End of training

all_results.json

@@ -1,30 +1,34 @@
 {
     "epoch": 10.0,
-    "eval_PREDICATE_f1": 0.9649805447470817,
+    "eval_PREDICATE_f1": 0.9727626459143969,
     "eval_PREDICATE_number": 126,
-    "eval_PREDICATE_precision": 0.9465648854961832,
-    "eval_PREDICATE_recall": 0.9841269841269841,
-    "eval_loss": 0.18749289214611053,
-    "eval_overall_accuracy": 0.9576365663322185,
-    "eval_overall_f1": 0.8493150684931506,
-    "eval_overall_precision": 0.8188679245283019,
-    "eval_overall_recall": 0.8821138211382114,
-    "eval_runtime": 1.816,
+    "eval_PREDICATE_precision": 0.9541984732824428,
+    "eval_PREDICATE_recall": 0.9920634920634921,
+    "eval_loss": 0.20896820724010468,
+    "eval_overall_accuracy": 0.9609810479375697,
+    "eval_overall_f1": 0.8588469184890657,
+    "eval_overall_precision": 0.8404669260700389,
+    "eval_overall_recall": 0.8780487804878049,
+    "eval_runtime": 1.8982,
     "eval_samples": 99,
-    "eval_samples_per_second": 54.515,
-    "eval_steps_per_second": 7.159,
+    "eval_samples_per_second": 52.155,
+    "eval_steps_per_second": 6.849,
+    "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
+    "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
+    "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
+    "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
     "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0,
     "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 0,
     "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0,
     "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0,
-    "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7123287671232876,
+    "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7142857142857144,
     "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
-    "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7027027027027027,
-    "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.7222222222222222,
-    "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8817204301075268,
-    "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 44,
-    "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.8367346938775511,
-    "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.9318181818181818,
+    "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7352941176470589,
+    "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.6944444444444444,
+    "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8200000000000001,
+    "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
+    "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7735849056603774,
+    "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8723404255319149,
     "eval_\u0434\u0440\u0443\u0433\u043e\u0435_f1": 0.0,
     "eval_\u0434\u0440\u0443\u0433\u043e\u0435_number": 1,
     "eval_\u0434\u0440\u0443\u0433\u043e\u0435_precision": 0.0,
@@ -33,9 +37,9 @@
     "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_number": 3,
     "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_precision": 0.0,
     "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_recall": 0.0,
-    "train_loss": 0.13130194339568593,
-    "train_runtime": 174.9188,
+    "train_loss": 0.11747416195326618,
+    "train_runtime": 191.5912,
     "train_samples": 884,
-    "train_samples_per_second": 50.538,
-    "train_steps_per_second": 3.201
+    "train_samples_per_second": 46.14,
+    "train_steps_per_second": 2.923
 }

config.json

@@ -12,27 +12,23 @@
   "hidden_size": 768,
   "id2label": {
     "0": "B-PREDICATE",
-    "1": "B-\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422",
-    "2": "B-\u041a\u0410\u0423\u0417\u0410\u0422\u0418\u0412",
-    "3": "B-\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420",
-    "4": "B-\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420",
-    "5": "B-\u0434\u0440\u0443\u0433\u043e\u0435",
-    "6": "B-\u043a\u0430\u0443\u0437\u0430\u0442\u043e\u0440",
-    "7": "B-\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440",
-    "8": "O"
+    "1": "B-\u0414\u0420\u0423\u0413\u041e\u0415",
+    "2": "B-\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422",
+    "3": "B-\u041a\u0410\u0423\u0417\u0410\u0422\u0418\u0412",
+    "4": "B-\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420",
+    "5": "B-\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420",
+    "6": "O"
   },
   "initializer_range": 0.02,
   "intermediate_size": 3072,
   "label2id": {
     "B-PREDICATE": 0,
-    "B-\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422": 1,
-    "B-\u041a\u0410\u0423\u0417\u0410\u0422\u0418\u0412": 2,
-    "B-\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420": 3,
-    "B-\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420": 4,
-    "B-\u0434\u0440\u0443\u0433\u043e\u0435": 5,
-    "B-\u043a\u0430\u0443\u0437\u0430\u0442\u043e\u0440": 6,
-    "B-\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440": 7,
-    "O": 8
+    "B-\u0414\u0420\u0423\u0413\u041e\u0415": 1,
+    "B-\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422": 2,
+    "B-\u041a\u0410\u0423\u0417\u0410\u0422\u0418\u0412": 3,
+    "B-\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420": 4,
+    "B-\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420": 5,
+    "O": 6
   },
   "layer_norm_eps": 1e-12,
   "max_position_embeddings": 512,

eval_results.json

@@ -1,36 +1,32 @@
 {
     "epoch": 10.0,
-    "eval_PREDICATE_f1": 0.9649805447470817,
+    "eval_PREDICATE_f1": 0.9727626459143969,
     "eval_PREDICATE_number": 126,
-    "eval_PREDICATE_precision": 0.9465648854961832,
-    "eval_PREDICATE_recall": 0.9841269841269841,
-    "eval_loss": 0.18749289214611053,
-    "eval_overall_accuracy": 0.9576365663322185,
-    "eval_overall_f1": 0.8493150684931506,
-    "eval_overall_precision": 0.8188679245283019,
-    "eval_overall_recall": 0.8821138211382114,
-    "eval_runtime": 1.816,
+    "eval_PREDICATE_precision": 0.9541984732824428,
+    "eval_PREDICATE_recall": 0.9920634920634921,
+    "eval_loss": 0.20896820724010468,
+    "eval_overall_accuracy": 0.9609810479375697,
+    "eval_overall_f1": 0.8588469184890657,
+    "eval_overall_precision": 0.8404669260700389,
+    "eval_overall_recall": 0.8780487804878049,
+    "eval_runtime": 1.8982,
     "eval_samples": 99,
-    "eval_samples_per_second": 54.515,
-    "eval_steps_per_second": 7.159,
+    "eval_samples_per_second": 52.155,
+    "eval_steps_per_second": 6.849,
+    "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
+    "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
+    "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
+    "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
     "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0,
     "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 0,
     "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0,
     "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0,
-    "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7123287671232876,
+    "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7142857142857144,
     "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
-    "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7027027027027027,
-    "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.7222222222222222,
-    "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8817204301075268,
-    "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 44,
-    "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.8367346938775511,
-    "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.9318181818181818,
-    "eval_\u0434\u0440\u0443\u0433\u043e\u0435_f1": 0.0,
-    "eval_\u0434\u0440\u0443\u0433\u043e\u0435_number": 1,
-    "eval_\u0434\u0440\u0443\u0433\u043e\u0435_precision": 0.0,
-    "eval_\u0434\u0440\u0443\u0433\u043e\u0435_recall": 0.0,
-    "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_f1": 0.0,
-    "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_number": 3,
-    "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_precision": 0.0,
-    "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_recall": 0.0
+    "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7352941176470589,
+    "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.6944444444444444,
+    "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8200000000000001,
+    "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
+    "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7735849056603774,
+    "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8723404255319149
 }

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7af8e323efcb030ab0858fb32ffefc6630dfeb0bcd200f59f45e8017ab2e3616
-size 710978033
+oid sha256:c9b1cd89ef795b2afec3e5af34d3494ffb554c05e4a7697e698196da63d1bde6
+size 710971889

train_results.json

@@ -1,8 +1,8 @@
 {
     "epoch": 10.0,
-    "train_loss": 0.13130194339568593,
-    "train_runtime": 174.9188,
+    "train_loss": 0.11747416195326618,
+    "train_runtime": 191.5912,
     "train_samples": 884,
-    "train_samples_per_second": 50.538,
-    "train_steps_per_second": 3.201
+    "train_samples_per_second": 46.14,
+    "train_steps_per_second": 2.923
 }

trainer_state.json

@@ -10,269 +10,261 @@
     {
       "epoch": 0.02,
       "learning_rate": 1.4705882352941177e-06,
-      "loss": 2.233,
+      "loss": 1.9437,
       "step": 1
     },
     {
       "epoch": 0.09,
       "learning_rate": 7.3529411764705884e-06,
-      "loss": 2.1869,
+      "loss": 1.8725,
       "step": 5
     },
     {
       "epoch": 0.18,
       "learning_rate": 1.4705882352941177e-05,
-      "loss": 1.9259,
+      "loss": 1.6377,
       "step": 10
     },
     {
       "epoch": 0.27,
       "learning_rate": 2.2058823529411766e-05,
-      "loss": 1.4271,
+      "loss": 1.189,
       "step": 15
     },
     {
       "epoch": 0.36,
       "learning_rate": 2.9411764705882354e-05,
-      "loss": 0.7441,
+      "loss": 0.6265,
       "step": 20
     },
     {
       "epoch": 0.45,
       "learning_rate": 3.6764705882352945e-05,
-      "loss": 0.9447,
+      "loss": 0.8537,
       "step": 25
     },
     {
       "epoch": 0.54,
       "learning_rate": 4.411764705882353e-05,
-      "loss": 0.531,
+      "loss": 0.4837,
       "step": 30
     },
     {
       "epoch": 0.62,
       "learning_rate": 4.9999554100506465e-05,
-      "loss": 0.428,
+      "loss": 0.3967,
       "step": 35
     },
     {
       "epoch": 0.71,
       "learning_rate": 4.998394928831034e-05,
-      "loss": 0.4043,
+      "loss": 0.3544,
       "step": 40
     },
     {
       "epoch": 0.8,
       "learning_rate": 4.9946065404995485e-05,
-      "loss": 0.2392,
+      "loss": 0.2233,
       "step": 45
     },
     {
       "epoch": 0.89,
       "learning_rate": 4.988593623296038e-05,
-      "loss": 0.3225,
+      "loss": 0.3092,
       "step": 50
     },
     {
       "epoch": 0.98,
       "learning_rate": 4.980361539151416e-05,
-      "loss": 0.289,
+      "loss": 0.2689,
       "step": 55
     },
     {
       "epoch": 1.0,
-      "eval_PREDICATE_f1": 0.9498069498069498,
+      "eval_PREDICATE_f1": 0.9429657794676807,
       "eval_PREDICATE_number": 126,
-      "eval_PREDICATE_precision": 0.924812030075188,
-      "eval_PREDICATE_recall": 0.9761904761904762,
-      "eval_loss": 0.16499629616737366,
-      "eval_overall_accuracy": 0.947603121516165,
-      "eval_overall_f1": 0.7853881278538812,
-      "eval_overall_precision": 0.8958333333333334,
-      "eval_overall_recall": 0.6991869918699187,
-      "eval_runtime": 0.4866,
-      "eval_samples_per_second": 203.444,
-      "eval_steps_per_second": 26.715,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.42424242424242425,
+      "eval_PREDICATE_precision": 0.9051094890510949,
+      "eval_PREDICATE_recall": 0.9841269841269841,
+      "eval_loss": 0.1595052033662796,
+      "eval_overall_accuracy": 0.9503901895206243,
+      "eval_overall_f1": 0.8093220338983051,
+      "eval_overall_precision": 0.8451327433628318,
+      "eval_overall_recall": 0.7764227642276422,
+      "eval_runtime": 0.523,
+      "eval_samples_per_second": 189.296,
+      "eval_steps_per_second": 24.857,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.543859649122807,
       "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7777777777777778,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.2916666666666667,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.7368421052631579,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 44,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.875,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.6363636363636364,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_f1": 0.0,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_number": 1,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_precision": 0.0,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_recall": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_f1": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_number": 3,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_precision": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_recall": 0.0,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7380952380952381,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.4305555555555556,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.7659574468085105,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7659574468085106,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.7659574468085106,
       "step": 56
     },
     {
       "epoch": 1.07,
       "learning_rate": 4.969917628906234e-05,
-      "loss": 0.1803,
+      "loss": 0.1956,
       "step": 60
     },
     {
       "epoch": 1.16,
       "learning_rate": 4.9572712057646e-05,
-      "loss": 0.2196,
+      "loss": 0.219,
       "step": 65
     },
     {
       "epoch": 1.25,
       "learning_rate": 4.9424335469892594e-05,
-      "loss": 0.1454,
+      "loss": 0.1433,
       "step": 70
     },
     {
       "epoch": 1.34,
       "learning_rate": 4.925417883845266e-05,
-      "loss": 0.1431,
+      "loss": 0.1429,
       "step": 75
     },
     {
       "epoch": 1.43,
       "learning_rate": 4.9062393898011904e-05,
-      "loss": 0.2021,
+      "loss": 0.2062,
       "step": 80
     },
     {
       "epoch": 1.52,
       "learning_rate": 4.884915166998405e-05,
-      "loss": 0.1312,
+      "loss": 0.1129,
       "step": 85
     },
     {
       "epoch": 1.61,
       "learning_rate": 4.8614642310004974e-05,
-      "loss": 0.1555,
+      "loss": 0.1351,
       "step": 90
     },
     {
       "epoch": 1.7,
       "learning_rate": 4.835907493836421e-05,
-      "loss": 0.2674,
+      "loss": 0.2069,
       "step": 95
     },
     {
       "epoch": 1.79,
       "learning_rate": 4.808267745352502e-05,
-      "loss": 0.1073,
+      "loss": 0.1009,
       "step": 100
     },
     {
       "epoch": 1.88,
       "learning_rate": 4.778569632889919e-05,
-      "loss": 0.1676,
+      "loss": 0.1587,
       "step": 105
     },
     {
       "epoch": 1.96,
       "learning_rate": 4.746839639305808e-05,
-      "loss": 0.1543,
+      "loss": 0.1624,
       "step": 110
     },
     {
       "epoch": 2.0,
-      "eval_PREDICATE_f1": 0.9689922480620156,
+      "eval_PREDICATE_f1": 0.9652509652509653,
       "eval_PREDICATE_number": 126,
-      "eval_PREDICATE_precision": 0.946969696969697,
+      "eval_PREDICATE_precision": 0.9398496240601504,
       "eval_PREDICATE_recall": 0.9920634920634921,
-      "eval_loss": 0.1350417286157608,
-      "eval_overall_accuracy": 0.9587513935340022,
-      "eval_overall_f1": 0.852589641434263,
-      "eval_overall_precision": 0.8359375,
-      "eval_overall_recall": 0.8699186991869918,
-      "eval_runtime": 0.4934,
-      "eval_samples_per_second": 200.654,
-      "eval_steps_per_second": 26.348,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7019867549668874,
+      "eval_loss": 0.12706787884235382,
+      "eval_overall_accuracy": 0.9604236343366778,
+      "eval_overall_f1": 0.8565737051792829,
+      "eval_overall_precision": 0.83984375,
+      "eval_overall_recall": 0.8739837398373984,
+      "eval_runtime": 0.5187,
+      "eval_samples_per_second": 190.846,
+      "eval_steps_per_second": 25.061,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7225806451612904,
       "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.6708860759493671,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.7361111111111112,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8089887640449439,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 44,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.8,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8181818181818182,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_f1": 0.0,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_number": 1,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_precision": 0.0,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_recall": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_f1": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_number": 3,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_precision": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_recall": 0.0,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.6746987951807228,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.7777777777777778,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.7816091954022989,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.85,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.723404255319149,
       "step": 112
     },
     {
       "epoch": 2.05,
       "learning_rate": 4.713106059357554e-05,
-      "loss": 0.0796,
+      "loss": 0.0869,
       "step": 115
     },
     {
       "epoch": 2.14,
       "learning_rate": 4.6773989744713596e-05,
-      "loss": 0.112,
+      "loss": 0.1126,
       "step": 120
     },
     {
       "epoch": 2.23,
       "learning_rate": 4.639750225917573e-05,
-      "loss": 0.1316,
+      "loss": 0.1287,
       "step": 125
     },
     {
       "epoch": 2.32,
       "learning_rate": 4.600193386416697e-05,
-      "loss": 0.0973,
+      "loss": 0.1152,
       "step": 130
     },
     {
       "epoch": 2.41,
       "learning_rate": 4.558763730201413e-05,
-      "loss": 0.1587,
+      "loss": 0.1302,
       "step": 135
     },
     {
       "epoch": 2.5,
       "learning_rate": 4.515498201561296e-05,
-      "loss": 0.0902,
+      "loss": 0.083,
       "step": 140
     },
     {
       "epoch": 2.59,
       "learning_rate": 4.470435381898287e-05,
-      "loss": 0.0881,
+      "loss": 0.0791,
       "step": 145
     },
     {
       "epoch": 2.68,
       "learning_rate": 4.423615455322293e-05,
-      "loss": 0.1606,
+      "loss": 0.1257,
       "step": 150
     },
     {
       "epoch": 2.77,
       "learning_rate": 4.375080172817591e-05,
-      "loss": 0.0843,
+      "loss": 0.0822,
       "step": 155
     },
     {
       "epoch": 2.86,
       "learning_rate": 4.324872815012005e-05,
-      "loss": 0.0939,
+      "loss": 0.0852,
       "step": 160
     },
     {
       "epoch": 2.95,
       "learning_rate": 4.273038153582034e-05,
-      "loss": 0.1525,
+      "loss": 0.1288,
       "step": 165
     },
     {
@@ -281,96 +273,92 @@
       "eval_PREDICATE_number": 126,
       "eval_PREDICATE_precision": 0.9541984732824428,
       "eval_PREDICATE_recall": 0.9920634920634921,
-      "eval_loss": 0.14091716706752777,
-      "eval_overall_accuracy": 0.9548494983277592,
-      "eval_overall_f1": 0.8396946564885496,
-      "eval_overall_precision": 0.7913669064748201,
-      "eval_overall_recall": 0.8943089430894309,
-      "eval_runtime": 1.7665,
-      "eval_samples_per_second": 56.044,
-      "eval_steps_per_second": 7.359,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.6707317073170731,
+      "eval_loss": 0.1378643959760666,
+      "eval_overall_accuracy": 0.9537346711259754,
+      "eval_overall_f1": 0.8415094339622642,
+      "eval_overall_precision": 0.7852112676056338,
+      "eval_overall_recall": 0.9065040650406504,
+      "eval_runtime": 1.8837,
+      "eval_samples_per_second": 52.556,
+      "eval_steps_per_second": 6.901,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.6666666666666666,
       "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.5978260869565217,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.7638888888888888,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8421052631578948,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 44,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7843137254901961,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.9090909090909091,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_f1": 0.0,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_number": 1,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_precision": 0.0,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_recall": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_f1": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_number": 3,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_precision": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_recall": 0.0,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.5833333333333334,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.7777777777777778,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8076923076923077,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7368421052631579,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8936170212765957,
       "step": 168
     },
     {
       "epoch": 3.04,
       "learning_rate": 4.21962241132837e-05,
-      "loss": 0.0602,
+      "loss": 0.0644,
       "step": 170
     },
     {
       "epoch": 3.12,
       "learning_rate": 4.164673220957387e-05,
-      "loss": 0.0662,
+      "loss": 0.0634,
       "step": 175
     },
     {
       "epoch": 3.21,
       "learning_rate": 4.108239582605374e-05,
-      "loss": 0.0804,
+      "loss": 0.0714,
       "step": 180
     },
     {
       "epoch": 3.3,
       "learning_rate": 4.050371820143384e-05,
-      "loss": 0.0492,
+      "loss": 0.0538,
       "step": 185
     },
     {
       "epoch": 3.39,
       "learning_rate": 3.991121536301653e-05,
-      "loss": 0.0752,
+      "loss": 0.0776,
       "step": 190
     },
     {
       "epoch": 3.48,
       "learning_rate": 3.9305415666536236e-05,
-      "loss": 0.0595,
+      "loss": 0.0338,
       "step": 195
     },
     {
       "epoch": 3.57,
       "learning_rate": 3.868685932500595e-05,
-      "loss": 0.0829,
+      "loss": 0.0795,
       "step": 200
     },
     {
       "epoch": 3.66,
       "learning_rate": 3.80560979269902e-05,
-      "loss": 0.0923,
+      "loss": 0.083,
       "step": 205
     },
     {
       "epoch": 3.75,
       "learning_rate": 3.7413693944733995e-05,
-      "loss": 0.033,
+      "loss": 0.0322,
       "step": 210
     },
     {
       "epoch": 3.84,
       "learning_rate": 3.676022023258643e-05,
-      "loss": 0.0674,
+      "loss": 0.081,
       "step": 215
     },
     {
       "epoch": 3.93,
       "learning_rate": 3.609625951616623e-05,
-      "loss": 0.088,
+      "loss": 0.0829,
       "step": 220
     },
     {
@@ -379,190 +367,182 @@
       "eval_PREDICATE_number": 126,
       "eval_PREDICATE_precision": 0.946969696969697,
       "eval_PREDICATE_recall": 0.9920634920634921,
-      "eval_loss": 0.13493739068508148,
+      "eval_loss": 0.14865602552890778,
       "eval_overall_accuracy": 0.959866220735786,
-      "eval_overall_f1": 0.8548707753479126,
-      "eval_overall_precision": 0.8365758754863813,
-      "eval_overall_recall": 0.8739837398373984,
-      "eval_runtime": 0.5164,
-      "eval_samples_per_second": 191.71,
-      "eval_steps_per_second": 25.174,
+      "eval_overall_f1": 0.8554455445544554,
+      "eval_overall_precision": 0.833976833976834,
+      "eval_overall_recall": 0.8780487804878049,
+      "eval_runtime": 0.5199,
+      "eval_samples_per_second": 190.431,
+      "eval_steps_per_second": 25.006,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7123287671232876,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.6938775510204083,
       "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7027027027027027,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.7222222222222222,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8444444444444444,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 44,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.8260869565217391,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8636363636363636,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_f1": 0.0,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_number": 1,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_precision": 0.0,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_recall": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_f1": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_number": 3,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_precision": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_recall": 0.0,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.68,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.7083333333333334,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8163265306122448,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7843137254901961,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.851063829787234,
       "step": 224
     },
     {
       "epoch": 4.02,
       "learning_rate": 3.5422403872724686e-05,
-      "loss": 0.0347,
+      "loss": 0.0289,
       "step": 225
     },
     {
       "epoch": 4.11,
       "learning_rate": 3.473925420316946e-05,
-      "loss": 0.042,
+      "loss": 0.0402,
       "step": 230
     },
     {
       "epoch": 4.2,
       "learning_rate": 3.404741969621996e-05,
-      "loss": 0.037,
+      "loss": 0.0258,
       "step": 235
     },
     {
       "epoch": 4.29,
       "learning_rate": 3.334751728517223e-05,
-      "loss": 0.0259,
+      "loss": 0.0196,
       "step": 240
     },
     {
       "epoch": 4.38,
       "learning_rate": 3.264017109775771e-05,
-      "loss": 0.0471,
+      "loss": 0.0444,
       "step": 245
     },
     {
       "epoch": 4.46,
       "learning_rate": 3.1926011899586485e-05,
-      "loss": 0.0315,
+      "loss": 0.0362,
       "step": 250
     },
     {
       "epoch": 4.55,
       "learning_rate": 3.120567653167127e-05,
-      "loss": 0.0374,
+      "loss": 0.0429,
       "step": 255
     },
     {
       "epoch": 4.64,
       "learning_rate": 3.0479807342533722e-05,
-      "loss": 0.0492,
+      "loss": 0.0534,
       "step": 260
     },
     {
       "epoch": 4.73,
       "learning_rate": 2.9749051615399637e-05,
-      "loss": 0.0336,
+      "loss": 0.0249,
       "step": 265
     },
     {
       "epoch": 4.82,
       "learning_rate": 2.9014060990993496e-05,
-      "loss": 0.0518,
+      "loss": 0.0339,
       "step": 270
     },
     {
       "epoch": 4.91,
       "learning_rate": 2.827549088644748e-05,
-      "loss": 0.0492,
+      "loss": 0.0484,
       "step": 275
     },
     {
       "epoch": 5.0,
       "learning_rate": 2.7533999910842766e-05,
-      "loss": 0.0454,
+      "loss": 0.0324,
       "step": 280
     },
     {
       "epoch": 5.0,
-      "eval_PREDICATE_f1": 0.9765625,
+      "eval_PREDICATE_f1": 0.96875,
       "eval_PREDICATE_number": 126,
-      "eval_PREDICATE_precision": 0.9615384615384616,
-      "eval_PREDICATE_recall": 0.9920634920634921,
-      "eval_loss": 0.15792229771614075,
-      "eval_overall_accuracy": 0.959866220735786,
-      "eval_overall_f1": 0.8537074148296593,
-      "eval_overall_precision": 0.841897233201581,
-      "eval_overall_recall": 0.8658536585365854,
-      "eval_runtime": 0.5032,
-      "eval_samples_per_second": 196.725,
-      "eval_steps_per_second": 25.833,
+      "eval_PREDICATE_precision": 0.9538461538461539,
+      "eval_PREDICATE_recall": 0.9841269841269841,
+      "eval_loss": 0.15630854666233063,
+      "eval_overall_accuracy": 0.9643255295429208,
+      "eval_overall_f1": 0.8702594810379242,
+      "eval_overall_precision": 0.8549019607843137,
+      "eval_overall_recall": 0.8861788617886179,
+      "eval_runtime": 0.5225,
+      "eval_samples_per_second": 189.481,
+      "eval_steps_per_second": 24.881,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7218045112781954,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7571428571428572,
       "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7868852459016393,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.6666666666666666,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8333333333333333,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 44,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7692307692307693,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.9090909090909091,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_f1": 0.0,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_number": 1,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_precision": 0.0,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_recall": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_f1": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_number": 3,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_precision": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_recall": 0.0,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7794117647058824,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.7361111111111112,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.7961165048543688,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7321428571428571,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8723404255319149,
       "step": 280
     },
     {
       "epoch": 5.09,
       "learning_rate": 2.6790249277904582e-05,
-      "loss": 0.0142,
+      "loss": 0.0201,
       "step": 285
     },
     {
       "epoch": 5.18,
       "learning_rate": 2.6044902216374496e-05,
-      "loss": 0.0234,
+      "loss": 0.0211,
       "step": 290
     },
     {
       "epoch": 5.27,
       "learning_rate": 2.5298623378585884e-05,
-      "loss": 0.0142,
+      "loss": 0.0145,
       "step": 295
     },
     {
       "epoch": 5.36,
       "learning_rate": 2.4552078247770007e-05,
-      "loss": 0.0325,
+      "loss": 0.0353,
       "step": 300
     },
     {
       "epoch": 5.45,
       "learning_rate": 2.3805932544621002e-05,
-      "loss": 0.0394,
+      "loss": 0.0428,
       "step": 305
     },
     {
       "epoch": 5.54,
       "learning_rate": 2.3060851633649245e-05,
-      "loss": 0.016,
+      "loss": 0.0153,
       "step": 310
     },
     {
       "epoch": 5.62,
       "learning_rate": 2.2317499929852273e-05,
-      "loss": 0.0354,
+      "loss": 0.0276,
       "step": 315
     },
     {
       "epoch": 5.71,
       "learning_rate": 2.1576540306232418e-05,
-      "loss": 0.0132,
+      "loss": 0.02,
       "step": 320
     },
     {
@@ -574,478 +554,458 @@
     {
       "epoch": 5.89,
       "learning_rate": 2.0104437536815883e-05,
-      "loss": 0.0343,
+      "loss": 0.0295,
       "step": 330
     },
     {
       "epoch": 5.98,
       "learning_rate": 1.9374607117118485e-05,
-      "loss": 0.0297,
+      "loss": 0.0361,
       "step": 335
     },
     {
       "epoch": 6.0,
-      "eval_PREDICATE_f1": 0.968503937007874,
+      "eval_PREDICATE_f1": 0.9727626459143969,
       "eval_PREDICATE_number": 126,
-      "eval_PREDICATE_precision": 0.9609375,
-      "eval_PREDICATE_recall": 0.9761904761904762,
-      "eval_loss": 0.16979572176933289,
-      "eval_overall_accuracy": 0.9587513935340022,
-      "eval_overall_f1": 0.8509803921568627,
-      "eval_overall_precision": 0.821969696969697,
-      "eval_overall_recall": 0.8821138211382114,
-      "eval_runtime": 0.5095,
-      "eval_samples_per_second": 194.318,
-      "eval_steps_per_second": 25.516,
+      "eval_PREDICATE_precision": 0.9541984732824428,
+      "eval_PREDICATE_recall": 0.9920634920634921,
+      "eval_loss": 0.17225639522075653,
+      "eval_overall_accuracy": 0.9620958751393534,
+      "eval_overall_f1": 0.8617234468937877,
+      "eval_overall_precision": 0.849802371541502,
+      "eval_overall_recall": 0.8739837398373984,
+      "eval_runtime": 0.518,
+      "eval_samples_per_second": 191.119,
+      "eval_steps_per_second": 25.096,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7083333333333334,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7101449275362319,
       "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7083333333333334,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.7083333333333334,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8453608247422681,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 44,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7735849056603774,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.9318181818181818,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_f1": 0.0,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_number": 1,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_precision": 0.0,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_recall": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_f1": 0.3333333333333333,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_number": 3,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_precision": 0.2222222222222222,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_recall": 0.6666666666666666,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7424242424242424,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.6805555555555556,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8367346938775511,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.803921568627451,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8723404255319149,
       "step": 336
     },
     {
       "epoch": 6.07,
       "learning_rate": 1.864979305919248e-05,
-      "loss": 0.0104,
+      "loss": 0.0098,
       "step": 340
     },
     {
       "epoch": 6.16,
       "learning_rate": 1.7930641705365836e-05,
-      "loss": 0.0247,
+      "loss": 0.0277,
       "step": 345
     },
     {
       "epoch": 6.25,
       "learning_rate": 1.721779434833299e-05,
-      "loss": 0.0066,
+      "loss": 0.0118,
       "step": 350
     },
     {
       "epoch": 6.34,
       "learning_rate": 1.651188665929151e-05,
-      "loss": 0.0246,
+      "loss": 0.0223,
       "step": 355
     },
     {
       "epoch": 6.43,
       "learning_rate": 1.5813548121091622e-05,
-      "loss": 0.017,
+      "loss": 0.0128,
       "step": 360
     },
     {
       "epoch": 6.52,
       "learning_rate": 1.512340146690402e-05,
-      "loss": 0.0117,
+      "loss": 0.0121,
       "step": 365
     },
     {
       "epoch": 6.61,
       "learning_rate": 1.4442062124906763e-05,
-      "loss": 0.0273,
+      "loss": 0.0243,
       "step": 370
     },
     {
       "epoch": 6.7,
       "learning_rate": 1.3770137669486171e-05,
-      "loss": 0.0214,
+      "loss": 0.0216,
       "step": 375
     },
     {
       "epoch": 6.79,
       "learning_rate": 1.3108227279441243e-05,
-      "loss": 0.0132,
+      "loss": 0.0093,
       "step": 380
     },
     {
       "epoch": 6.88,
       "learning_rate": 1.2456921203674715e-05,
-      "loss": 0.0276,
+      "loss": 0.0208,
       "step": 385
     },
     {
       "epoch": 6.96,
       "learning_rate": 1.1816800234847303e-05,
-      "loss": 0.0161,
+      "loss": 0.0074,
       "step": 390
     },
     {
       "epoch": 7.0,
-      "eval_PREDICATE_f1": 0.9725490196078431,
+      "eval_PREDICATE_f1": 0.9727626459143969,
       "eval_PREDICATE_number": 126,
-      "eval_PREDICATE_precision": 0.9612403100775194,
-      "eval_PREDICATE_recall": 0.9841269841269841,
-      "eval_loss": 0.1633814424276352,
-      "eval_overall_accuracy": 0.9637681159420289,
-      "eval_overall_f1": 0.8674698795180722,
-      "eval_overall_precision": 0.8571428571428571,
-      "eval_overall_recall": 0.8780487804878049,
-      "eval_runtime": 0.5173,
-      "eval_samples_per_second": 191.382,
-      "eval_steps_per_second": 25.131,
+      "eval_PREDICATE_precision": 0.9541984732824428,
+      "eval_PREDICATE_recall": 0.9920634920634921,
+      "eval_loss": 0.19959960877895355,
+      "eval_overall_accuracy": 0.959866220735786,
+      "eval_overall_f1": 0.8560157790927021,
+      "eval_overall_precision": 0.8314176245210728,
+      "eval_overall_recall": 0.8821138211382114,
+      "eval_runtime": 0.5027,
+      "eval_samples_per_second": 196.942,
+      "eval_steps_per_second": 25.861,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7338129496402879,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.723404255319149,
       "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7611940298507462,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7391304347826086,
       "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.7083333333333334,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8695652173913043,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 44,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.8333333333333334,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.9090909090909091,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_f1": 0.0,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_number": 1,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_precision": 0.0,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_recall": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_f1": 0.2222222222222222,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_number": 3,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_precision": 0.16666666666666666,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_recall": 0.3333333333333333,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.7884615384615385,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7192982456140351,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8723404255319149,
       "step": 392
     },
     {
       "epoch": 7.05,
       "learning_rate": 1.1188435191464303e-05,
-      "loss": 0.0116,
+      "loss": 0.0102,
       "step": 395
     },
     {
       "epoch": 7.14,
       "learning_rate": 1.0572386408856552e-05,
-      "loss": 0.0228,
+      "loss": 0.0193,
       "step": 400
     },
     {
       "epoch": 7.23,
       "learning_rate": 9.969203239509592e-06,
-      "loss": 0.0138,
+      "loss": 0.0082,
       "step": 405
     },
     {
       "epoch": 7.32,
       "learning_rate": 9.379423563186652e-06,
-      "loss": 0.009,
+      "loss": 0.0076,
       "step": 410
     },
     {
       "epoch": 7.41,
       "learning_rate": 8.803573307282209e-06,
-      "loss": 0.0096,
+      "loss": 0.0192,
       "step": 415
     },
     {
       "epoch": 7.5,
       "learning_rate": 8.242165977833974e-06,
-      "loss": 0.0071,
+      "loss": 0.0064,
       "step": 420
     },
     {
       "epoch": 7.59,
       "learning_rate": 7.695702201611338e-06,
-      "loss": 0.0164,
+      "loss": 0.0206,
       "step": 425
     },
     {
       "epoch": 7.68,
       "learning_rate": 7.1646692796888456e-06,
-      "loss": 0.0073,
+      "loss": 0.0041,
       "step": 430
     },
     {
       "epoch": 7.77,
       "learning_rate": 6.649540752902544e-06,
-      "loss": 0.0065,
+      "loss": 0.0039,
       "step": 435
     },
     {
       "epoch": 7.86,
       "learning_rate": 6.150775979576906e-06,
-      "loss": 0.0291,
+      "loss": 0.0236,
       "step": 440
     },
     {
       "epoch": 7.95,
       "learning_rate": 5.668819725898747e-06,
-      "loss": 0.0128,
+      "loss": 0.0073,
       "step": 445
     },
     {
       "epoch": 8.0,
-      "eval_PREDICATE_f1": 0.9689922480620156,
+      "eval_PREDICATE_f1": 0.9727626459143969,
       "eval_PREDICATE_number": 126,
-      "eval_PREDICATE_precision": 0.946969696969697,
+      "eval_PREDICATE_precision": 0.9541984732824428,
       "eval_PREDICATE_recall": 0.9920634920634921,
-      "eval_loss": 0.18599779903888702,
-      "eval_overall_accuracy": 0.9587513935340022,
-      "eval_overall_f1": 0.853228962818004,
-      "eval_overall_precision": 0.8226415094339623,
+      "eval_loss": 0.21042802929878235,
+      "eval_overall_accuracy": 0.9609810479375697,
+      "eval_overall_f1": 0.8599605522682447,
+      "eval_overall_precision": 0.8352490421455939,
       "eval_overall_recall": 0.8861788617886179,
-      "eval_runtime": 0.4889,
-      "eval_samples_per_second": 202.488,
-      "eval_steps_per_second": 26.589,
+      "eval_runtime": 0.5148,
+      "eval_samples_per_second": 192.299,
+      "eval_steps_per_second": 25.251,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7222222222222222,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7272727272727272,
       "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7222222222222222,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7323943661971831,
       "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.7222222222222222,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8817204301075268,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 44,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.8367346938775511,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.9318181818181818,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_f1": 0.0,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_number": 1,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_precision": 0.0,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_recall": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_f1": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_number": 3,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_precision": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_recall": 0.0,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8118811881188118,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7592592592592593,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8723404255319149,
       "step": 448
     },
     {
       "epoch": 8.04,
       "learning_rate": 5.204101769303474e-06,
-      "loss": 0.0195,
+      "loss": 0.014,
       "step": 450
     },
     {
       "epoch": 8.12,
       "learning_rate": 4.757036515227381e-06,
-      "loss": 0.018,
+      "loss": 0.0087,
       "step": 455
     },
     {
       "epoch": 8.21,
       "learning_rate": 4.328022627567657e-06,
-      "loss": 0.0124,
+      "loss": 0.0103,
       "step": 460
     },
     {
       "epoch": 8.3,
       "learning_rate": 3.91744267317968e-06,
-      "loss": 0.0071,
+      "loss": 0.0068,
       "step": 465
     },
     {
       "epoch": 8.39,
       "learning_rate": 3.5256627807286085e-06,
-      "loss": 0.0189,
+      "loss": 0.0155,
       "step": 470
     },
     {
       "epoch": 8.48,
       "learning_rate": 3.1530323141995553e-06,
-      "loss": 0.0053,
+      "loss": 0.0046,
       "step": 475
     },
     {
       "epoch": 8.57,
       "learning_rate": 2.7998835613573137e-06,
-      "loss": 0.0147,
+      "loss": 0.0171,
       "step": 480
     },
     {
       "epoch": 8.66,
       "learning_rate": 2.4665314374336335e-06,
-      "loss": 0.0111,
+      "loss": 0.0091,
       "step": 485
     },
     {
       "epoch": 8.75,
       "learning_rate": 2.153273204306153e-06,
-      "loss": 0.0043,
+      "loss": 0.0032,
       "step": 490
     },
     {
       "epoch": 8.84,
       "learning_rate": 1.8603882054194915e-06,
-      "loss": 0.0123,
+      "loss": 0.0146,
       "step": 495
     },
     {
       "epoch": 8.93,
       "learning_rate": 1.5881376166848149e-06,
-      "loss": 0.0064,
+      "loss": 0.0081,
       "step": 500
     },
     {
       "epoch": 9.0,
-      "eval_PREDICATE_f1": 0.9649805447470817,
+      "eval_PREDICATE_f1": 0.9727626459143969,
       "eval_PREDICATE_number": 126,
-      "eval_PREDICATE_precision": 0.9465648854961832,
-      "eval_PREDICATE_recall": 0.9841269841269841,
-      "eval_loss": 0.18625301122665405,
-      "eval_overall_accuracy": 0.9576365663322185,
-      "eval_overall_f1": 0.8493150684931506,
-      "eval_overall_precision": 0.8188679245283019,
-      "eval_overall_recall": 0.8821138211382114,
-      "eval_runtime": 0.5207,
-      "eval_samples_per_second": 190.14,
-      "eval_steps_per_second": 24.968,
+      "eval_PREDICATE_precision": 0.9541984732824428,
+      "eval_PREDICATE_recall": 0.9920634920634921,
+      "eval_loss": 0.20735101401805878,
+      "eval_overall_accuracy": 0.9620958751393534,
+      "eval_overall_f1": 0.8622754491017964,
+      "eval_overall_precision": 0.8470588235294118,
+      "eval_overall_recall": 0.8780487804878049,
+      "eval_runtime": 0.5139,
+      "eval_samples_per_second": 192.63,
+      "eval_steps_per_second": 25.295,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7123287671232876,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7194244604316546,
       "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7027027027027027,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.7222222222222222,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8817204301075268,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 44,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.8367346938775511,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.9318181818181818,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_f1": 0.0,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_number": 1,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_precision": 0.0,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_recall": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_f1": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_number": 3,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_precision": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_recall": 0.0,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.746268656716418,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.6944444444444444,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8200000000000001,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7735849056603774,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8723404255319149,
       "step": 504
     },
     {
       "epoch": 9.02,
       "learning_rate": 1.3367642135800329e-06,
-      "loss": 0.0049,
+      "loss": 0.0039,
       "step": 505
     },
     {
       "epoch": 9.11,
       "learning_rate": 1.106492154658323e-06,
-      "loss": 0.0062,
+      "loss": 0.0094,
       "step": 510
     },
     {
       "epoch": 9.2,
       "learning_rate": 8.975267816580201e-07,
-      "loss": 0.0139,
+      "loss": 0.007,
       "step": 515
     },
     {
       "epoch": 9.29,
       "learning_rate": 7.100544363921324e-07,
-      "loss": 0.0062,
+      "loss": 0.0045,
       "step": 520
     },
     {
       "epoch": 9.38,
       "learning_rate": 5.442422945807462e-07,
-      "loss": 0.026,
+      "loss": 0.0275,
       "step": 525
     },
     {
       "epoch": 9.46,
       "learning_rate": 4.002382167745428e-07,
-      "loss": 0.0046,
+      "loss": 0.0052,
       "step": 530
     },
     {
       "epoch": 9.55,
       "learning_rate": 2.781706165022879e-07,
-      "loss": 0.0073,
+      "loss": 0.008,
       "step": 535
     },
     {
       "epoch": 9.64,
       "learning_rate": 1.7814834575997363e-07,
-      "loss": 0.018,
+      "loss": 0.0151,
       "step": 540
     },
     {
       "epoch": 9.73,
       "learning_rate": 1.0026059794363551e-07,
-      "loss": 0.0058,
+      "loss": 0.0048,
       "step": 545
     },
     {
       "epoch": 9.82,
       "learning_rate": 4.457682831244259e-08,
-      "loss": 0.0058,
+      "loss": 0.0038,
       "step": 550
     },
     {
       "epoch": 9.91,
       "learning_rate": 1.114669205298513e-08,
-      "loss": 0.0108,
+      "loss": 0.0072,
       "step": 555
     },
     {
       "epoch": 10.0,
       "learning_rate": 0.0,
-      "loss": 0.0155,
+      "loss": 0.0126,
       "step": 560
     },
     {
       "epoch": 10.0,
-      "eval_PREDICATE_f1": 0.9649805447470817,
+      "eval_PREDICATE_f1": 0.9727626459143969,
       "eval_PREDICATE_number": 126,
-      "eval_PREDICATE_precision": 0.9465648854961832,
-      "eval_PREDICATE_recall": 0.9841269841269841,
-      "eval_loss": 0.18749289214611053,
-      "eval_overall_accuracy": 0.9576365663322185,
-      "eval_overall_f1": 0.8493150684931506,
-      "eval_overall_precision": 0.8188679245283019,
-      "eval_overall_recall": 0.8821138211382114,
-      "eval_runtime": 0.4979,
-      "eval_samples_per_second": 198.83,
-      "eval_steps_per_second": 26.109,
+      "eval_PREDICATE_precision": 0.9541984732824428,
+      "eval_PREDICATE_recall": 0.9920634920634921,
+      "eval_loss": 0.20896820724010468,
+      "eval_overall_accuracy": 0.9609810479375697,
+      "eval_overall_f1": 0.8588469184890657,
+      "eval_overall_precision": 0.8404669260700389,
+      "eval_overall_recall": 0.8780487804878049,
+      "eval_runtime": 0.4971,
+      "eval_samples_per_second": 199.154,
+      "eval_steps_per_second": 26.152,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_f1": 0.0,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_number": 1,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_precision": 0.0,
+      "eval_\u0414\u0420\u0423\u0413\u041e\u0415_recall": 0.0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_f1": 0.0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_number": 0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_precision": 0.0,
       "eval_\u0418\u041d\u0421\u0422\u0420\u0423\u041c\u0415\u041d\u0422_recall": 0.0,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7123287671232876,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_f1": 0.7142857142857144,
       "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_number": 72,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7027027027027027,
-      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.7222222222222222,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8817204301075268,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 44,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.8367346938775511,
-      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.9318181818181818,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_f1": 0.0,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_number": 1,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_precision": 0.0,
-      "eval_\u0434\u0440\u0443\u0433\u043e\u0435_recall": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_f1": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_number": 3,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_precision": 0.0,
-      "eval_\u044d\u043a\u0441\u043f\u0435\u0440\u0438\u0435\u043d\u0446\u0435\u0440_recall": 0.0,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_precision": 0.7352941176470589,
+      "eval_\u041a\u0410\u0423\u0417\u0410\u0422\u041e\u0420_recall": 0.6944444444444444,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_f1": 0.8200000000000001,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_number": 47,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_precision": 0.7735849056603774,
+      "eval_\u042d\u041a\u0421\u041f\u0415\u0420\u0418\u0415\u041d\u0426\u0415\u0420_recall": 0.8723404255319149,
       "step": 560
     },
     {
       "epoch": 10.0,
       "step": 560,
-      "total_flos": 132577707663864.0,
-      "train_loss": 0.13130194339568593,
-      "train_runtime": 174.9188,
-      "train_samples_per_second": 50.538,
-      "train_steps_per_second": 3.201
+      "total_flos": 132575310561672.0,
+      "train_loss": 0.11747416195326618,
+      "train_runtime": 191.5912,
+      "train_samples_per_second": 46.14,
+      "train_steps_per_second": 2.923
     }
   ],
   "max_steps": 560,
   "num_train_epochs": 10,
-  "total_flos": 132577707663864.0,
+  "total_flos": 132575310561672.0,
   "trial_name": null,
   "trial_params": null
 }