Upload batch 12/24: mrpc/mBERT/mha_attention/seed_44 ... mrpc/mBERT/multi_branch_average/seed_44

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

@@ -0,0 +1,17 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8578431372549019,
+    "eval_combined_score": 0.8767632952461559,
+    "eval_f1": 0.89568345323741,
+    "eval_loss": 0.34328851103782654,
+    "eval_runtime": 1.3366,
+    "eval_samples": 408,
+    "eval_samples_per_second": 305.257,
+    "eval_steps_per_second": 3.741,
+    "total_flos": 632409357956112.0,
+    "train_loss": 0.38678704608570447,
+    "train_runtime": 34.893,
+    "train_samples": 3668,
+    "train_samples_per_second": 315.364,
+    "train_steps_per_second": 6.62
+}

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

@@ -0,0 +1,11 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8578431372549019,
+    "eval_combined_score": 0.8767632952461559,
+    "eval_f1": 0.89568345323741,
+    "eval_loss": 0.34328851103782654,
+    "eval_runtime": 1.3366,
+    "eval_samples": 408,
+    "eval_samples_per_second": 305.257,
+    "eval_steps_per_second": 3.741
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mha_attention/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/mha_attention/seed_44 --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_44/train_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 3.0,
+    "total_flos": 632409357956112.0,
+    "train_loss": 0.38678704608570447,
+    "train_runtime": 34.893,
+    "train_samples": 3668,
+    "train_samples_per_second": 315.364,
+    "train_steps_per_second": 6.62
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/mha_attention/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": 2.5029401779174805,
+      "learning_rate": 2.6753246753246755e-05,
+      "loss": 0.6023,
+      "step": 25
+    },
+    {
+      "epoch": 0.6493506493506493,
+      "grad_norm": 2.046332836151123,
+      "learning_rate": 2.350649350649351e-05,
+      "loss": 0.532,
+      "step": 50
+    },
+    {
+      "epoch": 0.974025974025974,
+      "grad_norm": 3.2743406295776367,
+      "learning_rate": 2.025974025974026e-05,
+      "loss": 0.4732,
+      "step": 75
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.7818627450980392,
+      "eval_combined_score": 0.8003000079869015,
+      "eval_f1": 0.8187372708757638,
+      "eval_loss": 0.4533577859401703,
+      "eval_runtime": 1.8731,
+      "eval_samples_per_second": 217.821,
+      "eval_steps_per_second": 2.669,
+      "step": 77
+    },
+    {
+      "epoch": 1.2987012987012987,
+      "grad_norm": 3.9565742015838623,
+      "learning_rate": 1.7012987012987013e-05,
+      "loss": 0.3855,
+      "step": 100
+    },
+    {
+      "epoch": 1.6233766233766234,
+      "grad_norm": 3.86584734916687,
+      "learning_rate": 1.3766233766233767e-05,
+      "loss": 0.3788,
+      "step": 125
+    },
+    {
+      "epoch": 1.948051948051948,
+      "grad_norm": 4.910215377807617,
+      "learning_rate": 1.051948051948052e-05,
+      "loss": 0.388,
+      "step": 150
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.8357843137254902,
+      "eval_combined_score": 0.858600121464515,
+      "eval_f1": 0.8814159292035398,
+      "eval_loss": 0.3557942807674408,
+      "eval_runtime": 1.8551,
+      "eval_samples_per_second": 219.935,
+      "eval_steps_per_second": 2.695,
+      "step": 154
+    },
+    {
+      "epoch": 2.2727272727272725,
+      "grad_norm": 5.348351955413818,
+      "learning_rate": 7.272727272727273e-06,
+      "loss": 0.2601,
+      "step": 175
+    },
+    {
+      "epoch": 2.5974025974025974,
+      "grad_norm": 8.830328941345215,
+      "learning_rate": 4.025974025974026e-06,
+      "loss": 0.2558,
+      "step": 200
+    },
+    {
+      "epoch": 2.9220779220779223,
+      "grad_norm": 3.9247565269470215,
+      "learning_rate": 7.792207792207793e-07,
+      "loss": 0.2408,
+      "step": 225
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.8578431372549019,
+      "eval_combined_score": 0.8767632952461559,
+      "eval_f1": 0.89568345323741,
+      "eval_loss": 0.34328851103782654,
+      "eval_runtime": 1.2657,
+      "eval_samples_per_second": 322.359,
+      "eval_steps_per_second": 3.95,
+      "step": 231
+    },
+    {
+      "epoch": 3.0,
+      "step": 231,
+      "total_flos": 632409357956112.0,
+      "train_loss": 0.38678704608570447,
+      "train_runtime": 34.893,
+      "train_samples_per_second": 315.364,
+      "train_steps_per_second": 6.62
+    }
+  ],
+  "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": 632409357956112.0,
+  "train_batch_size": 48,
+  "trial_name": null,
+  "trial_params": null
+}

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

@@ -0,0 +1,17 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8504901960784313,
+    "eval_combined_score": 0.8714532109139952,
+    "eval_f1": 0.892416225749559,
+    "eval_loss": 0.3461981415748596,
+    "eval_runtime": 1.9053,
+    "eval_samples": 408,
+    "eval_samples_per_second": 214.136,
+    "eval_steps_per_second": 2.624,
+    "total_flos": 650809756309752.0,
+    "train_loss": 0.38643745581309,
+    "train_runtime": 27.189,
+    "train_samples": 3668,
+    "train_samples_per_second": 404.722,
+    "train_steps_per_second": 8.496
+}

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

@@ -0,0 +1,11 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8504901960784313,
+    "eval_combined_score": 0.8714532109139952,
+    "eval_f1": 0.892416225749559,
+    "eval_loss": 0.3461981415748596,
+    "eval_runtime": 1.9053,
+    "eval_samples": 408,
+    "eval_samples_per_second": 214.136,
+    "eval_steps_per_second": 2.624
+}

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

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

@@ -0,0 +1,9 @@
+{
+    "epoch": 3.0,
+    "total_flos": 650809756309752.0,
+    "train_loss": 0.38643745581309,
+    "train_runtime": 27.189,
+    "train_samples": 3668,
+    "train_samples_per_second": 404.722,
+    "train_steps_per_second": 8.496
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/multi_branch_average/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": 3.312774181365967,
+      "learning_rate": 2.6753246753246755e-05,
+      "loss": 0.612,
+      "step": 25
+    },
+    {
+      "epoch": 0.6493506493506493,
+      "grad_norm": 7.789833068847656,
+      "learning_rate": 2.350649350649351e-05,
+      "loss": 0.5163,
+      "step": 50
+    },
+    {
+      "epoch": 0.974025974025974,
+      "grad_norm": 4.037544250488281,
+      "learning_rate": 2.025974025974026e-05,
+      "loss": 0.5178,
+      "step": 75
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.7696078431372549,
+      "eval_combined_score": 0.7876964835520985,
+      "eval_f1": 0.8057851239669421,
+      "eval_loss": 0.4581141471862793,
+      "eval_runtime": 1.6688,
+      "eval_samples_per_second": 244.493,
+      "eval_steps_per_second": 2.996,
+      "step": 77
+    },
+    {
+      "epoch": 1.2987012987012987,
+      "grad_norm": 5.829493522644043,
+      "learning_rate": 1.7012987012987013e-05,
+      "loss": 0.4002,
+      "step": 100
+    },
+    {
+      "epoch": 1.6233766233766234,
+      "grad_norm": 5.771215438842773,
+      "learning_rate": 1.3766233766233767e-05,
+      "loss": 0.3759,
+      "step": 125
+    },
+    {
+      "epoch": 1.948051948051948,
+      "grad_norm": 10.886384963989258,
+      "learning_rate": 1.051948051948052e-05,
+      "loss": 0.3424,
+      "step": 150
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.8627450980392157,
+      "eval_combined_score": 0.8820767743717206,
+      "eval_f1": 0.9014084507042254,
+      "eval_loss": 0.3224419951438904,
+      "eval_runtime": 1.2952,
+      "eval_samples_per_second": 315.006,
+      "eval_steps_per_second": 3.86,
+      "step": 154
+    },
+    {
+      "epoch": 2.2727272727272725,
+      "grad_norm": 5.9191155433654785,
+      "learning_rate": 7.272727272727273e-06,
+      "loss": 0.2623,
+      "step": 175
+    },
+    {
+      "epoch": 2.5974025974025974,
+      "grad_norm": 11.66323184967041,
+      "learning_rate": 4.025974025974026e-06,
+      "loss": 0.2459,
+      "step": 200
+    },
+    {
+      "epoch": 2.9220779220779223,
+      "grad_norm": 7.454559803009033,
+      "learning_rate": 7.792207792207793e-07,
+      "loss": 0.2329,
+      "step": 225
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.8504901960784313,
+      "eval_combined_score": 0.8714532109139952,
+      "eval_f1": 0.892416225749559,
+      "eval_loss": 0.3461981415748596,
+      "eval_runtime": 1.5564,
+      "eval_samples_per_second": 262.141,
+      "eval_steps_per_second": 3.213,
+      "step": 231
+    },
+    {
+      "epoch": 3.0,
+      "step": 231,
+      "total_flos": 650809756309752.0,
+      "train_loss": 0.38643745581309,
+      "train_runtime": 27.189,
+      "train_samples_per_second": 404.722,
+      "train_steps_per_second": 8.496
+    }
+  ],
+  "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": 650809756309752.0,
+  "train_batch_size": 48,
+  "trial_name": null,
+  "trial_params": null
+}

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

@@ -0,0 +1,17 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8455882352941176,
+    "eval_combined_score": 0.8670419052576783,
+    "eval_f1": 0.8884955752212389,
+    "eval_loss": 0.35154372453689575,
+    "eval_runtime": 1.8306,
+    "eval_samples": 408,
+    "eval_samples_per_second": 222.879,
+    "eval_steps_per_second": 2.731,
+    "total_flos": 648725514745320.0,
+    "train_loss": 0.4033502381601375,
+    "train_runtime": 38.279,
+    "train_samples": 3668,
+    "train_samples_per_second": 287.468,
+    "train_steps_per_second": 6.035
+}

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

@@ -0,0 +1,11 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8455882352941176,
+    "eval_combined_score": 0.8670419052576783,
+    "eval_f1": 0.8884955752212389,
+    "eval_loss": 0.35154372453689575,
+    "eval_runtime": 1.8306,
+    "eval_samples": 408,
+    "eval_samples_per_second": 222.879,
+    "eval_steps_per_second": 2.731
+}

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

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

@@ -0,0 +1,9 @@
+{
+    "epoch": 3.0,
+    "total_flos": 648725514745320.0,
+    "train_loss": 0.4033502381601375,
+    "train_runtime": 38.279,
+    "train_samples": 3668,
+    "train_samples_per_second": 287.468,
+    "train_steps_per_second": 6.035
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/multi_branch_average/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": 2.91154408454895,
+      "learning_rate": 2.6753246753246755e-05,
+      "loss": 0.6178,
+      "step": 25
+    },
+    {
+      "epoch": 0.6493506493506493,
+      "grad_norm": 4.3118157386779785,
+      "learning_rate": 2.350649350649351e-05,
+      "loss": 0.5435,
+      "step": 50
+    },
+    {
+      "epoch": 0.974025974025974,
+      "grad_norm": 5.843005180358887,
+      "learning_rate": 2.025974025974026e-05,
+      "loss": 0.5119,
+      "step": 75
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.8063725490196079,
+      "eval_combined_score": 0.8374624808325992,
+      "eval_f1": 0.8685524126455907,
+      "eval_loss": 0.435122013092041,
+      "eval_runtime": 1.7876,
+      "eval_samples_per_second": 228.24,
+      "eval_steps_per_second": 2.797,
+      "step": 77
+    },
+    {
+      "epoch": 1.2987012987012987,
+      "grad_norm": 6.603716850280762,
+      "learning_rate": 1.7012987012987013e-05,
+      "loss": 0.4224,
+      "step": 100
+    },
+    {
+      "epoch": 1.6233766233766234,
+      "grad_norm": 2.9754514694213867,
+      "learning_rate": 1.3766233766233767e-05,
+      "loss": 0.4026,
+      "step": 125
+    },
+    {
+      "epoch": 1.948051948051948,
+      "grad_norm": 10.035341262817383,
+      "learning_rate": 1.051948051948052e-05,
+      "loss": 0.3765,
+      "step": 150
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.8333333333333334,
+      "eval_combined_score": 0.8546228710462287,
+      "eval_f1": 0.8759124087591241,
+      "eval_loss": 0.34409093856811523,
+      "eval_runtime": 1.6765,
+      "eval_samples_per_second": 243.366,
+      "eval_steps_per_second": 2.982,
+      "step": 154
+    },
+    {
+      "epoch": 2.2727272727272725,
+      "grad_norm": 5.992510795593262,
+      "learning_rate": 7.272727272727273e-06,
+      "loss": 0.2721,
+      "step": 175
+    },
+    {
+      "epoch": 2.5974025974025974,
+      "grad_norm": 6.607265949249268,
+      "learning_rate": 4.025974025974026e-06,
+      "loss": 0.2613,
+      "step": 200
+    },
+    {
+      "epoch": 2.9220779220779223,
+      "grad_norm": 10.01749324798584,
+      "learning_rate": 7.792207792207793e-07,
+      "loss": 0.2655,
+      "step": 225
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.8455882352941176,
+      "eval_combined_score": 0.8670419052576783,
+      "eval_f1": 0.8884955752212389,
+      "eval_loss": 0.35154372453689575,
+      "eval_runtime": 2.0795,
+      "eval_samples_per_second": 196.205,
+      "eval_steps_per_second": 2.404,
+      "step": 231
+    },
+    {
+      "epoch": 3.0,
+      "step": 231,
+      "total_flos": 648725514745320.0,
+      "train_loss": 0.4033502381601375,
+      "train_runtime": 38.279,
+      "train_samples_per_second": 287.468,
+      "train_steps_per_second": 6.035
+    }
+  ],
+  "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": 648725514745320.0,
+  "train_batch_size": 48,
+  "trial_name": null,
+  "trial_params": null
+}

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

@@ -0,0 +1,17 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8602941176470589,
+    "eval_combined_score": 0.878422920892495,
+    "eval_f1": 0.896551724137931,
+    "eval_loss": 0.33914050459861755,
+    "eval_runtime": 1.9691,
+    "eval_samples": 408,
+    "eval_samples_per_second": 207.202,
+    "eval_steps_per_second": 2.539,
+    "total_flos": 649682390473776.0,
+    "train_loss": 0.39075296472161364,
+    "train_runtime": 39.956,
+    "train_samples": 3668,
+    "train_samples_per_second": 275.403,
+    "train_steps_per_second": 5.781
+}

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

@@ -0,0 +1,11 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8602941176470589,
+    "eval_combined_score": 0.878422920892495,
+    "eval_f1": 0.896551724137931,
+    "eval_loss": 0.33914050459861755,
+    "eval_runtime": 1.9691,
+    "eval_samples": 408,
+    "eval_samples_per_second": 207.202,
+    "eval_steps_per_second": 2.539
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/multi_branch_average/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/multi_branch_average/seed_44 --overwrite_output_dir --trust_remote_code true --bf16 --tf32 true --skip_model_save --use_mha true --pooling_strategy multi_branch_average --n_heads 4 --head_dim 32 --classifier_hidden_dim 512 --head_dropout 0.05

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

@@ -0,0 +1,9 @@
+{
+    "epoch": 3.0,
+    "total_flos": 649682390473776.0,
+    "train_loss": 0.39075296472161364,
+    "train_runtime": 39.956,
+    "train_samples": 3668,
+    "train_samples_per_second": 275.403,
+    "train_steps_per_second": 5.781
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/multi_branch_average/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": 2.525108575820923,
+      "learning_rate": 2.6753246753246755e-05,
+      "loss": 0.6028,
+      "step": 25
+    },
+    {
+      "epoch": 0.6493506493506493,
+      "grad_norm": 2.730511426925659,
+      "learning_rate": 2.350649350649351e-05,
+      "loss": 0.538,
+      "step": 50
+    },
+    {
+      "epoch": 0.974025974025974,
+      "grad_norm": 4.872249603271484,
+      "learning_rate": 2.025974025974026e-05,
+      "loss": 0.4702,
+      "step": 75
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.7769607843137255,
+      "eval_combined_score": 0.7950512340459798,
+      "eval_f1": 0.813141683778234,
+      "eval_loss": 0.4498041272163391,
+      "eval_runtime": 2.0371,
+      "eval_samples_per_second": 200.281,
+      "eval_steps_per_second": 2.454,
+      "step": 77
+    },
+    {
+      "epoch": 1.2987012987012987,
+      "grad_norm": 3.917067527770996,
+      "learning_rate": 1.7012987012987013e-05,
+      "loss": 0.3866,
+      "step": 100
+    },
+    {
+      "epoch": 1.6233766233766234,
+      "grad_norm": 3.84232759475708,
+      "learning_rate": 1.3766233766233767e-05,
+      "loss": 0.3887,
+      "step": 125
+    },
+    {
+      "epoch": 1.948051948051948,
+      "grad_norm": 4.048624515533447,
+      "learning_rate": 1.051948051948052e-05,
+      "loss": 0.3842,
+      "step": 150
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.8455882352941176,
+      "eval_combined_score": 0.8662411553490337,
+      "eval_f1": 0.8868940754039497,
+      "eval_loss": 0.33782169222831726,
+      "eval_runtime": 2.1985,
+      "eval_samples_per_second": 185.577,
+      "eval_steps_per_second": 2.274,
+      "step": 154
+    },
+    {
+      "epoch": 2.2727272727272725,
+      "grad_norm": 5.455221652984619,
+      "learning_rate": 7.272727272727273e-06,
+      "loss": 0.2804,
+      "step": 175
+    },
+    {
+      "epoch": 2.5974025974025974,
+      "grad_norm": 8.960445404052734,
+      "learning_rate": 4.025974025974026e-06,
+      "loss": 0.2439,
+      "step": 200
+    },
+    {
+      "epoch": 2.9220779220779223,
+      "grad_norm": 4.210287570953369,
+      "learning_rate": 7.792207792207793e-07,
+      "loss": 0.256,
+      "step": 225
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.8602941176470589,
+      "eval_combined_score": 0.878422920892495,
+      "eval_f1": 0.896551724137931,
+      "eval_loss": 0.33914050459861755,
+      "eval_runtime": 1.8302,
+      "eval_samples_per_second": 222.925,
+      "eval_steps_per_second": 2.732,
+      "step": 231
+    },
+    {
+      "epoch": 3.0,
+      "step": 231,
+      "total_flos": 649682390473776.0,
+      "train_loss": 0.39075296472161364,
+      "train_runtime": 39.956,
+      "train_samples_per_second": 275.403,
+      "train_steps_per_second": 5.781
+    }
+  ],
+  "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": 649682390473776.0,
+  "train_batch_size": 48,
+  "trial_name": null,
+  "trial_params": null
+}