Upload batch 8/24: mrpc/mBERT/cls/seed_43 ... mrpc/mBERT/gated_multi_branch/seed_43

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

@@ -0,0 +1,17 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8455882352941176,
+    "eval_combined_score": 0.8658320923306031,
+    "eval_f1": 0.8860759493670886,
+    "eval_loss": 0.3552929759025574,
+    "eval_runtime": 1.4664,
+    "eval_samples": 408,
+    "eval_samples_per_second": 278.234,
+    "eval_steps_per_second": 3.41,
+    "total_flos": 628599578480640.0,
+    "train_loss": 0.38404927728496074,
+    "train_runtime": 38.3271,
+    "train_samples": 3668,
+    "train_samples_per_second": 287.108,
+    "train_steps_per_second": 6.027
+}

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

@@ -0,0 +1,11 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8455882352941176,
+    "eval_combined_score": 0.8658320923306031,
+    "eval_f1": 0.8860759493670886,
+    "eval_loss": 0.3552929759025574,
+    "eval_runtime": 1.4664,
+    "eval_samples": 408,
+    "eval_samples_per_second": 278.234,
+    "eval_steps_per_second": 3.41
+}

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

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

@@ -0,0 +1,9 @@
+{
+    "epoch": 3.0,
+    "total_flos": 628599578480640.0,
+    "train_loss": 0.38404927728496074,
+    "train_runtime": 38.3271,
+    "train_samples": 3668,
+    "train_samples_per_second": 287.108,
+    "train_steps_per_second": 6.027
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/cls/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": 3.477534055709839,
+      "learning_rate": 2.6753246753246755e-05,
+      "loss": 0.5857,
+      "step": 25
+    },
+    {
+      "epoch": 0.6493506493506493,
+      "grad_norm": 5.610284805297852,
+      "learning_rate": 2.350649350649351e-05,
+      "loss": 0.5375,
+      "step": 50
+    },
+    {
+      "epoch": 0.974025974025974,
+      "grad_norm": 3.8447365760803223,
+      "learning_rate": 2.025974025974026e-05,
+      "loss": 0.4838,
+      "step": 75
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.8063725490196079,
+      "eval_combined_score": 0.8378970183114568,
+      "eval_f1": 0.8694214876033057,
+      "eval_loss": 0.39049965143203735,
+      "eval_runtime": 1.948,
+      "eval_samples_per_second": 209.445,
+      "eval_steps_per_second": 2.567,
+      "step": 77
+    },
+    {
+      "epoch": 1.2987012987012987,
+      "grad_norm": 8.335118293762207,
+      "learning_rate": 1.7012987012987013e-05,
+      "loss": 0.385,
+      "step": 100
+    },
+    {
+      "epoch": 1.6233766233766234,
+      "grad_norm": 5.412845134735107,
+      "learning_rate": 1.3766233766233767e-05,
+      "loss": 0.3788,
+      "step": 125
+    },
+    {
+      "epoch": 1.948051948051948,
+      "grad_norm": 14.577869415283203,
+      "learning_rate": 1.051948051948052e-05,
+      "loss": 0.3731,
+      "step": 150
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.8333333333333334,
+      "eval_combined_score": 0.8527568922305764,
+      "eval_f1": 0.8721804511278195,
+      "eval_loss": 0.3599551320075989,
+      "eval_runtime": 1.6301,
+      "eval_samples_per_second": 250.291,
+      "eval_steps_per_second": 3.067,
+      "step": 154
+    },
+    {
+      "epoch": 2.2727272727272725,
+      "grad_norm": 6.724405288696289,
+      "learning_rate": 7.272727272727273e-06,
+      "loss": 0.2544,
+      "step": 175
+    },
+    {
+      "epoch": 2.5974025974025974,
+      "grad_norm": 5.845699310302734,
+      "learning_rate": 4.025974025974026e-06,
+      "loss": 0.2513,
+      "step": 200
+    },
+    {
+      "epoch": 2.9220779220779223,
+      "grad_norm": 11.254636764526367,
+      "learning_rate": 7.792207792207793e-07,
+      "loss": 0.2426,
+      "step": 225
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.8455882352941176,
+      "eval_combined_score": 0.8658320923306031,
+      "eval_f1": 0.8860759493670886,
+      "eval_loss": 0.3552929759025574,
+      "eval_runtime": 1.6621,
+      "eval_samples_per_second": 245.469,
+      "eval_steps_per_second": 3.008,
+      "step": 231
+    },
+    {
+      "epoch": 3.0,
+      "step": 231,
+      "total_flos": 628599578480640.0,
+      "train_loss": 0.38404927728496074,
+      "train_runtime": 38.3271,
+      "train_samples_per_second": 287.108,
+      "train_steps_per_second": 6.027
+    }
+  ],
+  "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/cls/seed_44/all_results.json

@@ -0,0 +1,17 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8480392156862745,
+    "eval_combined_score": 0.8676559714795009,
+    "eval_f1": 0.8872727272727273,
+    "eval_loss": 0.337443470954895,
+    "eval_runtime": 1.669,
+    "eval_samples": 408,
+    "eval_samples_per_second": 244.452,
+    "eval_steps_per_second": 2.996,
+    "total_flos": 629526768279552.0,
+    "train_loss": 0.3921152353286743,
+    "train_runtime": 37.5948,
+    "train_samples": 3668,
+    "train_samples_per_second": 292.7,
+    "train_steps_per_second": 6.144
+}

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

@@ -0,0 +1,11 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8480392156862745,
+    "eval_combined_score": 0.8676559714795009,
+    "eval_f1": 0.8872727272727273,
+    "eval_loss": 0.337443470954895,
+    "eval_runtime": 1.669,
+    "eval_samples": 408,
+    "eval_samples_per_second": 244.452,
+    "eval_steps_per_second": 2.996
+}

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

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

@@ -0,0 +1,9 @@
+{
+    "epoch": 3.0,
+    "total_flos": 629526768279552.0,
+    "train_loss": 0.3921152353286743,
+    "train_runtime": 37.5948,
+    "train_samples": 3668,
+    "train_samples_per_second": 292.7,
+    "train_steps_per_second": 6.144
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/cls/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": 2.307070732116699,
+      "learning_rate": 2.6753246753246755e-05,
+      "loss": 0.6162,
+      "step": 25
+    },
+    {
+      "epoch": 0.6493506493506493,
+      "grad_norm": 3.4895105361938477,
+      "learning_rate": 2.350649350649351e-05,
+      "loss": 0.566,
+      "step": 50
+    },
+    {
+      "epoch": 0.974025974025974,
+      "grad_norm": 4.919632434844971,
+      "learning_rate": 2.025974025974026e-05,
+      "loss": 0.4645,
+      "step": 75
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.7769607843137255,
+      "eval_combined_score": 0.7961883029073699,
+      "eval_f1": 0.8154158215010142,
+      "eval_loss": 0.4386051893234253,
+      "eval_runtime": 1.8335,
+      "eval_samples_per_second": 222.522,
+      "eval_steps_per_second": 2.727,
+      "step": 77
+    },
+    {
+      "epoch": 1.2987012987012987,
+      "grad_norm": 4.3901472091674805,
+      "learning_rate": 1.7012987012987013e-05,
+      "loss": 0.3849,
+      "step": 100
+    },
+    {
+      "epoch": 1.6233766233766234,
+      "grad_norm": 4.093626022338867,
+      "learning_rate": 1.3766233766233767e-05,
+      "loss": 0.3885,
+      "step": 125
+    },
+    {
+      "epoch": 1.948051948051948,
+      "grad_norm": 5.644167423248291,
+      "learning_rate": 1.051948051948052e-05,
+      "loss": 0.3752,
+      "step": 150
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.8602941176470589,
+      "eval_combined_score": 0.880059185361315,
+      "eval_f1": 0.8998242530755711,
+      "eval_loss": 0.3410514295101166,
+      "eval_runtime": 1.4199,
+      "eval_samples_per_second": 287.344,
+      "eval_steps_per_second": 3.521,
+      "step": 154
+    },
+    {
+      "epoch": 2.2727272727272725,
+      "grad_norm": 7.801192283630371,
+      "learning_rate": 7.272727272727273e-06,
+      "loss": 0.265,
+      "step": 175
+    },
+    {
+      "epoch": 2.5974025974025974,
+      "grad_norm": 6.680178642272949,
+      "learning_rate": 4.025974025974026e-06,
+      "loss": 0.2518,
+      "step": 200
+    },
+    {
+      "epoch": 2.9220779220779223,
+      "grad_norm": 3.9861366748809814,
+      "learning_rate": 7.792207792207793e-07,
+      "loss": 0.2507,
+      "step": 225
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.8480392156862745,
+      "eval_combined_score": 0.8676559714795009,
+      "eval_f1": 0.8872727272727273,
+      "eval_loss": 0.337443470954895,
+      "eval_runtime": 1.4638,
+      "eval_samples_per_second": 278.73,
+      "eval_steps_per_second": 3.416,
+      "step": 231
+    },
+    {
+      "epoch": 3.0,
+      "step": 231,
+      "total_flos": 629526768279552.0,
+      "train_loss": 0.3921152353286743,
+      "train_runtime": 37.5948,
+      "train_samples_per_second": 292.7,
+      "train_steps_per_second": 6.144
+    }
+  ],
+  "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/gated_multi_branch/seed_42/all_results.json

@@ -0,0 +1,17 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8382352941176471,
+    "eval_combined_score": 0.859977862112587,
+    "eval_f1": 0.8817204301075269,
+    "eval_loss": 0.3540915548801422,
+    "eval_runtime": 1.5523,
+    "eval_samples": 408,
+    "eval_samples_per_second": 262.84,
+    "eval_steps_per_second": 3.221,
+    "total_flos": 650826664755696.0,
+    "train_loss": 0.3987341035496105,
+    "train_runtime": 41.3742,
+    "train_samples": 3668,
+    "train_samples_per_second": 265.963,
+    "train_steps_per_second": 5.583
+}

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

@@ -0,0 +1,11 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8382352941176471,
+    "eval_combined_score": 0.859977862112587,
+    "eval_f1": 0.8817204301075269,
+    "eval_loss": 0.3540915548801422,
+    "eval_runtime": 1.5523,
+    "eval_samples": 408,
+    "eval_samples_per_second": 262.84,
+    "eval_steps_per_second": 3.221
+}

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

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

@@ -0,0 +1,9 @@
+{
+    "epoch": 3.0,
+    "total_flos": 650826664755696.0,
+    "train_loss": 0.3987341035496105,
+    "train_runtime": 41.3742,
+    "train_samples": 3668,
+    "train_samples_per_second": 265.963,
+    "train_steps_per_second": 5.583
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/gated_multi_branch/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": 5.213130474090576,
+      "learning_rate": 2.6753246753246755e-05,
+      "loss": 0.6228,
+      "step": 25
+    },
+    {
+      "epoch": 0.6493506493506493,
+      "grad_norm": 5.876645565032959,
+      "learning_rate": 2.350649350649351e-05,
+      "loss": 0.5401,
+      "step": 50
+    },
+    {
+      "epoch": 0.974025974025974,
+      "grad_norm": 3.7962708473205566,
+      "learning_rate": 2.025974025974026e-05,
+      "loss": 0.53,
+      "step": 75
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.7720588235294118,
+      "eval_combined_score": 0.78975611983924,
+      "eval_f1": 0.8074534161490683,
+      "eval_loss": 0.49036601185798645,
+      "eval_runtime": 1.9676,
+      "eval_samples_per_second": 207.358,
+      "eval_steps_per_second": 2.541,
+      "step": 77
+    },
+    {
+      "epoch": 1.2987012987012987,
+      "grad_norm": 4.581447124481201,
+      "learning_rate": 1.7012987012987013e-05,
+      "loss": 0.4199,
+      "step": 100
+    },
+    {
+      "epoch": 1.6233766233766234,
+      "grad_norm": 8.923229217529297,
+      "learning_rate": 1.3766233766233767e-05,
+      "loss": 0.397,
+      "step": 125
+    },
+    {
+      "epoch": 1.948051948051948,
+      "grad_norm": 7.470593452453613,
+      "learning_rate": 1.051948051948052e-05,
+      "loss": 0.3516,
+      "step": 150
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.8504901960784313,
+      "eval_combined_score": 0.8725680686782484,
+      "eval_f1": 0.8946459412780656,
+      "eval_loss": 0.34462273120880127,
+      "eval_runtime": 2.1505,
+      "eval_samples_per_second": 189.723,
+      "eval_steps_per_second": 2.325,
+      "step": 154
+    },
+    {
+      "epoch": 2.2727272727272725,
+      "grad_norm": 4.34853458404541,
+      "learning_rate": 7.272727272727273e-06,
+      "loss": 0.2772,
+      "step": 175
+    },
+    {
+      "epoch": 2.5974025974025974,
+      "grad_norm": 11.729267120361328,
+      "learning_rate": 4.025974025974026e-06,
+      "loss": 0.2533,
+      "step": 200
+    },
+    {
+      "epoch": 2.9220779220779223,
+      "grad_norm": 7.546706676483154,
+      "learning_rate": 7.792207792207793e-07,
+      "loss": 0.2313,
+      "step": 225
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.8382352941176471,
+      "eval_combined_score": 0.859977862112587,
+      "eval_f1": 0.8817204301075269,
+      "eval_loss": 0.3540915548801422,
+      "eval_runtime": 1.6467,
+      "eval_samples_per_second": 247.767,
+      "eval_steps_per_second": 3.036,
+      "step": 231
+    },
+    {
+      "epoch": 3.0,
+      "step": 231,
+      "total_flos": 650826664755696.0,
+      "train_loss": 0.3987341035496105,
+      "train_runtime": 41.3742,
+      "train_samples_per_second": 265.963,
+      "train_steps_per_second": 5.583
+    }
+  ],
+  "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": 650826664755696.0,
+  "train_batch_size": 48,
+  "trial_name": null,
+  "trial_params": null
+}

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

@@ -0,0 +1,17 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8333333333333334,
+    "eval_combined_score": 0.8565959952885749,
+    "eval_f1": 0.8798586572438163,
+    "eval_loss": 0.3669602572917938,
+    "eval_runtime": 1.2886,
+    "eval_samples": 408,
+    "eval_samples_per_second": 316.624,
+    "eval_steps_per_second": 3.88,
+    "total_flos": 648742369041360.0,
+    "train_loss": 0.3874885788211575,
+    "train_runtime": 40.7954,
+    "train_samples": 3668,
+    "train_samples_per_second": 269.736,
+    "train_steps_per_second": 5.662
+}

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

@@ -0,0 +1,11 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.8333333333333334,
+    "eval_combined_score": 0.8565959952885749,
+    "eval_f1": 0.8798586572438163,
+    "eval_loss": 0.3669602572917938,
+    "eval_runtime": 1.2886,
+    "eval_samples": 408,
+    "eval_samples_per_second": 316.624,
+    "eval_steps_per_second": 3.88
+}

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

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

@@ -0,0 +1,9 @@
+{
+    "epoch": 3.0,
+    "total_flos": 648742369041360.0,
+    "train_loss": 0.3874885788211575,
+    "train_runtime": 40.7954,
+    "train_samples": 3668,
+    "train_samples_per_second": 269.736,
+    "train_steps_per_second": 5.662
+}

mbert_rtx6000_metrics/result_ablation_mbert/mrpc/mBERT/gated_multi_branch/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": 5.391201496124268,
+      "learning_rate": 2.6753246753246755e-05,
+      "loss": 0.6214,
+      "step": 25
+    },
+    {
+      "epoch": 0.6493506493506493,
+      "grad_norm": 2.0320703983306885,
+      "learning_rate": 2.350649350649351e-05,
+      "loss": 0.5583,
+      "step": 50
+    },
+    {
+      "epoch": 0.974025974025974,
+      "grad_norm": 4.242596626281738,
+      "learning_rate": 2.025974025974026e-05,
+      "loss": 0.494,
+      "step": 75
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.7990196078431373,
+      "eval_combined_score": 0.8327345596218945,
+      "eval_f1": 0.8664495114006515,
+      "eval_loss": 0.4170113503932953,
+      "eval_runtime": 2.0181,
+      "eval_samples_per_second": 202.168,
+      "eval_steps_per_second": 2.478,
+      "step": 77
+    },
+    {
+      "epoch": 1.2987012987012987,
+      "grad_norm": 7.784523010253906,
+      "learning_rate": 1.7012987012987013e-05,
+      "loss": 0.397,
+      "step": 100
+    },
+    {
+      "epoch": 1.6233766233766234,
+      "grad_norm": 4.2649006843566895,
+      "learning_rate": 1.3766233766233767e-05,
+      "loss": 0.3888,
+      "step": 125
+    },
+    {
+      "epoch": 1.948051948051948,
+      "grad_norm": 8.975968360900879,
+      "learning_rate": 1.051948051948052e-05,
+      "loss": 0.3506,
+      "step": 150
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.8284313725490197,
+      "eval_combined_score": 0.8505793226381462,
+      "eval_f1": 0.8727272727272727,
+      "eval_loss": 0.35200247168540955,
+      "eval_runtime": 1.838,
+      "eval_samples_per_second": 221.978,
+      "eval_steps_per_second": 2.72,
+      "step": 154
+    },
+    {
+      "epoch": 2.2727272727272725,
+      "grad_norm": 6.934613227844238,
+      "learning_rate": 7.272727272727273e-06,
+      "loss": 0.2421,
+      "step": 175
+    },
+    {
+      "epoch": 2.5974025974025974,
+      "grad_norm": 7.820345401763916,
+      "learning_rate": 4.025974025974026e-06,
+      "loss": 0.24,
+      "step": 200
+    },
+    {
+      "epoch": 2.9220779220779223,
+      "grad_norm": 12.318483352661133,
+      "learning_rate": 7.792207792207793e-07,
+      "loss": 0.2461,
+      "step": 225
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.8333333333333334,
+      "eval_combined_score": 0.8565959952885749,
+      "eval_f1": 0.8798586572438163,
+      "eval_loss": 0.3669602572917938,
+      "eval_runtime": 1.3078,
+      "eval_samples_per_second": 311.984,
+      "eval_steps_per_second": 3.823,
+      "step": 231
+    },
+    {
+      "epoch": 3.0,
+      "step": 231,
+      "total_flos": 648742369041360.0,
+      "train_loss": 0.3874885788211575,
+      "train_runtime": 40.7954,
+      "train_samples_per_second": 269.736,
+      "train_steps_per_second": 5.662
+    }
+  ],
+  "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": 648742369041360.0,
+  "train_batch_size": 48,
+  "trial_name": null,
+  "trial_params": null
+}