🍻 cheers

README.md

@@ -1,7 +1,11 @@
 ---
+language:
+- eng
 license: apache-2.0
 base_model: facebook/dinov2-large
 tags:
+- multilabel-image-classification
+- multilabel
 - generated_from_trainer
 metrics:
 - accuracy
@@ -15,13 +19,13 @@ should probably proofread and complete it, then remove this comment. -->
 
 # dinov2-large-2024_01_05-kornia_img-size518_batch-size32_epochs70_freeze
 
-This model is a fine-tuned version of [facebook/dinov2-large](https://huggingface.co/facebook/dinov2-large) on the None dataset.
+This model is a fine-tuned version of [facebook/dinov2-large](https://huggingface.co/facebook/dinov2-large) on the multilabel_complete_dataset dataset.
 It achieves the following results on the evaluation set:
-- Loss: 0.0819
-- F1 Micro: 0.8564
-- F1 Macro: 0.7560
-- Roc Auc: 0.9061
-- Accuracy: 0.5656
+- Loss: 0.0825
+- F1 Micro: 0.8570
+- F1 Macro: 0.7430
+- Roc Auc: 0.9080
+- Accuracy: 0.5739
 - Learning Rate: 0.0000
 
 ## Model description

all_results.json

@@ -1,16 +1,16 @@
 {
-    "epoch": 70.0,
-    "eval_accuracy": 0.5605742296918768,
-    "eval_f1_macro": 0.7342630546801885,
-    "eval_f1_micro": 0.8543162417321499,
-    "eval_loss": 0.08401281386613846,
-    "eval_roc_auc": 0.9076857807628663,
-    "eval_runtime": 670.4543,
-    "eval_samples_per_second": 4.26,
-    "eval_steps_per_second": 0.134,
-    "learning_rate": 0.0001,
-    "train_loss": 0.11672632308896316,
-    "train_runtime": 200748.2354,
-    "train_samples_per_second": 3.057,
-    "train_steps_per_second": 0.096
+    "epoch": 114.0,
+    "eval_accuracy": 0.5738795518207283,
+    "eval_f1_macro": 0.7429818572746157,
+    "eval_f1_micro": 0.8570132153593103,
+    "eval_loss": 0.08252906054258347,
+    "eval_roc_auc": 0.9080498428032193,
+    "eval_runtime": 681.2636,
+    "eval_samples_per_second": 4.192,
+    "eval_steps_per_second": 0.132,
+    "learning_rate": 1.0000000000000002e-07,
+    "train_loss": 0.033337813025782426,
+    "train_runtime": 131818.0427,
+    "train_samples_per_second": 9.311,
+    "train_steps_per_second": 0.291
 }

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "facebook/dinov2-large",
+  "_name_or_path": "facebook/dinov2-large2024_01_08",
   "apply_layernorm": true,
   "architectures": [
     "NewheadDinov2ForImageClassification"

eval_results.json

@@ -1,12 +1,12 @@
 {
-    "epoch": 70.0,
-    "eval_accuracy": 0.5605742296918768,
-    "eval_f1_macro": 0.7342630546801885,
-    "eval_f1_micro": 0.8543162417321499,
-    "eval_loss": 0.08401281386613846,
-    "eval_roc_auc": 0.9076857807628663,
-    "eval_runtime": 670.4543,
-    "eval_samples_per_second": 4.26,
-    "eval_steps_per_second": 0.134,
-    "learning_rate": 0.0001
+    "epoch": 114.0,
+    "eval_accuracy": 0.5738795518207283,
+    "eval_f1_macro": 0.7429818572746157,
+    "eval_f1_micro": 0.8570132153593103,
+    "eval_loss": 0.08252906054258347,
+    "eval_roc_auc": 0.9080498428032193,
+    "eval_runtime": 681.2636,
+    "eval_samples_per_second": 4.192,
+    "eval_steps_per_second": 0.132,
+    "learning_rate": 1.0000000000000002e-07
 }

train_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 70.0,
-    "learning_rate": 0.0001,
-    "train_loss": 0.11672632308896316,
-    "train_runtime": 200748.2354,
-    "train_samples_per_second": 3.057,
-    "train_steps_per_second": 0.096
+    "epoch": 114.0,
+    "learning_rate": 1.0000000000000002e-07,
+    "train_loss": 0.033337813025782426,
+    "train_runtime": 131818.0427,
+    "train_samples_per_second": 9.311,
+    "train_steps_per_second": 0.291
 }

trainer_state.json

@@ -1,9 +1,9 @@
 {
-  "best_metric": 0.08306006342172623,
-  "best_model_checkpoint": "/home1/datawork/mcontini/models/multilabel/huggingface/dinov2-large-2024_01_05-kornia_img-size518_batch-size32_epochs70_freeze/checkpoint-19180",
-  "epoch": 70.0,
+  "best_metric": 0.08124219626188278,
+  "best_model_checkpoint": "/home1/datawork/mcontini/models/multilabel/huggingface/dinov2-large-2024_01_05-kornia_img-size518_batch-size32_epochs70_freeze/checkpoint-28496",
+  "epoch": 114.0,
   "eval_steps": 500,
-  "global_step": 19180,
+  "global_step": 31236,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -1147,21 +1147,737 @@
       "step": 19180
     },
     {
-      "epoch": 70.0,
+      "epoch": 71.0,
+      "eval_accuracy": 0.557920446615492,
+      "eval_f1_macro": 0.74841480686329,
+      "eval_f1_micro": 0.8557041347283614,
+      "eval_loss": 0.08348394185304642,
+      "eval_roc_auc": 0.9102149107350765,
+      "eval_runtime": 688.9991,
+      "eval_samples_per_second": 4.16,
+      "eval_steps_per_second": 0.131,
+      "learning_rate": 0.0001,
+      "step": 19454
+    },
+    {
+      "epoch": 71.17,
+      "learning_rate": 0.0001,
+      "loss": 0.0907,
+      "step": 19500
+    },
+    {
+      "epoch": 72.0,
+      "eval_accuracy": 0.5610607117934403,
+      "eval_f1_macro": 0.7445560004670461,
+      "eval_f1_micro": 0.8531667140017041,
+      "eval_loss": 0.08317266404628754,
+      "eval_roc_auc": 0.9036835353705291,
+      "eval_runtime": 686.0467,
+      "eval_samples_per_second": 4.178,
+      "eval_steps_per_second": 0.131,
+      "learning_rate": 0.0001,
+      "step": 19728
+    },
+    {
+      "epoch": 72.99,
+      "learning_rate": 0.0001,
+      "loss": 0.0905,
+      "step": 20000
+    },
+    {
+      "epoch": 73.0,
+      "eval_accuracy": 0.5575715282623867,
+      "eval_f1_macro": 0.7511661878734329,
+      "eval_f1_micro": 0.8557848235754626,
+      "eval_loss": 0.0826636552810669,
+      "eval_roc_auc": 0.9104516775814524,
+      "eval_runtime": 696.2462,
+      "eval_samples_per_second": 4.116,
+      "eval_steps_per_second": 0.129,
+      "learning_rate": 0.0001,
+      "step": 20002
+    },
+    {
+      "epoch": 74.0,
+      "eval_accuracy": 0.5589672016748081,
+      "eval_f1_macro": 0.7519078199055363,
+      "eval_f1_micro": 0.8547880690737834,
+      "eval_loss": 0.08345863223075867,
+      "eval_roc_auc": 0.9089978296204139,
+      "eval_runtime": 684.5044,
+      "eval_samples_per_second": 4.187,
+      "eval_steps_per_second": 0.131,
+      "learning_rate": 0.0001,
+      "step": 20276
+    },
+    {
+      "epoch": 74.82,
+      "learning_rate": 0.0001,
+      "loss": 0.0896,
+      "step": 20500
+    },
+    {
+      "epoch": 75.0,
+      "eval_accuracy": 0.5565247732030705,
+      "eval_f1_macro": 0.7427640048071817,
+      "eval_f1_micro": 0.8535384963511908,
+      "eval_loss": 0.08291995525360107,
+      "eval_roc_auc": 0.9052642407433178,
+      "eval_runtime": 693.0369,
+      "eval_samples_per_second": 4.135,
+      "eval_steps_per_second": 0.13,
+      "learning_rate": 0.0001,
+      "step": 20550
+    },
+    {
+      "epoch": 76.0,
+      "eval_accuracy": 0.5642009769713887,
+      "eval_f1_macro": 0.7449358839801685,
+      "eval_f1_micro": 0.8560516708789666,
+      "eval_loss": 0.08278567343950272,
+      "eval_roc_auc": 0.9091009598641902,
+      "eval_runtime": 689.5027,
+      "eval_samples_per_second": 4.157,
+      "eval_steps_per_second": 0.131,
+      "learning_rate": 0.0001,
+      "step": 20824
+    },
+    {
+      "epoch": 76.64,
+      "learning_rate": 0.0001,
+      "loss": 0.089,
+      "step": 21000
+    },
+    {
+      "epoch": 77.0,
+      "eval_accuracy": 0.5603628750872296,
+      "eval_f1_macro": 0.7507123347472292,
+      "eval_f1_micro": 0.8567585743106928,
+      "eval_loss": 0.08268015086650848,
+      "eval_roc_auc": 0.9102413784270108,
+      "eval_runtime": 691.3385,
+      "eval_samples_per_second": 4.146,
+      "eval_steps_per_second": 0.13,
+      "learning_rate": 0.0001,
+      "step": 21098
+    },
+    {
+      "epoch": 78.0,
+      "eval_accuracy": 0.557920446615492,
+      "eval_f1_macro": 0.7435504236075495,
+      "eval_f1_micro": 0.8528667079277555,
+      "eval_loss": 0.0833190307021141,
+      "eval_roc_auc": 0.9067310989812636,
+      "eval_runtime": 698.1968,
+      "eval_samples_per_second": 4.105,
+      "eval_steps_per_second": 0.129,
+      "learning_rate": 0.0001,
+      "step": 21372
+    },
+    {
+      "epoch": 78.47,
+      "learning_rate": 0.0001,
+      "loss": 0.0892,
+      "step": 21500
+    },
+    {
+      "epoch": 79.0,
+      "eval_accuracy": 0.5589672016748081,
+      "eval_f1_macro": 0.7501683098189544,
+      "eval_f1_micro": 0.8540393754243042,
+      "eval_loss": 0.08301213383674622,
+      "eval_roc_auc": 0.905519599881194,
+      "eval_runtime": 693.3725,
+      "eval_samples_per_second": 4.133,
+      "eval_steps_per_second": 0.13,
       "learning_rate": 0.0001,
-      "step": 19180,
-      "total_flos": 9.099793269879256e+20,
-      "train_loss": 0.11672632308896316,
-      "train_runtime": 200748.2354,
-      "train_samples_per_second": 3.057,
-      "train_steps_per_second": 0.096
+      "step": 21646
+    },
+    {
+      "epoch": 80.0,
+      "eval_accuracy": 0.5600139567341242,
+      "eval_f1_macro": 0.7461247141006309,
+      "eval_f1_micro": 0.8548240635641317,
+      "eval_loss": 0.08266153186559677,
+      "eval_roc_auc": 0.9048757931076342,
+      "eval_runtime": 692.1533,
+      "eval_samples_per_second": 4.141,
+      "eval_steps_per_second": 0.13,
+      "learning_rate": 1e-05,
+      "step": 21920
+    },
+    {
+      "epoch": 80.29,
+      "learning_rate": 1e-05,
+      "loss": 0.0879,
+      "step": 22000
+    },
+    {
+      "epoch": 81.0,
+      "eval_accuracy": 0.560711793440335,
+      "eval_f1_macro": 0.7542778186740821,
+      "eval_f1_micro": 0.8576379185065117,
+      "eval_loss": 0.08230110257863998,
+      "eval_roc_auc": 0.9116332184274908,
+      "eval_runtime": 691.9752,
+      "eval_samples_per_second": 4.142,
+      "eval_steps_per_second": 0.13,
+      "learning_rate": 1e-05,
+      "step": 22194
+    },
+    {
+      "epoch": 82.0,
+      "eval_accuracy": 0.5631542219120725,
+      "eval_f1_macro": 0.7536111063550082,
+      "eval_f1_micro": 0.8576225654801881,
+      "eval_loss": 0.08222728967666626,
+      "eval_roc_auc": 0.9111816666216643,
+      "eval_runtime": 688.2687,
+      "eval_samples_per_second": 4.164,
+      "eval_steps_per_second": 0.131,
+      "learning_rate": 1e-05,
+      "step": 22468
+    },
+    {
+      "epoch": 82.12,
+      "learning_rate": 1e-05,
+      "loss": 0.0867,
+      "step": 22500
+    },
+    {
+      "epoch": 83.0,
+      "eval_accuracy": 0.5624563852058618,
+      "eval_f1_macro": 0.7519980936807484,
+      "eval_f1_micro": 0.8553623024191264,
+      "eval_loss": 0.08224428445100784,
+      "eval_roc_auc": 0.9057647902327958,
+      "eval_runtime": 675.1087,
+      "eval_samples_per_second": 4.245,
+      "eval_steps_per_second": 0.133,
+      "learning_rate": 1e-05,
+      "step": 22742
+    },
+    {
+      "epoch": 83.94,
+      "learning_rate": 1e-05,
+      "loss": 0.0864,
+      "step": 23000
+    },
+    {
+      "epoch": 84.0,
+      "eval_accuracy": 0.5638520586182834,
+      "eval_f1_macro": 0.751088987860079,
+      "eval_f1_micro": 0.8550642905481616,
+      "eval_loss": 0.08214889466762543,
+      "eval_roc_auc": 0.9071555340236915,
+      "eval_runtime": 684.6287,
+      "eval_samples_per_second": 4.186,
+      "eval_steps_per_second": 0.131,
+      "learning_rate": 1e-05,
+      "step": 23016
+    },
+    {
+      "epoch": 85.0,
+      "eval_accuracy": 0.5617585484996511,
+      "eval_f1_macro": 0.7532850236770345,
+      "eval_f1_micro": 0.8560117633751838,
+      "eval_loss": 0.08197268098592758,
+      "eval_roc_auc": 0.9067157953185159,
+      "eval_runtime": 685.9962,
+      "eval_samples_per_second": 4.178,
+      "eval_steps_per_second": 0.131,
+      "learning_rate": 1e-05,
+      "step": 23290
+    },
+    {
+      "epoch": 85.77,
+      "learning_rate": 1e-05,
+      "loss": 0.0865,
+      "step": 23500
+    },
+    {
+      "epoch": 86.0,
+      "eval_accuracy": 0.5600139567341242,
+      "eval_f1_macro": 0.7495768803790436,
+      "eval_f1_micro": 0.8553152949167893,
+      "eval_loss": 0.0821395218372345,
+      "eval_roc_auc": 0.9060269285615214,
+      "eval_runtime": 676.0683,
+      "eval_samples_per_second": 4.239,
+      "eval_steps_per_second": 0.133,
+      "learning_rate": 1e-05,
+      "step": 23564
+    },
+    {
+      "epoch": 87.0,
+      "eval_accuracy": 0.5586182833217027,
+      "eval_f1_macro": 0.7518865385404515,
+      "eval_f1_micro": 0.8558878188883257,
+      "eval_loss": 0.08169202506542206,
+      "eval_roc_auc": 0.9080884949227034,
+      "eval_runtime": 683.932,
+      "eval_samples_per_second": 4.19,
+      "eval_steps_per_second": 0.132,
+      "learning_rate": 1e-05,
+      "step": 23838
+    },
+    {
+      "epoch": 87.59,
+      "learning_rate": 1e-05,
+      "loss": 0.0868,
+      "step": 24000
+    },
+    {
+      "epoch": 88.0,
+      "eval_accuracy": 0.5621074668527565,
+      "eval_f1_macro": 0.7526313975322289,
+      "eval_f1_micro": 0.8558239036198839,
+      "eval_loss": 0.08171343803405762,
+      "eval_roc_auc": 0.9081685470699702,
+      "eval_runtime": 687.6304,
+      "eval_samples_per_second": 4.168,
+      "eval_steps_per_second": 0.131,
+      "learning_rate": 1e-05,
+      "step": 24112
+    },
+    {
+      "epoch": 89.0,
+      "eval_accuracy": 0.5638520586182834,
+      "eval_f1_macro": 0.7535945449960157,
+      "eval_f1_micro": 0.8569978572234127,
+      "eval_loss": 0.08180436491966248,
+      "eval_roc_auc": 0.9082570227224631,
+      "eval_runtime": 685.8987,
+      "eval_samples_per_second": 4.178,
+      "eval_steps_per_second": 0.131,
+      "learning_rate": 1e-05,
+      "step": 24386
+    },
+    {
+      "epoch": 89.42,
+      "learning_rate": 1e-05,
+      "loss": 0.0857,
+      "step": 24500
+    },
+    {
+      "epoch": 90.0,
+      "eval_accuracy": 0.5617585484996511,
+      "eval_f1_macro": 0.7521712775972677,
+      "eval_f1_micro": 0.8557914296976182,
+      "eval_loss": 0.08184907585382462,
+      "eval_roc_auc": 0.9080738403314199,
+      "eval_runtime": 2199.2107,
+      "eval_samples_per_second": 1.303,
+      "eval_steps_per_second": 0.041,
+      "learning_rate": 1e-05,
+      "step": 24660
+    },
+    {
+      "epoch": 91.0,
+      "eval_accuracy": 0.5631542219120725,
+      "eval_f1_macro": 0.7496381590553177,
+      "eval_f1_micro": 0.856868937514099,
+      "eval_loss": 0.08175913989543915,
+      "eval_roc_auc": 0.9081476613926293,
+      "eval_runtime": 685.7142,
+      "eval_samples_per_second": 4.18,
+      "eval_steps_per_second": 0.131,
+      "learning_rate": 1e-05,
+      "step": 24934
+    },
+    {
+      "epoch": 91.24,
+      "learning_rate": 1e-05,
+      "loss": 0.0862,
+      "step": 25000
+    },
+    {
+      "epoch": 92.0,
+      "eval_accuracy": 0.5648988136775994,
+      "eval_f1_macro": 0.7551884126903309,
+      "eval_f1_micro": 0.8566131025957974,
+      "eval_loss": 0.08205177634954453,
+      "eval_roc_auc": 0.9092762668547993,
+      "eval_runtime": 688.0975,
+      "eval_samples_per_second": 4.165,
+      "eval_steps_per_second": 0.131,
+      "learning_rate": 1e-05,
+      "step": 25208
+    },
+    {
+      "epoch": 93.0,
+      "eval_accuracy": 0.5628053035589672,
+      "eval_f1_macro": 0.7579638727848569,
+      "eval_f1_micro": 0.8588642195693407,
+      "eval_loss": 0.08153587579727173,
+      "eval_roc_auc": 0.9129889700545543,
+      "eval_runtime": 686.4355,
+      "eval_samples_per_second": 4.175,
+      "eval_steps_per_second": 0.131,
+      "learning_rate": 1e-05,
+      "step": 25482
+    },
+    {
+      "epoch": 93.07,
+      "learning_rate": 1e-05,
+      "loss": 0.0851,
+      "step": 25500
+    },
+    {
+      "epoch": 94.0,
+      "eval_accuracy": 0.5600139567341242,
+      "eval_f1_macro": 0.7565755790887538,
+      "eval_f1_micro": 0.8570630932439977,
+      "eval_loss": 0.0816139355301857,
+      "eval_roc_auc": 0.911680023691974,
+      "eval_runtime": 686.9681,
+      "eval_samples_per_second": 4.172,
+      "eval_steps_per_second": 0.131,
+      "learning_rate": 1e-05,
+      "step": 25756
+    },
+    {
+      "epoch": 94.89,
+      "learning_rate": 1e-05,
+      "loss": 0.0854,
+      "step": 26000
+    },
+    {
+      "epoch": 95.0,
+      "eval_accuracy": 0.5631542219120725,
+      "eval_f1_macro": 0.7553215815461111,
+      "eval_f1_micro": 0.8563582357226924,
+      "eval_loss": 0.08153104782104492,
+      "eval_roc_auc": 0.9100006738804182,
+      "eval_runtime": 792.6719,
+      "eval_samples_per_second": 3.616,
+      "eval_steps_per_second": 0.114,
+      "learning_rate": 1e-05,
+      "step": 26030
+    },
+    {
+      "epoch": 96.0,
+      "eval_accuracy": 0.5621074668527565,
+      "eval_f1_macro": 0.7584737150271473,
+      "eval_f1_micro": 0.8576208800401539,
+      "eval_loss": 0.081536203622818,
+      "eval_roc_auc": 0.9123942619313182,
+      "eval_runtime": 689.1211,
+      "eval_samples_per_second": 4.159,
+      "eval_steps_per_second": 0.131,
+      "learning_rate": 1e-05,
+      "step": 26304
+    },
+    {
+      "epoch": 96.72,
+      "learning_rate": 1e-05,
+      "loss": 0.0854,
+      "step": 26500
+    },
+    {
+      "epoch": 97.0,
+      "eval_accuracy": 0.5628053035589672,
+      "eval_f1_macro": 0.7578675241138624,
+      "eval_f1_micro": 0.8575591168889387,
+      "eval_loss": 0.0817214846611023,
+      "eval_roc_auc": 0.9107227875201963,
+      "eval_runtime": 791.8215,
+      "eval_samples_per_second": 3.62,
+      "eval_steps_per_second": 0.114,
+      "learning_rate": 1e-05,
+      "step": 26578
+    },
+    {
+      "epoch": 98.0,
+      "eval_accuracy": 0.5638520586182834,
+      "eval_f1_macro": 0.7526862577512524,
+      "eval_f1_micro": 0.8570947651910453,
+      "eval_loss": 0.08161807805299759,
+      "eval_roc_auc": 0.9100232038721354,
+      "eval_runtime": 689.7672,
+      "eval_samples_per_second": 4.155,
+      "eval_steps_per_second": 0.13,
+      "learning_rate": 1e-05,
+      "step": 26852
+    },
+    {
+      "epoch": 98.54,
+      "learning_rate": 1e-05,
+      "loss": 0.0855,
+      "step": 27000
+    },
+    {
+      "epoch": 99.0,
+      "eval_accuracy": 0.5642009769713887,
+      "eval_f1_macro": 0.7556285126609241,
+      "eval_f1_micro": 0.8578035986237239,
+      "eval_loss": 0.08184286206960678,
+      "eval_roc_auc": 0.9086036990034648,
+      "eval_runtime": 685.3364,
+      "eval_samples_per_second": 4.182,
+      "eval_steps_per_second": 0.131,
+      "learning_rate": 1e-05,
+      "step": 27126
+    },
+    {
+      "epoch": 100.0,
+      "eval_accuracy": 0.5631542219120725,
+      "eval_f1_macro": 0.753319281703831,
+      "eval_f1_micro": 0.857062242537103,
+      "eval_loss": 0.08161789923906326,
+      "eval_roc_auc": 0.9080422377630124,
+      "eval_runtime": 680.1579,
+      "eval_samples_per_second": 4.214,
+      "eval_steps_per_second": 0.132,
+      "learning_rate": 1.0000000000000002e-06,
+      "step": 27400
+    },
+    {
+      "epoch": 100.36,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 0.0837,
+      "step": 27500
+    },
+    {
+      "epoch": 101.0,
+      "eval_accuracy": 0.5645498953244941,
+      "eval_f1_macro": 0.7553159980269647,
+      "eval_f1_micro": 0.8575286968264686,
+      "eval_loss": 0.08142262697219849,
+      "eval_roc_auc": 0.9092807526598085,
+      "eval_runtime": 693.9432,
+      "eval_samples_per_second": 4.13,
+      "eval_steps_per_second": 0.13,
+      "learning_rate": 1.0000000000000002e-06,
+      "step": 27674
+    },
+    {
+      "epoch": 102.0,
+      "eval_accuracy": 0.5652477320307048,
+      "eval_f1_macro": 0.7559444477928223,
+      "eval_f1_micro": 0.8571749382438806,
+      "eval_loss": 0.08143333345651627,
+      "eval_roc_auc": 0.9098557722819599,
+      "eval_runtime": 695.0171,
+      "eval_samples_per_second": 4.124,
+      "eval_steps_per_second": 0.129,
+      "learning_rate": 1.0000000000000002e-06,
+      "step": 27948
+    },
+    {
+      "epoch": 102.19,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 0.085,
+      "step": 28000
+    },
+    {
+      "epoch": 103.0,
+      "eval_accuracy": 0.5645498953244941,
+      "eval_f1_macro": 0.7565514506209431,
+      "eval_f1_micro": 0.8570462613399448,
+      "eval_loss": 0.08157742768526077,
+      "eval_roc_auc": 0.9085338156424723,
+      "eval_runtime": 710.2061,
+      "eval_samples_per_second": 4.035,
+      "eval_steps_per_second": 0.127,
+      "learning_rate": 1.0000000000000002e-06,
+      "step": 28222
+    },
+    {
+      "epoch": 104.0,
+      "eval_accuracy": 0.5645498953244941,
+      "eval_f1_macro": 0.7572978759334409,
+      "eval_f1_micro": 0.857623884617543,
+      "eval_loss": 0.08124219626188278,
+      "eval_roc_auc": 0.9102385369363781,
+      "eval_runtime": 703.8703,
+      "eval_samples_per_second": 4.072,
+      "eval_steps_per_second": 0.128,
+      "learning_rate": 1.0000000000000002e-06,
+      "step": 28496
+    },
+    {
+      "epoch": 104.01,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 0.0844,
+      "step": 28500
+    },
+    {
+      "epoch": 105.0,
+      "eval_accuracy": 0.5603628750872296,
+      "eval_f1_macro": 0.7588882568234135,
+      "eval_f1_micro": 0.8571747004736696,
+      "eval_loss": 0.08170615136623383,
+      "eval_roc_auc": 0.9124156957134514,
+      "eval_runtime": 693.4706,
+      "eval_samples_per_second": 4.133,
+      "eval_steps_per_second": 0.13,
+      "learning_rate": 1.0000000000000002e-06,
+      "step": 28770
+    },
+    {
+      "epoch": 105.84,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 0.0845,
+      "step": 29000
+    },
+    {
+      "epoch": 106.0,
+      "eval_accuracy": 0.5628053035589672,
+      "eval_f1_macro": 0.7514055460147794,
+      "eval_f1_micro": 0.8562891131532955,
+      "eval_loss": 0.08144387602806091,
+      "eval_roc_auc": 0.9079250010327448,
+      "eval_runtime": 693.924,
+      "eval_samples_per_second": 4.13,
+      "eval_steps_per_second": 0.13,
+      "learning_rate": 1.0000000000000002e-06,
+      "step": 29044
+    },
+    {
+      "epoch": 107.0,
+      "eval_accuracy": 0.5635031402651779,
+      "eval_f1_macro": 0.7490061784836257,
+      "eval_f1_micro": 0.8557823129251702,
+      "eval_loss": 0.08168599754571915,
+      "eval_roc_auc": 0.9057833825242961,
+      "eval_runtime": 697.3665,
+      "eval_samples_per_second": 4.11,
+      "eval_steps_per_second": 0.129,
+      "learning_rate": 1.0000000000000002e-06,
+      "step": 29318
+    },
+    {
+      "epoch": 107.66,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 0.0854,
+      "step": 29500
+    },
+    {
+      "epoch": 108.0,
+      "eval_accuracy": 0.5642009769713887,
+      "eval_f1_macro": 0.7569164155873785,
+      "eval_f1_micro": 0.8569341425039335,
+      "eval_loss": 0.08160943537950516,
+      "eval_roc_auc": 0.9094149373671999,
+      "eval_runtime": 692.7895,
+      "eval_samples_per_second": 4.137,
+      "eval_steps_per_second": 0.13,
+      "learning_rate": 1.0000000000000002e-06,
+      "step": 29592
+    },
+    {
+      "epoch": 109.0,
+      "eval_accuracy": 0.5652477320307048,
+      "eval_f1_macro": 0.7558234898859766,
+      "eval_f1_micro": 0.8573669467787115,
+      "eval_loss": 0.08135490119457245,
+      "eval_roc_auc": 0.9106934783376293,
+      "eval_runtime": 695.7903,
+      "eval_samples_per_second": 4.119,
+      "eval_steps_per_second": 0.129,
+      "learning_rate": 1.0000000000000002e-06,
+      "step": 29866
+    },
+    {
+      "epoch": 109.49,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 0.0854,
+      "step": 30000
+    },
+    {
+      "epoch": 110.0,
+      "eval_accuracy": 0.5638520586182834,
+      "eval_f1_macro": 0.7564667466516518,
+      "eval_f1_micro": 0.8577973962116557,
+      "eval_loss": 0.08132224529981613,
+      "eval_roc_auc": 0.9118372864958749,
+      "eval_runtime": 682.0266,
+      "eval_samples_per_second": 4.202,
+      "eval_steps_per_second": 0.132,
+      "learning_rate": 1.0000000000000002e-06,
+      "step": 30140
+    },
+    {
+      "epoch": 111.0,
+      "eval_accuracy": 0.5638520586182834,
+      "eval_f1_macro": 0.7579481466413214,
+      "eval_f1_micro": 0.8576381122791322,
+      "eval_loss": 0.0814104825258255,
+      "eval_roc_auc": 0.911498485615307,
+      "eval_runtime": 686.5544,
+      "eval_samples_per_second": 4.174,
+      "eval_steps_per_second": 0.131,
+      "learning_rate": 1.0000000000000002e-07,
+      "step": 30414
+    },
+    {
+      "epoch": 111.31,
+      "learning_rate": 1.0000000000000002e-07,
+      "loss": 0.0851,
+      "step": 30500
+    },
+    {
+      "epoch": 112.0,
+      "eval_accuracy": 0.5631542219120725,
+      "eval_f1_macro": 0.7575801018117111,
+      "eval_f1_micro": 0.8580721134974485,
+      "eval_loss": 0.08170080929994583,
+      "eval_roc_auc": 0.91075603440298,
+      "eval_runtime": 746.778,
+      "eval_samples_per_second": 3.838,
+      "eval_steps_per_second": 0.121,
+      "learning_rate": 1.0000000000000002e-07,
+      "step": 30688
+    },
+    {
+      "epoch": 113.0,
+      "eval_accuracy": 0.5614096301465457,
+      "eval_f1_macro": 0.7562692232145891,
+      "eval_f1_micro": 0.8583059164668487,
+      "eval_loss": 0.081505186855793,
+      "eval_roc_auc": 0.912813663063945,
+      "eval_runtime": 687.4479,
+      "eval_samples_per_second": 4.169,
+      "eval_steps_per_second": 0.131,
+      "learning_rate": 1.0000000000000002e-07,
+      "step": 30962
+    },
+    {
+      "epoch": 113.14,
+      "learning_rate": 1.0000000000000002e-07,
+      "loss": 0.0848,
+      "step": 31000
+    },
+    {
+      "epoch": 114.0,
+      "eval_accuracy": 0.5655966503838102,
+      "eval_f1_macro": 0.7560403631861836,
+      "eval_f1_micro": 0.8563654914408797,
+      "eval_loss": 0.08186182379722595,
+      "eval_roc_auc": 0.9061407756963643,
+      "eval_runtime": 731.2304,
+      "eval_samples_per_second": 3.919,
+      "eval_steps_per_second": 0.123,
+      "learning_rate": 1.0000000000000002e-07,
+      "step": 31236
+    },
+    {
+      "epoch": 114.0,
+      "learning_rate": 1.0000000000000002e-07,
+      "step": 31236,
+      "total_flos": 1.4819663325231928e+21,
+      "train_loss": 0.033337813025782426,
+      "train_runtime": 131818.0427,
+      "train_samples_per_second": 9.311,
+      "train_steps_per_second": 0.291
     }
   ],
   "logging_steps": 500,
-  "max_steps": 19180,
-  "num_train_epochs": 70,
+  "max_steps": 38360,
+  "num_train_epochs": 140,
   "save_steps": 500,
-  "total_flos": 9.099793269879256e+20,
+  "total_flos": 1.4819663325231928e+21,
   "trial_name": null,
   "trial_params": null
 }