Upload batch 23/24: vsfc/mBERT/mean/seed_43 ... vsfc/mBERT/mha_attention/seed_43

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

@@ -0,0 +1,15 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.936197094125079,
+    "eval_loss": 0.22665680944919586,
+    "eval_runtime": 6.7728,
+    "eval_samples": 1583,
+    "eval_samples_per_second": 233.73,
+    "eval_steps_per_second": 2.51,
+    "total_flos": 1024705701150720.0,
+    "train_loss": 0.20191558268612234,
+    "train_runtime": 106.7947,
+    "train_samples": 11426,
+    "train_samples_per_second": 320.971,
+    "train_steps_per_second": 6.714
+}

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

@@ -0,0 +1,9 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.936197094125079,
+    "eval_loss": 0.22665680944919586,
+    "eval_runtime": 6.7728,
+    "eval_samples": 1583,
+    "eval_samples_per_second": 233.73,
+    "eval_steps_per_second": 2.51
+}

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

@@ -0,0 +1,9 @@
+{
+    "epoch": 3.0,
+    "total_flos": 1024705701150720.0,
+    "train_loss": 0.20191558268612234,
+    "train_runtime": 106.7947,
+    "train_samples": 11426,
+    "train_samples_per_second": 320.971,
+    "train_steps_per_second": 6.714
+}

mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/mean/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": 5.314708709716797,
+      "learning_rate": 2.8953974895397488e-05,
+      "loss": 0.5169,
+      "step": 25
+    },
+    {
+      "epoch": 0.20920502092050208,
+      "grad_norm": 8.301779747009277,
+      "learning_rate": 2.7907949790794982e-05,
+      "loss": 0.3416,
+      "step": 50
+    },
+    {
+      "epoch": 0.3138075313807531,
+      "grad_norm": 5.364657402038574,
+      "learning_rate": 2.686192468619247e-05,
+      "loss": 0.3466,
+      "step": 75
+    },
+    {
+      "epoch": 0.41841004184100417,
+      "grad_norm": 3.2444639205932617,
+      "learning_rate": 2.5815899581589957e-05,
+      "loss": 0.2759,
+      "step": 100
+    },
+    {
+      "epoch": 0.5230125523012552,
+      "grad_norm": 6.162262439727783,
+      "learning_rate": 2.4769874476987448e-05,
+      "loss": 0.2799,
+      "step": 125
+    },
+    {
+      "epoch": 0.6276150627615062,
+      "grad_norm": 4.790009498596191,
+      "learning_rate": 2.372384937238494e-05,
+      "loss": 0.2475,
+      "step": 150
+    },
+    {
+      "epoch": 0.7322175732217573,
+      "grad_norm": 7.008646488189697,
+      "learning_rate": 2.2677824267782426e-05,
+      "loss": 0.2744,
+      "step": 175
+    },
+    {
+      "epoch": 0.8368200836820083,
+      "grad_norm": 3.773528575897217,
+      "learning_rate": 2.1631799163179917e-05,
+      "loss": 0.225,
+      "step": 200
+    },
+    {
+      "epoch": 0.9414225941422594,
+      "grad_norm": 6.053400993347168,
+      "learning_rate": 2.0585774058577408e-05,
+      "loss": 0.2334,
+      "step": 225
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.9286165508528111,
+      "eval_loss": 0.22938770055770874,
+      "eval_runtime": 6.8869,
+      "eval_samples_per_second": 229.855,
+      "eval_steps_per_second": 2.468,
+      "step": 239
+    },
+    {
+      "epoch": 1.0460251046025104,
+      "grad_norm": 5.818469047546387,
+      "learning_rate": 1.9539748953974895e-05,
+      "loss": 0.203,
+      "step": 250
+    },
+    {
+      "epoch": 1.1506276150627615,
+      "grad_norm": 8.090319633483887,
+      "learning_rate": 1.8493723849372386e-05,
+      "loss": 0.1994,
+      "step": 275
+    },
+    {
+      "epoch": 1.2552301255230125,
+      "grad_norm": 3.0878026485443115,
+      "learning_rate": 1.7447698744769877e-05,
+      "loss": 0.1816,
+      "step": 300
+    },
+    {
+      "epoch": 1.3598326359832635,
+      "grad_norm": 3.3793857097625732,
+      "learning_rate": 1.6401673640167365e-05,
+      "loss": 0.1842,
+      "step": 325
+    },
+    {
+      "epoch": 1.4644351464435146,
+      "grad_norm": 3.8809702396392822,
+      "learning_rate": 1.5355648535564852e-05,
+      "loss": 0.1899,
+      "step": 350
+    },
+    {
+      "epoch": 1.5690376569037658,
+      "grad_norm": 6.736715316772461,
+      "learning_rate": 1.4309623430962344e-05,
+      "loss": 0.1819,
+      "step": 375
+    },
+    {
+      "epoch": 1.6736401673640167,
+      "grad_norm": 4.629181861877441,
+      "learning_rate": 1.3263598326359832e-05,
+      "loss": 0.1905,
+      "step": 400
+    },
+    {
+      "epoch": 1.778242677824268,
+      "grad_norm": 3.5737199783325195,
+      "learning_rate": 1.2217573221757323e-05,
+      "loss": 0.1988,
+      "step": 425
+    },
+    {
+      "epoch": 1.8828451882845187,
+      "grad_norm": 6.036787509918213,
+      "learning_rate": 1.1171548117154812e-05,
+      "loss": 0.1937,
+      "step": 450
+    },
+    {
+      "epoch": 1.98744769874477,
+      "grad_norm": 2.8429791927337646,
+      "learning_rate": 1.0125523012552301e-05,
+      "loss": 0.1477,
+      "step": 475
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.9317751105495894,
+      "eval_loss": 0.22606870532035828,
+      "eval_runtime": 6.4536,
+      "eval_samples_per_second": 245.291,
+      "eval_steps_per_second": 2.634,
+      "step": 478
+    },
+    {
+      "epoch": 2.092050209205021,
+      "grad_norm": 13.506583213806152,
+      "learning_rate": 9.079497907949792e-06,
+      "loss": 0.1236,
+      "step": 500
+    },
+    {
+      "epoch": 2.196652719665272,
+      "grad_norm": 4.173339366912842,
+      "learning_rate": 8.03347280334728e-06,
+      "loss": 0.1272,
+      "step": 525
+    },
+    {
+      "epoch": 2.301255230125523,
+      "grad_norm": 2.1523401737213135,
+      "learning_rate": 6.98744769874477e-06,
+      "loss": 0.1219,
+      "step": 550
+    },
+    {
+      "epoch": 2.405857740585774,
+      "grad_norm": 4.702078819274902,
+      "learning_rate": 5.941422594142259e-06,
+      "loss": 0.1243,
+      "step": 575
+    },
+    {
+      "epoch": 2.510460251046025,
+      "grad_norm": 2.8082313537597656,
+      "learning_rate": 4.895397489539749e-06,
+      "loss": 0.1243,
+      "step": 600
+    },
+    {
+      "epoch": 2.6150627615062763,
+      "grad_norm": 6.025881767272949,
+      "learning_rate": 3.849372384937238e-06,
+      "loss": 0.1141,
+      "step": 625
+    },
+    {
+      "epoch": 2.719665271966527,
+      "grad_norm": 5.893019199371338,
+      "learning_rate": 2.803347280334728e-06,
+      "loss": 0.1389,
+      "step": 650
+    },
+    {
+      "epoch": 2.8242677824267783,
+      "grad_norm": 5.248950481414795,
+      "learning_rate": 1.7573221757322176e-06,
+      "loss": 0.0995,
+      "step": 675
+    },
+    {
+      "epoch": 2.928870292887029,
+      "grad_norm": 3.4337127208709717,
+      "learning_rate": 7.112970711297071e-07,
+      "loss": 0.1231,
+      "step": 700
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.936197094125079,
+      "eval_loss": 0.22665680944919586,
+      "eval_runtime": 5.155,
+      "eval_samples_per_second": 307.079,
+      "eval_steps_per_second": 3.298,
+      "step": 717
+    },
+    {
+      "epoch": 3.0,
+      "step": 717,
+      "total_flos": 1024705701150720.0,
+      "train_loss": 0.20191558268612234,
+      "train_runtime": 106.7947,
+      "train_samples_per_second": 320.971,
+      "train_steps_per_second": 6.714
+    }
+  ],
+  "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/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 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/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/vsfc/mBERT/mha_attention/seed_43/all_results.json

@@ -0,0 +1,15 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.9387239418825016,
+    "eval_loss": 0.2242736667394638,
+    "eval_runtime": 6.8417,
+    "eval_samples": 1583,
+    "eval_samples_per_second": 231.376,
+    "eval_steps_per_second": 2.485,
+    "total_flos": 1027433247582168.0,
+    "train_loss": 0.20716787132757977,
+    "train_runtime": 96.0564,
+    "train_samples": 11426,
+    "train_samples_per_second": 356.853,
+    "train_steps_per_second": 7.464
+}

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

@@ -0,0 +1,9 @@
+{
+    "epoch": 3.0,
+    "eval_accuracy": 0.9387239418825016,
+    "eval_loss": 0.2242736667394638,
+    "eval_runtime": 6.8417,
+    "eval_samples": 1583,
+    "eval_samples_per_second": 231.376,
+    "eval_steps_per_second": 2.485
+}

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

@@ -0,0 +1,9 @@
+{
+    "epoch": 3.0,
+    "total_flos": 1027433247582168.0,
+    "train_loss": 0.20716787132757977,
+    "train_runtime": 96.0564,
+    "train_samples": 11426,
+    "train_samples_per_second": 356.853,
+    "train_steps_per_second": 7.464
+}

mbert_rtx6000_metrics/result_ablation_mbert/vsfc/mBERT/mha_attention/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": 6.308683395385742,
+      "learning_rate": 2.8953974895397488e-05,
+      "loss": 0.5152,
+      "step": 25
+    },
+    {
+      "epoch": 0.20920502092050208,
+      "grad_norm": 7.704500198364258,
+      "learning_rate": 2.7907949790794982e-05,
+      "loss": 0.3189,
+      "step": 50
+    },
+    {
+      "epoch": 0.3138075313807531,
+      "grad_norm": 6.910772323608398,
+      "learning_rate": 2.686192468619247e-05,
+      "loss": 0.3111,
+      "step": 75
+    },
+    {
+      "epoch": 0.41841004184100417,
+      "grad_norm": 4.743614673614502,
+      "learning_rate": 2.5815899581589957e-05,
+      "loss": 0.2751,
+      "step": 100
+    },
+    {
+      "epoch": 0.5230125523012552,
+      "grad_norm": 10.062735557556152,
+      "learning_rate": 2.4769874476987448e-05,
+      "loss": 0.3335,
+      "step": 125
+    },
+    {
+      "epoch": 0.6276150627615062,
+      "grad_norm": 3.8704612255096436,
+      "learning_rate": 2.372384937238494e-05,
+      "loss": 0.3015,
+      "step": 150
+    },
+    {
+      "epoch": 0.7322175732217573,
+      "grad_norm": 3.3860716819763184,
+      "learning_rate": 2.2677824267782426e-05,
+      "loss": 0.2698,
+      "step": 175
+    },
+    {
+      "epoch": 0.8368200836820083,
+      "grad_norm": 6.553400993347168,
+      "learning_rate": 2.1631799163179917e-05,
+      "loss": 0.2518,
+      "step": 200
+    },
+    {
+      "epoch": 0.9414225941422594,
+      "grad_norm": 3.3159918785095215,
+      "learning_rate": 2.0585774058577408e-05,
+      "loss": 0.2436,
+      "step": 225
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.9260897030953885,
+      "eval_loss": 0.23272140324115753,
+      "eval_runtime": 6.213,
+      "eval_samples_per_second": 254.79,
+      "eval_steps_per_second": 2.736,
+      "step": 239
+    },
+    {
+      "epoch": 1.0460251046025104,
+      "grad_norm": 5.275418281555176,
+      "learning_rate": 1.9539748953974895e-05,
+      "loss": 0.2365,
+      "step": 250
+    },
+    {
+      "epoch": 1.1506276150627615,
+      "grad_norm": 5.208725452423096,
+      "learning_rate": 1.8493723849372386e-05,
+      "loss": 0.1869,
+      "step": 275
+    },
+    {
+      "epoch": 1.2552301255230125,
+      "grad_norm": 9.752256393432617,
+      "learning_rate": 1.7447698744769877e-05,
+      "loss": 0.1835,
+      "step": 300
+    },
+    {
+      "epoch": 1.3598326359832635,
+      "grad_norm": 11.681489944458008,
+      "learning_rate": 1.6401673640167365e-05,
+      "loss": 0.1927,
+      "step": 325
+    },
+    {
+      "epoch": 1.4644351464435146,
+      "grad_norm": 5.413773059844971,
+      "learning_rate": 1.5355648535564852e-05,
+      "loss": 0.1698,
+      "step": 350
+    },
+    {
+      "epoch": 1.5690376569037658,
+      "grad_norm": 4.785913944244385,
+      "learning_rate": 1.4309623430962344e-05,
+      "loss": 0.196,
+      "step": 375
+    },
+    {
+      "epoch": 1.6736401673640167,
+      "grad_norm": 4.216114044189453,
+      "learning_rate": 1.3263598326359832e-05,
+      "loss": 0.1911,
+      "step": 400
+    },
+    {
+      "epoch": 1.778242677824268,
+      "grad_norm": 2.994622230529785,
+      "learning_rate": 1.2217573221757323e-05,
+      "loss": 0.1976,
+      "step": 425
+    },
+    {
+      "epoch": 1.8828451882845187,
+      "grad_norm": 4.442307949066162,
+      "learning_rate": 1.1171548117154812e-05,
+      "loss": 0.1704,
+      "step": 450
+    },
+    {
+      "epoch": 1.98744769874477,
+      "grad_norm": 6.980963230133057,
+      "learning_rate": 1.0125523012552301e-05,
+      "loss": 0.1704,
+      "step": 475
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.934301958307012,
+      "eval_loss": 0.21168287098407745,
+      "eval_runtime": 6.9147,
+      "eval_samples_per_second": 228.934,
+      "eval_steps_per_second": 2.459,
+      "step": 478
+    },
+    {
+      "epoch": 2.092050209205021,
+      "grad_norm": 6.22763204574585,
+      "learning_rate": 9.079497907949792e-06,
+      "loss": 0.1191,
+      "step": 500
+    },
+    {
+      "epoch": 2.196652719665272,
+      "grad_norm": 6.816147327423096,
+      "learning_rate": 8.03347280334728e-06,
+      "loss": 0.1273,
+      "step": 525
+    },
+    {
+      "epoch": 2.301255230125523,
+      "grad_norm": 0.2829854488372803,
+      "learning_rate": 6.98744769874477e-06,
+      "loss": 0.1263,
+      "step": 550
+    },
+    {
+      "epoch": 2.405857740585774,
+      "grad_norm": 6.346888065338135,
+      "learning_rate": 5.941422594142259e-06,
+      "loss": 0.1425,
+      "step": 575
+    },
+    {
+      "epoch": 2.510460251046025,
+      "grad_norm": 3.365602731704712,
+      "learning_rate": 4.895397489539749e-06,
+      "loss": 0.145,
+      "step": 600
+    },
+    {
+      "epoch": 2.6150627615062763,
+      "grad_norm": 3.679931640625,
+      "learning_rate": 3.849372384937238e-06,
+      "loss": 0.109,
+      "step": 625
+    },
+    {
+      "epoch": 2.719665271966527,
+      "grad_norm": 0.2995340824127197,
+      "learning_rate": 2.803347280334728e-06,
+      "loss": 0.1256,
+      "step": 650
+    },
+    {
+      "epoch": 2.8242677824267783,
+      "grad_norm": 4.675779342651367,
+      "learning_rate": 1.7573221757322176e-06,
+      "loss": 0.1201,
+      "step": 675
+    },
+    {
+      "epoch": 2.928870292887029,
+      "grad_norm": 3.4704439640045166,
+      "learning_rate": 7.112970711297071e-07,
+      "loss": 0.1452,
+      "step": 700
+    },
+    {
+      "epoch": 3.0,
+      "eval_accuracy": 0.9387239418825016,
+      "eval_loss": 0.2242736667394638,
+      "eval_runtime": 5.3531,
+      "eval_samples_per_second": 295.716,
+      "eval_steps_per_second": 3.176,
+      "step": 717
+    },
+    {
+      "epoch": 3.0,
+      "step": 717,
+      "total_flos": 1027433247582168.0,
+      "train_loss": 0.20716787132757977,
+      "train_runtime": 96.0564,
+      "train_samples_per_second": 356.853,
+      "train_steps_per_second": 7.464
+    }
+  ],
+  "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": 1027433247582168.0,
+  "train_batch_size": 48,
+  "trial_name": null,
+  "trial_params": null
+}