Training in progress, step 100

all_results.json

@@ -1,20 +1,20 @@
 {
     "epoch": 20.0,
-    "eval_accuracy": 81.15942028985508,
-    "eval_average_metrics": 81.15942028985508,
-    "eval_loss": 0.18789426982402802,
-    "eval_runtime": 0.6932,
+    "eval_accuracy": 84.05797101449275,
+    "eval_average_metrics": 84.05797101449275,
+    "eval_loss": 0.16963252425193787,
+    "eval_runtime": 0.8933,
     "eval_samples": 277,
-    "eval_samples_per_second": 199.08,
+    "eval_samples_per_second": 154.483,
     "eval_steps_per_second": 6.498,
-    "test_accuracy": 69.7841726618705,
-    "test_average_metrics": 69.7841726618705,
-    "test_loss": 0.23357094824314117,
-    "test_runtime": 0.4076,
-    "test_samples_per_second": 340.999,
-    "train_loss": 0.7619044189453125,
-    "train_runtime": 235.5565,
+    "test_accuracy": 73.38129496402878,
+    "test_average_metrics": 73.38129496402878,
+    "test_loss": 0.26427963376045227,
+    "test_runtime": 1.0271,
+    "test_samples_per_second": 135.33,
+    "train_loss": 0.1103338890350782,
+    "train_runtime": 529.9082,
     "train_samples": 2490,
-    "train_samples_per_second": 211.414,
-    "train_steps_per_second": 2.123
+    "train_samples_per_second": 93.979,
+    "train_steps_per_second": 2.944
 }

eval_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 20.0,
-    "eval_accuracy": 81.15942028985508,
-    "eval_average_metrics": 81.15942028985508,
-    "eval_loss": 0.18789426982402802,
-    "eval_runtime": 0.6932,
-    "eval_samples_per_second": 199.08
+    "eval_accuracy": 84.05797101449275,
+    "eval_average_metrics": 84.05797101449275,
+    "eval_loss": 0.16963252425193787,
+    "eval_runtime": 0.8933,
+    "eval_samples_per_second": 154.483
 }

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6451f5be2a2ca02ed37ae107a73a31c2c5ab273f08121393bc2d6acb53358c45
-size 7551621
+oid sha256:8a56d60abfbe8d6f12c35d6dc3c1f7e5f19ac92e88a47f2723a5b466546f1d6b
+size 2631685

runs/Feb01_01-06-20_node2/events.out.tfevents.1643648862.node2

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0ad109047ea799567cd3ccece208f937e9aef8abe8051fcb75b9ff09afd8230d
-size 9270
+oid sha256:0bb8a929b8cf2edc4928bb39f304c212a5e04e32a8a20db6de8f5598ea8043a8
+size 9946

runs/Feb01_01-06-20_node2/events.out.tfevents.1643649393.node2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba022c8eb07b0a85f141120d86cb3296f9308a58f5a4eec65c1fba63daa24b97
+size 684

runs/Feb02_15-36-39_node1/1643787523.6723077/events.out.tfevents.1643787523.node1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8efe16d778edae593e5250ce2717574822bdc665a58e824e5f5ab7e304be61db
+size 5008

runs/Feb02_15-36-39_node1/events.out.tfevents.1643787523.node1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c01199f169c7577076d1beb0685ddd4afe9c93fa1bdd4e81f459fe2d5ba555c
+size 4297

test_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 20.0,
-    "test_accuracy": 69.7841726618705,
-    "test_average_metrics": 69.7841726618705,
-    "test_loss": 0.23357094824314117,
-    "test_runtime": 0.4076,
-    "test_samples_per_second": 340.999
+    "test_accuracy": 73.38129496402878,
+    "test_average_metrics": 73.38129496402878,
+    "test_loss": 0.26427963376045227,
+    "test_runtime": 1.0271,
+    "test_samples_per_second": 135.33
 }

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 20.0,
-    "train_loss": 0.7619044189453125,
-    "train_runtime": 235.5565,
+    "train_loss": 0.1103338890350782,
+    "train_runtime": 529.9082,
     "train_samples": 2490,
-    "train_samples_per_second": 211.414,
-    "train_steps_per_second": 2.123
+    "train_samples_per_second": 93.979,
+    "train_steps_per_second": 2.944
 }

trainer_state.json

@@ -1,76 +1,178 @@
 {
-  "best_metric": 81.15942028985508,
-  "best_model_checkpoint": "outputs/bitfit/t5-base/rte/checkpoint-200",
+  "best_metric": 84.05797101449275,
+  "best_model_checkpoint": "outputs/bitfit/t5-base/rte/checkpoint-300",
   "epoch": 20.0,
-  "global_step": 500,
+  "global_step": 1560,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 4.0,
-      "eval_accuracy": 55.79710144927537,
-      "eval_average_metrics": 55.79710144927537,
-      "eval_loss": 0.2813531160354614,
-      "eval_runtime": 0.6184,
-      "eval_samples_per_second": 223.144,
+      "epoch": 1.28,
+      "eval_accuracy": 76.81159420289855,
+      "eval_average_metrics": 76.81159420289855,
+      "eval_loss": 0.21982769668102264,
+      "eval_runtime": 1.0149,
+      "eval_samples_per_second": 135.972,
       "step": 100
     },
     {
-      "epoch": 8.0,
-      "eval_accuracy": 81.15942028985508,
-      "eval_average_metrics": 81.15942028985508,
-      "eval_loss": 0.18789426982402802,
-      "eval_runtime": 0.5049,
-      "eval_samples_per_second": 273.329,
+      "epoch": 2.56,
+      "eval_accuracy": 78.26086956521739,
+      "eval_average_metrics": 78.26086956521739,
+      "eval_loss": 0.18932242691516876,
+      "eval_runtime": 0.6715,
+      "eval_samples_per_second": 205.52,
       "step": 200
     },
     {
-      "epoch": 12.0,
-      "eval_accuracy": 76.81159420289855,
-      "eval_average_metrics": 76.81159420289855,
-      "eval_loss": 0.18861950933933258,
-      "eval_runtime": 0.4727,
-      "eval_samples_per_second": 291.912,
+      "epoch": 3.85,
+      "eval_accuracy": 84.05797101449275,
+      "eval_average_metrics": 84.05797101449275,
+      "eval_loss": 0.16963252425193787,
+      "eval_runtime": 0.8653,
+      "eval_samples_per_second": 159.484,
       "step": 300
     },
     {
-      "epoch": 16.0,
-      "eval_accuracy": 78.26086956521739,
-      "eval_average_metrics": 78.26086956521739,
-      "eval_loss": 0.24881036579608917,
-      "eval_runtime": 0.474,
-      "eval_samples_per_second": 291.12,
+      "epoch": 5.13,
+      "eval_accuracy": 80.43478260869566,
+      "eval_average_metrics": 80.43478260869566,
+      "eval_loss": 0.18479692935943604,
+      "eval_runtime": 0.7171,
+      "eval_samples_per_second": 192.431,
       "step": 400
     },
     {
-      "epoch": 20.0,
-      "learning_rate": 0.0,
-      "loss": 0.7619,
+      "epoch": 6.41,
+      "learning_rate": 0.00020384615384615385,
+      "loss": 0.2327,
       "step": 500
     },
     {
-      "epoch": 20.0,
-      "eval_accuracy": 76.08695652173914,
-      "eval_average_metrics": 76.08695652173914,
-      "eval_loss": 0.3080892562866211,
-      "eval_runtime": 0.5158,
-      "eval_samples_per_second": 267.545,
+      "epoch": 6.41,
+      "eval_accuracy": 81.88405797101449,
+      "eval_average_metrics": 81.88405797101449,
+      "eval_loss": 0.23633617162704468,
+      "eval_runtime": 0.6787,
+      "eval_samples_per_second": 203.343,
       "step": 500
     },
+    {
+      "epoch": 7.69,
+      "eval_accuracy": 79.71014492753623,
+      "eval_average_metrics": 79.71014492753623,
+      "eval_loss": 0.23751917481422424,
+      "eval_runtime": 0.8803,
+      "eval_samples_per_second": 156.767,
+      "step": 600
+    },
+    {
+      "epoch": 8.97,
+      "eval_accuracy": 84.05797101449275,
+      "eval_average_metrics": 84.05797101449275,
+      "eval_loss": 0.23690848052501678,
+      "eval_runtime": 0.5629,
+      "eval_samples_per_second": 245.142,
+      "step": 700
+    },
+    {
+      "epoch": 10.26,
+      "eval_accuracy": 83.33333333333334,
+      "eval_average_metrics": 83.33333333333334,
+      "eval_loss": 0.2846378684043884,
+      "eval_runtime": 0.9472,
+      "eval_samples_per_second": 145.689,
+      "step": 800
+    },
+    {
+      "epoch": 11.54,
+      "eval_accuracy": 82.6086956521739,
+      "eval_average_metrics": 82.6086956521739,
+      "eval_loss": 0.2770601212978363,
+      "eval_runtime": 0.8799,
+      "eval_samples_per_second": 156.839,
+      "step": 900
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 0.00010769230769230768,
+      "loss": 0.0774,
+      "step": 1000
+    },
+    {
+      "epoch": 12.82,
+      "eval_accuracy": 82.6086956521739,
+      "eval_average_metrics": 82.6086956521739,
+      "eval_loss": 0.3579100966453552,
+      "eval_runtime": 0.9343,
+      "eval_samples_per_second": 147.71,
+      "step": 1000
+    },
+    {
+      "epoch": 14.1,
+      "eval_accuracy": 81.15942028985508,
+      "eval_average_metrics": 81.15942028985508,
+      "eval_loss": 0.4031548798084259,
+      "eval_runtime": 0.6763,
+      "eval_samples_per_second": 204.046,
+      "step": 1100
+    },
+    {
+      "epoch": 15.38,
+      "eval_accuracy": 81.15942028985508,
+      "eval_average_metrics": 81.15942028985508,
+      "eval_loss": 0.4019993841648102,
+      "eval_runtime": 0.9288,
+      "eval_samples_per_second": 148.583,
+      "step": 1200
+    },
+    {
+      "epoch": 16.67,
+      "eval_accuracy": 78.98550724637681,
+      "eval_average_metrics": 78.98550724637681,
+      "eval_loss": 0.47033587098121643,
+      "eval_runtime": 0.6015,
+      "eval_samples_per_second": 229.417,
+      "step": 1300
+    },
+    {
+      "epoch": 17.95,
+      "eval_accuracy": 81.88405797101449,
+      "eval_average_metrics": 81.88405797101449,
+      "eval_loss": 0.4693973958492279,
+      "eval_runtime": 0.8377,
+      "eval_samples_per_second": 164.743,
+      "step": 1400
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.1538461538461538e-05,
+      "loss": 0.0312,
+      "step": 1500
+    },
+    {
+      "epoch": 19.23,
+      "eval_accuracy": 79.71014492753623,
+      "eval_average_metrics": 79.71014492753623,
+      "eval_loss": 0.5053138136863708,
+      "eval_runtime": 1.053,
+      "eval_samples_per_second": 131.051,
+      "step": 1500
+    },
     {
       "epoch": 20.0,
-      "step": 500,
-      "total_flos": 7650655646515200.0,
-      "train_loss": 0.7619044189453125,
-      "train_runtime": 235.5565,
-      "train_samples_per_second": 211.414,
-      "train_steps_per_second": 2.123
+      "step": 1560,
+      "total_flos": 7649465032433664.0,
+      "train_loss": 0.1103338890350782,
+      "train_runtime": 529.9082,
+      "train_samples_per_second": 93.979,
+      "train_steps_per_second": 2.944
     }
   ],
-  "max_steps": 500,
+  "max_steps": 1560,
   "num_train_epochs": 20,
-  "total_flos": 7650655646515200.0,
+  "total_flos": 7649465032433664.0,
   "trial_name": null,
   "trial_params": null
 }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:2df0c1c1741f8ab42c8bcb5b0a2436f5275ba72a6f32b8533f2f0d10c2728a9f
+oid sha256:131be287b992cce8aa641373128179469bd2f96cf2d311ace3ba167fd100d6d1
 size 3183

training_config.json

@@ -1 +1 @@
-{"bottleneck_dim": 24, "dataset_config_name": ["en"], "delta_type": "adapter", "do_eval": true, "do_test": true, "do_train": true, "eval_dataset_config_name": ["en"], "eval_dataset_name": "rte", "eval_steps": 100, "evaluation_strategy": "steps", "greater_is_better": true, "learning_rate": 0.0003, "load_best_model_at_end": true, "max_source_length": 128, "metric_for_best_model": "average_metrics", "model_name_or_path": "../../../../plm_cache/t5-base", "num_train_epochs": 20, "output_dir": "outputs/bitfit/t5-base/rte", "overwrite_output_dir": true, "per_device_eval_batch_size": 32, "per_device_train_batch_size": 32, "predict_with_generate": true, "push_to_hub": true, "save_steps": 100, "save_strategy": "steps", "save_total_limit": 1, "seed": 42, "split_validation_test": true, "task_name": "rte", "test_dataset_config_name": ["en"], "test_dataset_name": "rte", "tokenizer_name": "../../../../plm_cache/t5-base", "unfrozen_modules": ["deltas", "layer_norm", "final_layer_norm"], "warmup_steps": 0}
+{"dataset_config_name": ["en"], "delta_type": "lora", "do_eval": true, "do_test": true, "do_train": true, "eval_dataset_config_name": ["en"], "eval_dataset_name": "rte", "eval_steps": 100, "evaluation_strategy": "steps", "greater_is_better": true, "learning_rate": 0.0003, "load_best_model_at_end": true, "lora_r": 8, "max_source_length": 128, "metric_for_best_model": "average_metrics", "model_name_or_path": "../../../../plm_cache/t5-base", "num_train_epochs": 20, "output_dir": "outputs/bitfit/t5-base/rte", "overwrite_output_dir": true, "per_device_eval_batch_size": 32, "per_device_train_batch_size": 32, "predict_with_generate": true, "push_to_hub": true, "save_steps": 100, "save_strategy": "steps", "save_total_limit": 1, "seed": 42, "split_validation_test": true, "task_name": "rte", "test_dataset_config_name": ["en"], "test_dataset_name": "rte", "tokenizer_name": "../../../../plm_cache/t5-base", "unfrozen_modules": ["deltas", "layer_norm", "final_layer_norm"], "warmup_steps": 0}