |
+
"loss": 0.2336,
|
| 72 |
+
"step": 225
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 1.0,
|
| 76 |
+
"eval_accuracy": 0.9210360075805433,
|
| 77 |
+
"eval_loss": 0.23681390285491943,
|
| 78 |
+
"eval_runtime": 6.3097,
|
| 79 |
+
"eval_samples_per_second": 250.882,
|
| 80 |
+
"eval_steps_per_second": 2.694,
|
| 81 |
+
"step": 239
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"epoch": 1.0460251046025104,
|
| 85 |
+
"grad_norm": 2.349342107772827,
|
| 86 |
+
"learning_rate": 1.9539748953974895e-05,
|
| 87 |
+
"loss": 0.2232,
|
| 88 |
+
"step": 250
|
| 89 |
+
},
|
| 90 |
+
{
|
| 91 |
+
"epoch": 1.1506276150627615,
|
| 92 |
+
"grad_norm": 6.752340316772461,
|
| 93 |
+
"learning_rate": 1.8493723849372386e-05,
|
| 94 |
+
"loss": 0.2006,
|
| 95 |
+
"step": 275
|
| 96 |
+
},
|
| 97 |
+
{
|
| 98 |
+
"epoch": 1.2552301255230125,
|
| 99 |
+
"grad_norm": 2.9140703678131104,
|
| 100 |
+
"learning_rate": 1.7447698744769877e-05,
|
| 101 |
+
"loss": 0.2007,
|
| 102 |
+
"step": 300
|
| 103 |
+
},
|
| 104 |
+
{
|
| 105 |
+
"epoch": 1.3598326359832635,
|
| 106 |
+
"grad_norm": 2.5294418334960938,
|
| 107 |
+
"learning_rate": 1.6401673640167365e-05,
|
| 108 |
+
"loss": 0.1928,
|
| 109 |
+
"step": 325
|
| 110 |
+
},
|
| 111 |
+
{
|
| 112 |
+
"epoch": 1.4644351464435146,
|
| 113 |
+
"grad_norm": 2.4641706943511963,
|
| 114 |
+
"learning_rate": 1.5355648535564852e-05,
|
| 115 |
+
"loss": 0.2016,
|
| 116 |
+
"step": 350
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"epoch": 1.5690376569037658,
|
| 120 |
+
"grad_norm": 8.905129432678223,
|
| 121 |
+
"learning_rate": 1.4309623430962344e-05,
|
| 122 |
+
"loss": 0.1905,
|
| 123 |
+
"step": 375
|
| 124 |
+
},
|
| 125 |
+
{
|
| 126 |
+
"epoch": 1.6736401673640167,
|
| 127 |
+
"grad_norm": 5.136508941650391,
|
| 128 |
+
"learning_rate": 1.3263598326359832e-05,
|
| 129 |
+
"loss": 0.2027,
|
| 130 |
+
"step": 400
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 1.778242677824268,
|
| 134 |
+
"grad_norm": 4.05654764175415,
|
| 135 |
+
"learning_rate": 1.2217573221757323e-05,
|
| 136 |
+
"loss": 0.1966,
|
| 137 |
+
"step": 425
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.8828451882845187,
|
| 141 |
+
"grad_norm": 4.408506393432617,
|
| 142 |
+
"learning_rate": 1.1171548117154812e-05,
|
| 143 |
+
"loss": 0.2049,
|
| 144 |
+
"step": 450
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 1.98744769874477,
|
| 148 |
+
"grad_norm": 4.980363368988037,
|
| 149 |
+
"learning_rate": 1.0125523012552301e-05,
|
| 150 |
+
"loss": 0.1658,
|
| 151 |
+
"step": 475
|
| 152 |
+
},
|
| 153 |
+
{
|
| 154 |
+
"epoch": 2.0,
|
| 155 |
+
"eval_accuracy": 0.930511686670878,
|
| 156 |
+
"eval_loss": 0.22501271963119507,
|
| 157 |
+
"eval_runtime": 5.8626,
|
| 158 |
+
"eval_samples_per_second": 270.018,
|
| 159 |
+
"eval_steps_per_second": 2.9,
|
| 160 |
+
"step": 478
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 2.092050209205021,
|
| 164 |
+
"grad_norm": 6.472335338592529,
|
| 165 |
+
"learning_rate": 9.079497907949792e-06,
|
| 166 |
+
"loss": 0.13,
|
| 167 |
+
"step": 500
|
| 168 |
+
},
|
| 169 |
+
{
|
| 170 |
+
"epoch": 2.196652719665272,
|
| 171 |
+
"grad_norm": 4.2256760597229,
|
| 172 |
+
"learning_rate": 8.03347280334728e-06,
|
| 173 |
+
"loss": 0.1348,
|
| 174 |
+
"step": 525
|
| 175 |
+
},
|
| 176 |
+
{
|
| 177 |
+
"epoch": 2.301255230125523,
|
| 178 |
+
"grad_norm": 1.4343773126602173,
|
| 179 |
+
"learning_rate": 6.98744769874477e-06,
|
| 180 |
+
"loss": 0.1256,
|
| 181 |
+
"step": 550
|
| 182 |
+
},
|
| 183 |
+
{
|
| 184 |
+
"epoch": 2.405857740585774,
|
| 185 |
+
"grad_norm": 6.69666862487793,
|
| 186 |
+
"learning_rate": 5.941422594142259e-06,
|
| 187 |
+
"loss": 0.1357,
|
| 188 |
+
"step": 575
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"epoch": 2.510460251046025,
|
| 192 |
+
"grad_norm": 4.778820991516113,
|
| 193 |
+
"learning_rate": 4.895397489539749e-06,
|
| 194 |
+
"loss": 0.1353,
|
| 195 |
+
"step": 600
|
| 196 |
+
},
|
| 197 |
+
{
|
| 198 |
+
"epoch": 2.6150627615062763,
|
| 199 |
+
"grad_norm": 6.657689094543457,
|
| 200 |
+
"learning_rate": 3.849372384937238e-06,
|
| 201 |
+
"loss": 0.1236,
|
| 202 |
+
"step": 625
|
| 203 |
+
},
|
| 204 |
+
{
|
| 205 |
+
"epoch": 2.719665271966527,
|
| 206 |
+
"grad_norm": 4.8483405113220215,
|
| 207 |
+
"learning_rate": 2.803347280334728e-06,
|
| 208 |
+
"loss": 0.1533,
|
| 209 |
+
"step": 650
|
| 210 |
+
},
|
| 211 |
+
{
|
| 212 |
+
"epoch": 2.8242677824267783,
|
| 213 |
+
"grad_norm": 4.537696361541748,
|
| 214 |
+
"learning_rate": 1.7573221757322176e-06,
|
| 215 |
+
"loss": 0.1008,
|
| 216 |
+
"step": 675
|
| 217 |
+
},
|
| 218 |
+
{
|
| 219 |
+
"epoch": 2.928870292887029,
|
| 220 |
+
"grad_norm": 4.3347978591918945,
|
| 221 |
+
"learning_rate": 7.112970711297071e-07,
|
| 222 |
+
"loss": 0.1255,
|
| 223 |
+
"step": 700
|
| 224 |
+
},
|
| 225 |
+
{
|
| 226 |
+
"epoch": 3.0,
|
| 227 |
+
"eval_accuracy": 0.932406822488945,
|
| 228 |
+
"eval_loss": 0.23924019932746887,
|
| 229 |
+
"eval_runtime": 6.4354,
|
| 230 |
+
"eval_samples_per_second": 245.982,
|
| 231 |
+
"eval_steps_per_second": 2.642,
|
| 232 |
+
"step": 717
|
| 233 |
+
},
|
| 234 |
+
{
|
| 235 |
+
"epoch": 3.0,
|
| 236 |
+
"step": 717,
|
| 237 |
+
"total_flos": 1020016719882720.0,
|
| 238 |
+
"train_loss": 0.21428075062846205,
|
| 239 |
+
"train_runtime": 119.0204,
|
| 240 |
+
"train_samples_per_second": 288.001,
|
| 241 |
+
"train_steps_per_second": 6.024
|
| 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": 1020016719882720.0,
|
| 262 |
+
"train_batch_size": 48,
|
| 263 |
+
"trial_name": null,
|
| 264 |
+
"trial_params": null
|
| 265 |
+
}
|