Training in progress, step 2862

run-16/checkpoint-2862/tokenizer.json

@@ -1,11 +1,6 @@
 {
   "version": "1.0",
-  "truncation": {
-    "direction": "Right",
-    "max_length": 512,
-    "strategy": "LongestFirst",
-    "stride": 0
-  },
+  "truncation": null,
   "padding": null,
   "added_tokens": [
     {

run-16/checkpoint-2862/trainer_state.json

@@ -19,9 +19,9 @@
       "epoch": 1.0,
       "eval_accuracy": 0.6696774193548387,
       "eval_loss": 0.40919938683509827,
-      "eval_runtime": 1.6156,
-      "eval_samples_per_second": 1918.812,
-      "eval_steps_per_second": 40.233,
+      "eval_runtime": 1.562,
+      "eval_samples_per_second": 1984.622,
+      "eval_steps_per_second": 41.613,
       "step": 318
     },
     {
@@ -35,9 +35,9 @@
       "epoch": 2.0,
       "eval_accuracy": 0.844516129032258,
       "eval_loss": 0.13953416049480438,
-      "eval_runtime": 1.4219,
-      "eval_samples_per_second": 2180.152,
-      "eval_steps_per_second": 45.713,
+      "eval_runtime": 1.3987,
+      "eval_samples_per_second": 2216.309,
+      "eval_steps_per_second": 46.471,
       "step": 636
     },
     {
@@ -51,9 +51,9 @@
       "epoch": 3.0,
       "eval_accuracy": 0.8964516129032258,
       "eval_loss": 0.0722731500864029,
-      "eval_runtime": 1.4266,
-      "eval_samples_per_second": 2172.941,
-      "eval_steps_per_second": 45.562,
+      "eval_runtime": 1.4106,
+      "eval_samples_per_second": 2197.617,
+      "eval_steps_per_second": 46.079,
       "step": 954
     },
     {
@@ -67,9 +67,9 @@
       "epoch": 4.0,
       "eval_accuracy": 0.9180645161290323,
       "eval_loss": 0.050948865711688995,
-      "eval_runtime": 1.6166,
-      "eval_samples_per_second": 1917.656,
-      "eval_steps_per_second": 40.209,
+      "eval_runtime": 1.604,
+      "eval_samples_per_second": 1932.713,
+      "eval_steps_per_second": 40.525,
       "step": 1272
     },
     {
@@ -83,9 +83,9 @@
       "epoch": 5.0,
       "eval_accuracy": 0.9232258064516129,
       "eval_loss": 0.042863838374614716,
-      "eval_runtime": 1.431,
-      "eval_samples_per_second": 2166.391,
-      "eval_steps_per_second": 45.424,
+      "eval_runtime": 1.6025,
+      "eval_samples_per_second": 1934.471,
+      "eval_steps_per_second": 40.561,
       "step": 1590
     },
     {
@@ -99,9 +99,9 @@
       "epoch": 6.0,
       "eval_accuracy": 0.9261290322580645,
       "eval_loss": 0.03817891329526901,
-      "eval_runtime": 1.7938,
-      "eval_samples_per_second": 1728.153,
-      "eval_steps_per_second": 36.235,
+      "eval_runtime": 1.6062,
+      "eval_samples_per_second": 1930.001,
+      "eval_steps_per_second": 40.468,
       "step": 1908
     },
     {
@@ -115,9 +115,9 @@
       "epoch": 7.0,
       "eval_accuracy": 0.9303225806451613,
       "eval_loss": 0.035540465265512466,
-      "eval_runtime": 1.6169,
-      "eval_samples_per_second": 1917.217,
-      "eval_steps_per_second": 40.2,
+      "eval_runtime": 1.4224,
+      "eval_samples_per_second": 2179.412,
+      "eval_steps_per_second": 45.697,
       "step": 2226
     },
     {
@@ -131,9 +131,9 @@
       "epoch": 8.0,
       "eval_accuracy": 0.932258064516129,
       "eval_loss": 0.034126147627830505,
-      "eval_runtime": 1.5609,
-      "eval_samples_per_second": 1985.971,
-      "eval_steps_per_second": 41.641,
+      "eval_runtime": 1.422,
+      "eval_samples_per_second": 2180.024,
+      "eval_steps_per_second": 45.71,
       "step": 2544
     },
     {
@@ -161,11 +161,11 @@
       "attributes": {}
     }
   },
-  "total_flos": 742988563029384.0,
+  "total_flos": 742838622388212.0,
   "train_batch_size": 48,
   "trial_name": null,
   "trial_params": {
-    "alpha": 0.45730450441108567,
+    "alpha": 0.30033087334188296,
     "num_train_epochs": 9,
     "temperature": 2
   }

run-16/checkpoint-2862/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3fc3aa49ba06431aba065e94edc972c0714cf8bbe8237b664cb1781900dce380
+oid sha256:5c86f3d5ef41ce4b0c9fef4ceb006e512b2745b2eac6815d060edaed99ccb914
 size 5368

runs/Dec21_12-52-13_P920/events.out.tfevents.1734762203.P920.912596.17

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4fa933ecfb408cd33395883e029415cbf11cb8445429c4738b511bd3c4c615e5
+size 17738