Upload batch 22/24: vsfc/mBERT/max/seed_42 ... vsfc/mBERT/mean/seed_42

mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/max/seed_42/run_command.txt

@@ -0,0 +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/max/seed_42 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save --use_mha true --pooling_strategy max --n_heads 4 --head_dim 32 --classifier_hidden_dim 512 --head_dropout 0.05

mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/max/seed_43/all_results.json

@@ -0,0 +1,15 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.9317751105495894,
+    "eval_loss": 0.23970364034175873,
+    "eval_runtime": 5.8795,
+    "eval_samples": 1583,
+    "eval_samples_per_second": 269.241,
+    "eval_steps_per_second": 2.891,
+    "total_flos": 1022750124392448.0,
+    "train_loss": 0.20938005839764157,
+    "train_runtime": 114.2677,
+    "train_samples": 11426,
+    "train_samples_per_second": 299.98,
+    "train_steps_per_second": 6.275
+}

mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/max/seed_43/eval_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.9317751105495894,
+    "eval_loss": 0.23970364034175873,
+    "eval_runtime": 5.8795,
+    "eval_samples": 1583,
+    "eval_samples_per_second": 269.241,
+    "eval_steps_per_second": 2.891
+}

mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/max/seed_43/run_command.txt

@@ -0,0 +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/max/seed_43 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save --use_mha true --pooling_strategy max --n_heads 4 --head_dim 32 --classifier_hidden_dim 512 --head_dropout 0.05

mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/max/seed_43/train_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 3.0,
+    "total_flos": 1022750124392448.0,
+    "train_loss": 0.20938005839764157,
+    "train_runtime": 114.2677,
+    "train_samples": 11426,
+    "train_samples_per_second": 299.98,
+    "train_steps_per_second": 6.275
+}

mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/max/seed_43/trainer_state.json

@@ -0,0 +1,265 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 717,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.10460251046025104,
+      "grad_norm": 6.728825092315674,
+      "learning_rate": 2.8953974895397488e-05,
+      "loss": 0.5651,
+      "step": 25
+    },
+    {
+      "epoch": 0.20920502092050208,
+      "grad_norm": 7.908098220825195,
+      "learning_rate": 2.7907949790794982e-05,
+      "loss": 0.3427,
+      "step": 50
+    },
+    {
+      "epoch": 0.3138075313807531,
+      "grad_norm": 5.102977752685547,
+      "learning_rate": 2.686192468619247e-05,
+      "loss": 0.3305,
+      "step": 75
+    },
+    {
+      "epoch": 0.41841004184100417,
+      "grad_norm": 3.49076247215271,
+      "learning_rate": 2.5815899581589957e-05,
+      "loss": 0.3034,
+      "step": 100
+    },
+    {
+      "epoch": 0.5230125523012552,
+      "grad_norm": 8.612030029296875,
+      "learning_rate": 2.4769874476987448e-05,
+      "loss": 0.3254,
+      "step": 125
+    },
+    {
+      "epoch": 0.6276150627615062,
+      "grad_norm": 3.783565044403076,
+      "learning_rate": 2.372384937238494e-05,
+      "loss": 0.2945,
+      "step": 150
+    },
+    {
+      "epoch": 0.7322175732217573,
+      "grad_norm": 4.309210777282715,
+      "learning_rate": 2.2677824267782426e-05,
+      "loss": 0.264,
+      "step": 175
+    },
+    {
+      "epoch": 0.8368200836820083,
+      "grad_norm": 4.5513763427734375,
+      "learning_rate": 2.1631799163179917e-05,
+      "loss": 0.2276,
+      "step": 200
+    },
+    {
+      "epoch": 0.9414225941422594,
+      "grad_norm": 4.058280944824219,
+      "learning_rate": 2.0585774058577408e-05,
+      "loss": 0.2444,
+      "step": 225
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.9248262792166771,
+      "eval_loss": 0.22704137861728668,
+      "eval_runtime": 6.82,
+      "eval_samples_per_second": 232.113,
+      "eval_steps_per_second": 2.493,
+      "step": 239
+    },
+    {
+      "epoch": 1.0460251046025104,
+      "grad_norm": 3.737504720687866,
+      "learning_rate": 1.9539748953974895e-05,
+      "loss": 0.2356,
+      "step": 250
+    },
+    {
+      "epoch": 1.1506276150627615,
+      "grad_norm": 3.996934652328491,
+      "learning_rate": 1.8493723849372386e-05,
+      "loss": 0.1916,
+      "step": 275
+    },
+    {
+      "epoch": 1.2552301255230125,
+      "grad_norm": 6.380361557006836,
+      "learning_rate": 1.7447698744769877e-05,
+      "loss": 0.2191,
+      "step": 300
+    },
+    {
+      "epoch": 1.3598326359832635,
+      "grad_norm": 9.467670440673828,
+      "learning_rate": 1.6401673640167365e-05,
+      "loss": 0.1908,
+      "step": 325
+    },
+    {
+      "epoch": 1.4644351464435146,
+      "grad_norm": 3.292236566543579,
+      "learning_rate": 1.5355648535564852e-05,
+      "loss": 0.1608,
+      "step": 350
+    },
+    {
+      "epoch": 1.5690376569037658,
+      "grad_norm": 3.0692200660705566,
+      "learning_rate": 1.4309623430962344e-05,
+      "loss": 0.1928,
+      "step": 375
+    },
+    {
+      "epoch": 1.6736401673640167,
+      "grad_norm": 4.897969722747803,
+      "learning_rate": 1.3263598326359832e-05,
+      "loss": 0.1785,
+      "step": 400
+    },
+    {
+      "epoch": 1.778242677824268,
+      "grad_norm": 13.100573539733887,
+      "learning_rate": 1.2217573221757323e-05,
+      "loss": 0.1813,
+      "step": 425
+    },
+    {
+      "epoch": 1.8828451882845187,
+      "grad_norm": 3.577944755554199,
+      "learning_rate": 1.1171548117154812e-05,
+      "loss": 0.166,
+      "step": 450
+    },
+    {
+      "epoch": 1.98744769874477,
+      "grad_norm": 3.756112575531006,
+      "learning_rate": 1.0125523012552301e-05,
+      "loss": 0.177,
+      "step": 475
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.934301958307012,
+      "eval_loss": 0.22262610495090485,
+      "eval_runtime": 6.3532,
+      "eval_samples_per_second": 249.164,
+      "eval_steps_per_second": 2.676,
+      "step": 478
+    },
+    {
+      "epoch": 2.092050209205021,
+      "grad_norm": 7.469801425933838,
+      "learning_rate": 9.079497907949792e-06,
+      "loss": 0.1089,
+      "step": 500
+    },
+    {
+      "epoch": 2.196652719665272,
+      "grad_norm": 8.408180236816406,
+      "learning_rate": 8.03347280334728e-06,
+      "loss": 0.1333,
+      "step": 525
+    },
+    {
+      "epoch": 2.301255230125523,
+      "grad_norm": 0.488555908203125,
+      "learning_rate": 6.98744769874477e-06,
+      "loss": 0.1267,
+      "step": 550
+    },
+    {
+      "epoch": 2.405857740585774,
+      "grad_norm": 5.445498943328857,
+      "learning_rate": 5.941422594142259e-06,
+      "loss": 0.1297,
+      "step": 575
+    },
+    {
+      "epoch": 2.510460251046025,
+      "grad_norm": 4.109771728515625,
+      "learning_rate": 4.895397489539749e-06,
+      "loss": 0.1332,
+      "step": 600
+    },
+    {
+      "epoch": 2.6150627615062763,
+      "grad_norm": 3.7967689037323,
+      "learning_rate": 3.849372384937238e-06,
+      "loss": 0.1096,
+      "step": 625
+    },
+    {
+      "epoch": 2.719665271966527,
+      "grad_norm": 0.5358346700668335,
+      "learning_rate": 2.803347280334728e-06,
+      "loss": 0.1369,
+      "step": 650
+    },
+    {
+      "epoch": 2.8242677824267783,
+      "grad_norm": 4.2060651779174805,
+      "learning_rate": 1.7573221757322176e-06,
+      "loss": 0.1277,
+      "step": 675
+    },
+    {
+      "epoch": 2.928870292887029,
+      "grad_norm": 4.618159770965576,
+      "learning_rate": 7.112970711297071e-07,
+      "loss": 0.1439,
+      "step": 700
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.9317751105495894,
+      "eval_loss": 0.23970364034175873,
+      "eval_runtime": 5.5616,
+      "eval_samples_per_second": 284.628,
+      "eval_steps_per_second": 3.057,
+      "step": 717
+    },
+    {
+      "epoch": 3.0,
+      "step": 717,
+      "total_flos": 1022750124392448.0,
+      "train_loss": 0.20938005839764157,
+      "train_runtime": 114.2677,
+      "train_samples_per_second": 299.98,
+      "train_steps_per_second": 6.275
+    }
+  ],
+  "logging_steps": 25,
+  "max_steps": 717,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1022750124392448.0,
+  "train_batch_size": 48,
+  "trial_name": null,
+  "trial_params": null
+}

mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/max/seed_44/all_results.json

@@ -0,0 +1,15 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.9330385344283006,
+    "eval_loss": 0.2280229926109314,
+    "eval_runtime": 6.1519,
+    "eval_samples": 1583,
+    "eval_samples_per_second": 257.32,
+    "eval_steps_per_second": 2.763,
+    "total_flos": 1024705701150720.0,
+    "train_loss": 0.20747514532533484,
+    "train_runtime": 114.3541,
+    "train_samples": 11426,
+    "train_samples_per_second": 299.753,
+    "train_steps_per_second": 6.27
+}

mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/max/seed_44/eval_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.9330385344283006,
+    "eval_loss": 0.2280229926109314,
+    "eval_runtime": 6.1519,
+    "eval_samples": 1583,
+    "eval_samples_per_second": 257.32,
+    "eval_steps_per_second": 2.763
+}

mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/max/seed_44/run_command.txt

@@ -0,0 +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/max/seed_44 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save --use_mha true --pooling_strategy max --n_heads 4 --head_dim 32 --classifier_hidden_dim 512 --head_dropout 0.05

mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/max/seed_44/train_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 3.0,
+    "total_flos": 1024705701150720.0,
+    "train_loss": 0.20747514532533484,
+    "train_runtime": 114.3541,
+    "train_samples": 11426,
+    "train_samples_per_second": 299.753,
+    "train_steps_per_second": 6.27
+}

mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/max/seed_44/trainer_state.json

@@ -0,0 +1,265 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 717,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.10460251046025104,
+      "grad_norm": 6.769598484039307,
+      "learning_rate": 2.8953974895397488e-05,
+      "loss": 0.5558,
+      "step": 25
+    },
+    {
+      "epoch": 0.20920502092050208,
+      "grad_norm": 4.832651138305664,
+      "learning_rate": 2.7907949790794982e-05,
+      "loss": 0.3485,
+      "step": 50
+    },
+    {
+      "epoch": 0.3138075313807531,
+      "grad_norm": 4.488345623016357,
+      "learning_rate": 2.686192468619247e-05,
+      "loss": 0.3463,
+      "step": 75
+    },
+    {
+      "epoch": 0.41841004184100417,
+      "grad_norm": 2.1114728450775146,
+      "learning_rate": 2.5815899581589957e-05,
+      "loss": 0.279,
+      "step": 100
+    },
+    {
+      "epoch": 0.5230125523012552,
+      "grad_norm": 5.334822177886963,
+      "learning_rate": 2.4769874476987448e-05,
+      "loss": 0.2882,
+      "step": 125
+    },
+    {
+      "epoch": 0.6276150627615062,
+      "grad_norm": 4.173310279846191,
+      "learning_rate": 2.372384937238494e-05,
+      "loss": 0.2539,
+      "step": 150
+    },
+    {
+      "epoch": 0.7322175732217573,
+      "grad_norm": 7.422400951385498,
+      "learning_rate": 2.2677824267782426e-05,
+      "loss": 0.3082,
+      "step": 175
+    },
+    {
+      "epoch": 0.8368200836820083,
+      "grad_norm": 4.0469560623168945,
+      "learning_rate": 2.1631799163179917e-05,
+      "loss": 0.22,
+      "step": 200
+    },
+    {
+      "epoch": 0.9414225941422594,
+      "grad_norm": 6.489193439483643,
+      "learning_rate": 2.0585774058577408e-05,
+      "loss": 0.2361,
+      "step": 225
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.9216677195198989,
+      "eval_loss": 0.23060229420661926,
+      "eval_runtime": 6.7396,
+      "eval_samples_per_second": 234.882,
+      "eval_steps_per_second": 2.522,
+      "step": 239
+    },
+    {
+      "epoch": 1.0460251046025104,
+      "grad_norm": 5.142463684082031,
+      "learning_rate": 1.9539748953974895e-05,
+      "loss": 0.1979,
+      "step": 250
+    },
+    {
+      "epoch": 1.1506276150627615,
+      "grad_norm": 7.232017993927002,
+      "learning_rate": 1.8493723849372386e-05,
+      "loss": 0.1866,
+      "step": 275
+    },
+    {
+      "epoch": 1.2552301255230125,
+      "grad_norm": 3.043032169342041,
+      "learning_rate": 1.7447698744769877e-05,
+      "loss": 0.1892,
+      "step": 300
+    },
+    {
+      "epoch": 1.3598326359832635,
+      "grad_norm": 6.970592021942139,
+      "learning_rate": 1.6401673640167365e-05,
+      "loss": 0.1975,
+      "step": 325
+    },
+    {
+      "epoch": 1.4644351464435146,
+      "grad_norm": 4.755401611328125,
+      "learning_rate": 1.5355648535564852e-05,
+      "loss": 0.1931,
+      "step": 350
+    },
+    {
+      "epoch": 1.5690376569037658,
+      "grad_norm": 7.054227828979492,
+      "learning_rate": 1.4309623430962344e-05,
+      "loss": 0.179,
+      "step": 375
+    },
+    {
+      "epoch": 1.6736401673640167,
+      "grad_norm": 6.783203601837158,
+      "learning_rate": 1.3263598326359832e-05,
+      "loss": 0.1963,
+      "step": 400
+    },
+    {
+      "epoch": 1.778242677824268,
+      "grad_norm": 3.4103643894195557,
+      "learning_rate": 1.2217573221757323e-05,
+      "loss": 0.1996,
+      "step": 425
+    },
+    {
+      "epoch": 1.8828451882845187,
+      "grad_norm": 6.070463180541992,
+      "learning_rate": 1.1171548117154812e-05,
+      "loss": 0.1977,
+      "step": 450
+    },
+    {
+      "epoch": 1.98744769874477,
+      "grad_norm": 3.7371246814727783,
+      "learning_rate": 1.0125523012552301e-05,
+      "loss": 0.1638,
+      "step": 475
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.9330385344283006,
+      "eval_loss": 0.23419837653636932,
+      "eval_runtime": 6.5683,
+      "eval_samples_per_second": 241.007,
+      "eval_steps_per_second": 2.588,
+      "step": 478
+    },
+    {
+      "epoch": 2.092050209205021,
+      "grad_norm": 9.284645080566406,
+      "learning_rate": 9.079497907949792e-06,
+      "loss": 0.1358,
+      "step": 500
+    },
+    {
+      "epoch": 2.196652719665272,
+      "grad_norm": 3.2133939266204834,
+      "learning_rate": 8.03347280334728e-06,
+      "loss": 0.1323,
+      "step": 525
+    },
+    {
+      "epoch": 2.301255230125523,
+      "grad_norm": 2.1595170497894287,
+      "learning_rate": 6.98744769874477e-06,
+      "loss": 0.1245,
+      "step": 550
+    },
+    {
+      "epoch": 2.405857740585774,
+      "grad_norm": 3.053664207458496,
+      "learning_rate": 5.941422594142259e-06,
+      "loss": 0.1292,
+      "step": 575
+    },
+    {
+      "epoch": 2.510460251046025,
+      "grad_norm": 2.4191031455993652,
+      "learning_rate": 4.895397489539749e-06,
+      "loss": 0.11,
+      "step": 600
+    },
+    {
+      "epoch": 2.6150627615062763,
+      "grad_norm": 4.869324207305908,
+      "learning_rate": 3.849372384937238e-06,
+      "loss": 0.1259,
+      "step": 625
+    },
+    {
+      "epoch": 2.719665271966527,
+      "grad_norm": 4.810281276702881,
+      "learning_rate": 2.803347280334728e-06,
+      "loss": 0.1491,
+      "step": 650
+    },
+    {
+      "epoch": 2.8242677824267783,
+      "grad_norm": 3.5581533908843994,
+      "learning_rate": 1.7573221757322176e-06,
+      "loss": 0.1066,
+      "step": 675
+    },
+    {
+      "epoch": 2.928870292887029,
+      "grad_norm": 3.9768917560577393,
+      "learning_rate": 7.112970711297071e-07,
+      "loss": 0.1213,
+      "step": 700
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.9330385344283006,
+      "eval_loss": 0.2280229926109314,
+      "eval_runtime": 5.5941,
+      "eval_samples_per_second": 282.978,
+      "eval_steps_per_second": 3.039,
+      "step": 717
+    },
+    {
+      "epoch": 3.0,
+      "step": 717,
+      "total_flos": 1024705701150720.0,
+      "train_loss": 0.20747514532533484,
+      "train_runtime": 114.3541,
+      "train_samples_per_second": 299.753,
+      "train_steps_per_second": 6.27
+    }
+  ],
+  "logging_steps": 25,
+  "max_steps": 717,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1024705701150720.0,
+  "train_batch_size": 48,
+  "trial_name": null,
+  "trial_params": null
+}

mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/mean/seed_42/run_command.txt

@@ -0,0 +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/mean/seed_42 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save --use_mha true --pooling_strategy mean --n_heads 4 --head_dim 32 --classifier_hidden_dim 512 --head_dropout 0.05