Upload batch 11/24: mrpc/mBERT/mean/seed_43 ... mrpc/mBERT/mha_attention/seed_43

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mean/seed_43/all_results.json

@@ -0,0 +1,17 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8529411764705882,
+    "eval_combined_score": 0.8723189636865576,
+    "eval_f1": 0.8916967509025271,
+    "eval_loss": 0.35500794649124146,
+    "eval_runtime": 1.9712,
+    "eval_samples": 408,
+    "eval_samples_per_second": 206.986,
+    "eval_steps_per_second": 2.537,
+    "total_flos": 628599578480640.0,
+    "train_loss": 0.38332160159106893,
+    "train_runtime": 21.6161,
+    "train_samples": 3668,
+    "train_samples_per_second": 509.064,
+    "train_steps_per_second": 10.686
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mean/seed_43/eval_results.json

@@ -0,0 +1,11 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8529411764705882,
+    "eval_combined_score": 0.8723189636865576,
+    "eval_f1": 0.8916967509025271,
+    "eval_loss": 0.35500794649124146,
+    "eval_runtime": 1.9712,
+    "eval_samples": 408,
+    "eval_samples_per_second": 206.986,
+    "eval_steps_per_second": 2.537
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mean/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 mrpc --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/mrpc/mBERT/mean/seed_43 --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

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mean/seed_43/train_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 3.0,
+    "total_flos": 628599578480640.0,
+    "train_loss": 0.38332160159106893,
+    "train_runtime": 21.6161,
+    "train_samples": 3668,
+    "train_samples_per_second": 509.064,
+    "train_steps_per_second": 10.686
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mean/seed_43/trainer_state.json

@@ -0,0 +1,138 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 231,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.3246753246753247,
+      "grad_norm": 2.6699135303497314,
+      "learning_rate": 2.6753246753246755e-05,
+      "loss": 0.6059,
+      "step": 25
+    },
+    {
+      "epoch": 0.6493506493506493,
+      "grad_norm": 4.405506610870361,
+      "learning_rate": 2.350649350649351e-05,
+      "loss": 0.5354,
+      "step": 50
+    },
+    {
+      "epoch": 0.974025974025974,
+      "grad_norm": 4.275317668914795,
+      "learning_rate": 2.025974025974026e-05,
+      "loss": 0.4874,
+      "step": 75
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.7892156862745098,
+      "eval_combined_score": 0.8243464052287581,
+      "eval_f1": 0.8594771241830066,
+      "eval_loss": 0.399752140045166,
+      "eval_runtime": 1.4258,
+      "eval_samples_per_second": 286.162,
+      "eval_steps_per_second": 3.507,
+      "step": 77
+    },
+    {
+      "epoch": 1.2987012987012987,
+      "grad_norm": 10.583173751831055,
+      "learning_rate": 1.7012987012987013e-05,
+      "loss": 0.3923,
+      "step": 100
+    },
+    {
+      "epoch": 1.6233766233766234,
+      "grad_norm": 5.822242259979248,
+      "learning_rate": 1.3766233766233767e-05,
+      "loss": 0.4015,
+      "step": 125
+    },
+    {
+      "epoch": 1.948051948051948,
+      "grad_norm": 11.757969856262207,
+      "learning_rate": 1.051948051948052e-05,
+      "loss": 0.3483,
+      "step": 150
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.8333333333333334,
+      "eval_combined_score": 0.8548484848484849,
+      "eval_f1": 0.8763636363636363,
+      "eval_loss": 0.33591070771217346,
+      "eval_runtime": 1.3904,
+      "eval_samples_per_second": 293.441,
+      "eval_steps_per_second": 3.596,
+      "step": 154
+    },
+    {
+      "epoch": 2.2727272727272725,
+      "grad_norm": 4.565943241119385,
+      "learning_rate": 7.272727272727273e-06,
+      "loss": 0.2479,
+      "step": 175
+    },
+    {
+      "epoch": 2.5974025974025974,
+      "grad_norm": 8.083616256713867,
+      "learning_rate": 4.025974025974026e-06,
+      "loss": 0.2337,
+      "step": 200
+    },
+    {
+      "epoch": 2.9220779220779223,
+      "grad_norm": 13.684831619262695,
+      "learning_rate": 7.792207792207793e-07,
+      "loss": 0.2388,
+      "step": 225
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.8529411764705882,
+      "eval_combined_score": 0.8723189636865576,
+      "eval_f1": 0.8916967509025271,
+      "eval_loss": 0.35500794649124146,
+      "eval_runtime": 1.4953,
+      "eval_samples_per_second": 272.851,
+      "eval_steps_per_second": 3.344,
+      "step": 231
+    },
+    {
+      "epoch": 3.0,
+      "step": 231,
+      "total_flos": 628599578480640.0,
+      "train_loss": 0.38332160159106893,
+      "train_runtime": 21.6161,
+      "train_samples_per_second": 509.064,
+      "train_steps_per_second": 10.686
+    }
+  ],
+  "logging_steps": 25,
+  "max_steps": 231,
+  "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": 628599578480640.0,
+  "train_batch_size": 48,
+  "trial_name": null,
+  "trial_params": null
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mean/seed_44/all_results.json

@@ -0,0 +1,17 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8406862745098039,
+    "eval_combined_score": 0.862203602371181,
+    "eval_f1": 0.8837209302325582,
+    "eval_loss": 0.33879777789115906,
+    "eval_runtime": 1.9291,
+    "eval_samples": 408,
+    "eval_samples_per_second": 211.496,
+    "eval_steps_per_second": 2.592,
+    "total_flos": 629526768279552.0,
+    "train_loss": 0.3880394872648891,
+    "train_runtime": 33.5809,
+    "train_samples": 3668,
+    "train_samples_per_second": 327.687,
+    "train_steps_per_second": 6.879
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mean/seed_44/eval_results.json

@@ -0,0 +1,11 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8406862745098039,
+    "eval_combined_score": 0.862203602371181,
+    "eval_f1": 0.8837209302325582,
+    "eval_loss": 0.33879777789115906,
+    "eval_runtime": 1.9291,
+    "eval_samples": 408,
+    "eval_samples_per_second": 211.496,
+    "eval_steps_per_second": 2.592
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mean/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 mrpc --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/mrpc/mBERT/mean/seed_44 --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

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mean/seed_44/train_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 3.0,
+    "total_flos": 629526768279552.0,
+    "train_loss": 0.3880394872648891,
+    "train_runtime": 33.5809,
+    "train_samples": 3668,
+    "train_samples_per_second": 327.687,
+    "train_steps_per_second": 6.879
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mean/seed_44/trainer_state.json

@@ -0,0 +1,138 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 231,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.3246753246753247,
+      "grad_norm": 2.240074396133423,
+      "learning_rate": 2.6753246753246755e-05,
+      "loss": 0.6137,
+      "step": 25
+    },
+    {
+      "epoch": 0.6493506493506493,
+      "grad_norm": 3.1652536392211914,
+      "learning_rate": 2.350649350649351e-05,
+      "loss": 0.5645,
+      "step": 50
+    },
+    {
+      "epoch": 0.974025974025974,
+      "grad_norm": 5.1598896980285645,
+      "learning_rate": 2.025974025974026e-05,
+      "loss": 0.4544,
+      "step": 75
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.7745098039215687,
+      "eval_combined_score": 0.7918192173134815,
+      "eval_f1": 0.8091286307053942,
+      "eval_loss": 0.4683724343776703,
+      "eval_runtime": 1.6765,
+      "eval_samples_per_second": 243.36,
+      "eval_steps_per_second": 2.982,
+      "step": 77
+    },
+    {
+      "epoch": 1.2987012987012987,
+      "grad_norm": 4.91569185256958,
+      "learning_rate": 1.7012987012987013e-05,
+      "loss": 0.3834,
+      "step": 100
+    },
+    {
+      "epoch": 1.6233766233766234,
+      "grad_norm": 3.673845052719116,
+      "learning_rate": 1.3766233766233767e-05,
+      "loss": 0.386,
+      "step": 125
+    },
+    {
+      "epoch": 1.948051948051948,
+      "grad_norm": 3.953237295150757,
+      "learning_rate": 1.051948051948052e-05,
+      "loss": 0.3633,
+      "step": 150
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.8455882352941176,
+      "eval_combined_score": 0.866644385026738,
+      "eval_f1": 0.8877005347593583,
+      "eval_loss": 0.33358556032180786,
+      "eval_runtime": 1.788,
+      "eval_samples_per_second": 228.186,
+      "eval_steps_per_second": 2.796,
+      "step": 154
+    },
+    {
+      "epoch": 2.2727272727272725,
+      "grad_norm": 7.82267951965332,
+      "learning_rate": 7.272727272727273e-06,
+      "loss": 0.2645,
+      "step": 175
+    },
+    {
+      "epoch": 2.5974025974025974,
+      "grad_norm": 5.962333679199219,
+      "learning_rate": 4.025974025974026e-06,
+      "loss": 0.2523,
+      "step": 200
+    },
+    {
+      "epoch": 2.9220779220779223,
+      "grad_norm": 3.8463003635406494,
+      "learning_rate": 7.792207792207793e-07,
+      "loss": 0.2438,
+      "step": 225
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.8406862745098039,
+      "eval_combined_score": 0.862203602371181,
+      "eval_f1": 0.8837209302325582,
+      "eval_loss": 0.33879777789115906,
+      "eval_runtime": 2.0094,
+      "eval_samples_per_second": 203.043,
+      "eval_steps_per_second": 2.488,
+      "step": 231
+    },
+    {
+      "epoch": 3.0,
+      "step": 231,
+      "total_flos": 629526768279552.0,
+      "train_loss": 0.3880394872648891,
+      "train_runtime": 33.5809,
+      "train_samples_per_second": 327.687,
+      "train_steps_per_second": 6.879
+    }
+  ],
+  "logging_steps": 25,
+  "max_steps": 231,
+  "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": 629526768279552.0,
+  "train_batch_size": 48,
+  "trial_name": null,
+  "trial_params": null
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mha_attention/seed_42/all_results.json

@@ -0,0 +1,17 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8431372549019608,
+    "eval_combined_score": 0.8650315249774821,
+    "eval_f1": 0.8869257950530035,
+    "eval_loss": 0.36945641040802,
+    "eval_runtime": 1.8695,
+    "eval_samples": 408,
+    "eval_samples_per_second": 218.244,
+    "eval_steps_per_second": 2.675,
+    "total_flos": 633506750643624.0,
+    "train_loss": 0.3834533773975455,
+    "train_runtime": 38.5615,
+    "train_samples": 3668,
+    "train_samples_per_second": 285.362,
+    "train_steps_per_second": 5.99
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mha_attention/seed_42/eval_results.json

@@ -0,0 +1,11 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8431372549019608,
+    "eval_combined_score": 0.8650315249774821,
+    "eval_f1": 0.8869257950530035,
+    "eval_loss": 0.36945641040802,
+    "eval_runtime": 1.8695,
+    "eval_samples": 408,
+    "eval_samples_per_second": 218.244,
+    "eval_steps_per_second": 2.675
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mha_attention/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 mrpc --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/mrpc/mBERT/mha_attention/seed_42 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save --use_mha true --pooling_strategy mha_attention --n_heads 4 --head_dim 32 --classifier_hidden_dim 512 --head_dropout 0.05

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mha_attention/seed_42/train_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 3.0,
+    "total_flos": 633506750643624.0,
+    "train_loss": 0.3834533773975455,
+    "train_runtime": 38.5615,
+    "train_samples": 3668,
+    "train_samples_per_second": 285.362,
+    "train_steps_per_second": 5.99
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mha_attention/seed_42/trainer_state.json

@@ -0,0 +1,138 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 231,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.3246753246753247,
+      "grad_norm": 1.7152166366577148,
+      "learning_rate": 2.6753246753246755e-05,
+      "loss": 0.6228,
+      "step": 25
+    },
+    {
+      "epoch": 0.6493506493506493,
+      "grad_norm": 3.873400926589966,
+      "learning_rate": 2.350649350649351e-05,
+      "loss": 0.5341,
+      "step": 50
+    },
+    {
+      "epoch": 0.974025974025974,
+      "grad_norm": 5.573062419891357,
+      "learning_rate": 2.025974025974026e-05,
+      "loss": 0.4935,
+      "step": 75
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.8235294117647058,
+      "eval_combined_score": 0.844850207741089,
+      "eval_f1": 0.8661710037174721,
+      "eval_loss": 0.3878018260002136,
+      "eval_runtime": 2.2134,
+      "eval_samples_per_second": 184.331,
+      "eval_steps_per_second": 2.259,
+      "step": 77
+    },
+    {
+      "epoch": 1.2987012987012987,
+      "grad_norm": 7.445052146911621,
+      "learning_rate": 1.7012987012987013e-05,
+      "loss": 0.3892,
+      "step": 100
+    },
+    {
+      "epoch": 1.6233766233766234,
+      "grad_norm": 4.177548408508301,
+      "learning_rate": 1.3766233766233767e-05,
+      "loss": 0.3684,
+      "step": 125
+    },
+    {
+      "epoch": 1.948051948051948,
+      "grad_norm": 7.246552467346191,
+      "learning_rate": 1.051948051948052e-05,
+      "loss": 0.3352,
+      "step": 150
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.8431372549019608,
+      "eval_combined_score": 0.8635976129582268,
+      "eval_f1": 0.8840579710144928,
+      "eval_loss": 0.3435911238193512,
+      "eval_runtime": 1.8803,
+      "eval_samples_per_second": 216.99,
+      "eval_steps_per_second": 2.659,
+      "step": 154
+    },
+    {
+      "epoch": 2.2727272727272725,
+      "grad_norm": 2.5947506427764893,
+      "learning_rate": 7.272727272727273e-06,
+      "loss": 0.2683,
+      "step": 175
+    },
+    {
+      "epoch": 2.5974025974025974,
+      "grad_norm": 9.742546081542969,
+      "learning_rate": 4.025974025974026e-06,
+      "loss": 0.2405,
+      "step": 200
+    },
+    {
+      "epoch": 2.9220779220779223,
+      "grad_norm": 6.458611488342285,
+      "learning_rate": 7.792207792207793e-07,
+      "loss": 0.2363,
+      "step": 225
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.8431372549019608,
+      "eval_combined_score": 0.8650315249774821,
+      "eval_f1": 0.8869257950530035,
+      "eval_loss": 0.36945641040802,
+      "eval_runtime": 1.8386,
+      "eval_samples_per_second": 221.903,
+      "eval_steps_per_second": 2.719,
+      "step": 231
+    },
+    {
+      "epoch": 3.0,
+      "step": 231,
+      "total_flos": 633506750643624.0,
+      "train_loss": 0.3834533773975455,
+      "train_runtime": 38.5615,
+      "train_samples_per_second": 285.362,
+      "train_steps_per_second": 5.99
+    }
+  ],
+  "logging_steps": 25,
+  "max_steps": 231,
+  "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": 633506750643624.0,
+  "train_batch_size": 48,
+  "trial_name": null,
+  "trial_params": null
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mha_attention/seed_43/all_results.json

@@ -0,0 +1,17 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8480392156862745,
+    "eval_combined_score": 0.8674502647774438,
+    "eval_f1": 0.8868613138686131,
+    "eval_loss": 0.3478466868400574,
+    "eval_runtime": 1.4768,
+    "eval_samples": 408,
+    "eval_samples_per_second": 276.267,
+    "eval_steps_per_second": 3.386,
+    "total_flos": 631477922574840.0,
+    "train_loss": 0.37287821604575944,
+    "train_runtime": 38.9069,
+    "train_samples": 3668,
+    "train_samples_per_second": 282.829,
+    "train_steps_per_second": 5.937
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mha_attention/seed_43/eval_results.json

@@ -0,0 +1,11 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8480392156862745,
+    "eval_combined_score": 0.8674502647774438,
+    "eval_f1": 0.8868613138686131,
+    "eval_loss": 0.3478466868400574,
+    "eval_runtime": 1.4768,
+    "eval_samples": 408,
+    "eval_samples_per_second": 276.267,
+    "eval_steps_per_second": 3.386
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mha_attention/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 mrpc --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/mrpc/mBERT/mha_attention/seed_43 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save --use_mha true --pooling_strategy mha_attention --n_heads 4 --head_dim 32 --classifier_hidden_dim 512 --head_dropout 0.05

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mha_attention/seed_43/train_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 3.0,
+    "total_flos": 631477922574840.0,
+    "train_loss": 0.37287821604575944,
+    "train_runtime": 38.9069,
+    "train_samples": 3668,
+    "train_samples_per_second": 282.829,
+    "train_steps_per_second": 5.937
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mha_attention/seed_43/trainer_state.json

@@ -0,0 +1,138 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 231,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.3246753246753247,
+      "grad_norm": 2.556619644165039,
+      "learning_rate": 2.6753246753246755e-05,
+      "loss": 0.5842,
+      "step": 25
+    },
+    {
+      "epoch": 0.6493506493506493,
+      "grad_norm": 2.4965264797210693,
+      "learning_rate": 2.350649350649351e-05,
+      "loss": 0.5194,
+      "step": 50
+    },
+    {
+      "epoch": 0.974025974025974,
+      "grad_norm": 3.746295213699341,
+      "learning_rate": 2.025974025974026e-05,
+      "loss": 0.481,
+      "step": 75
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.8431372549019608,
+      "eval_combined_score": 0.8678773522831952,
+      "eval_f1": 0.8926174496644296,
+      "eval_loss": 0.39169251918792725,
+      "eval_runtime": 1.9878,
+      "eval_samples_per_second": 205.253,
+      "eval_steps_per_second": 2.515,
+      "step": 77
+    },
+    {
+      "epoch": 1.2987012987012987,
+      "grad_norm": 10.450796127319336,
+      "learning_rate": 1.7012987012987013e-05,
+      "loss": 0.3819,
+      "step": 100
+    },
+    {
+      "epoch": 1.6233766233766234,
+      "grad_norm": 4.166968822479248,
+      "learning_rate": 1.3766233766233767e-05,
+      "loss": 0.3799,
+      "step": 125
+    },
+    {
+      "epoch": 1.948051948051948,
+      "grad_norm": 9.58257007598877,
+      "learning_rate": 1.051948051948052e-05,
+      "loss": 0.343,
+      "step": 150
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.8529411764705882,
+      "eval_combined_score": 0.8694363677029747,
+      "eval_f1": 0.8859315589353612,
+      "eval_loss": 0.35060209035873413,
+      "eval_runtime": 1.8527,
+      "eval_samples_per_second": 220.215,
+      "eval_steps_per_second": 2.699,
+      "step": 154
+    },
+    {
+      "epoch": 2.2727272727272725,
+      "grad_norm": 7.541630268096924,
+      "learning_rate": 7.272727272727273e-06,
+      "loss": 0.2489,
+      "step": 175
+    },
+    {
+      "epoch": 2.5974025974025974,
+      "grad_norm": 5.439375877380371,
+      "learning_rate": 4.025974025974026e-06,
+      "loss": 0.2377,
+      "step": 200
+    },
+    {
+      "epoch": 2.9220779220779223,
+      "grad_norm": 15.038283348083496,
+      "learning_rate": 7.792207792207793e-07,
+      "loss": 0.2224,
+      "step": 225
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.8480392156862745,
+      "eval_combined_score": 0.8674502647774438,
+      "eval_f1": 0.8868613138686131,
+      "eval_loss": 0.3478466868400574,
+      "eval_runtime": 1.6255,
+      "eval_samples_per_second": 251.001,
+      "eval_steps_per_second": 3.076,
+      "step": 231
+    },
+    {
+      "epoch": 3.0,
+      "step": 231,
+      "total_flos": 631477922574840.0,
+      "train_loss": 0.37287821604575944,
+      "train_runtime": 38.9069,
+      "train_samples_per_second": 282.829,
+      "train_steps_per_second": 5.937
+    }
+  ],
+  "logging_steps": 25,
+  "max_steps": 231,
+  "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": 631477922574840.0,
+  "train_batch_size": 48,
+  "trial_name": null,
+  "trial_params": null
+}