Training in progress, epoch 2

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6b7007f61a373deb00621503390180c0e6f3f5a8b37facf65d670c423f04bf45
+oid sha256:035ac4309d3064fbdb87084e018ecfb9b3d389199a08cd4e7d6b1b98837a5829
 size 437958648

run-0/checkpoint-1670/model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:db2a60dc6bbb07338c4d0dfff0424ac1bfa4a30608dbd8bd9055ff57493d0093
+oid sha256:035ac4309d3064fbdb87084e018ecfb9b3d389199a08cd4e7d6b1b98837a5829
 size 437958648

run-0/checkpoint-1670/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:44763a5e2c5d3f7b655029eda2376b76307b1122b86d80084664b5b86917899f
+oid sha256:2227f041b070701f9b2e36be87e58d7d918a20662ec98ca096c3f3d84049b261
 size 876038394

run-0/checkpoint-1670/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:627ea2f7cd2aedb5e1b4e507ea04b02dd0820e238075ddd9d5b5edf37464fe6d
-size 14308
+oid sha256:6e9450ad04385d20fe34680dfade3c60745403e057ce0bf1f1a034d03735cabe
+size 14244

run-0/checkpoint-1670/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:366cda4d6512ba28553df7753df51ae469825d7fb82c5003880659bdb93dbc77
+oid sha256:b0a0124953825835361cdc6763f757f2b1d1a27a3f8fe88d27358245324a8d8a
 size 1064

run-0/checkpoint-1670/trainer_state.json

@@ -1,5 +1,5 @@
 {
-  "best_metric": 0.9820850808867585,
+  "best_metric": 0.9642899940083882,
   "best_model_checkpoint": "bert-base-uncased-finetuned-spam-real/run-0/checkpoint-1670",
   "epoch": 2.0,
   "eval_steps": 500,
@@ -10,56 +10,56 @@
   "log_history": [
     {
       "epoch": 0.6,
-      "grad_norm": 1.6855844259262085,
-      "learning_rate": 7.660219401214295e-06,
-      "loss": 0.2399,
+      "grad_norm": 0.30469822883605957,
+      "learning_rate": 3.878567975913952e-06,
+      "loss": 0.3154,
       "step": 500
     },
     {
       "epoch": 1.0,
-      "eval_accuracy": 0.9690832834032355,
-      "eval_loss": 0.1404273957014084,
-      "eval_runtime": 321.4031,
-      "eval_samples_per_second": 51.929,
-      "eval_steps_per_second": 3.248,
+      "eval_accuracy": 0.9488915518274416,
+      "eval_loss": 0.22129623591899872,
+      "eval_runtime": 335.9128,
+      "eval_samples_per_second": 49.686,
+      "eval_steps_per_second": 3.108,
       "step": 835
     },
     {
       "epoch": 1.2,
-      "grad_norm": 0.28391602635383606,
-      "learning_rate": 6.618012679960649e-06,
-      "loss": 0.1015,
+      "grad_norm": 0.06290946900844574,
+      "learning_rate": 2.2210602938994423e-06,
+      "loss": 0.1341,
       "step": 1000
     },
     {
       "epoch": 1.8,
-      "grad_norm": 0.011191217228770256,
-      "learning_rate": 5.575805958707004e-06,
-      "loss": 0.0775,
+      "grad_norm": 0.40462997555732727,
+      "learning_rate": 5.635526118849332e-07,
+      "loss": 0.1332,
       "step": 1500
     },
     {
       "epoch": 2.0,
-      "eval_accuracy": 0.9820850808867585,
-      "eval_loss": 0.08610038459300995,
-      "eval_runtime": 318.1892,
-      "eval_samples_per_second": 52.453,
-      "eval_steps_per_second": 3.281,
+      "eval_accuracy": 0.9642899940083882,
+      "eval_loss": 0.16416475176811218,
+      "eval_runtime": 335.3225,
+      "eval_samples_per_second": 49.773,
+      "eval_steps_per_second": 3.113,
       "step": 1670
     }
   ],
   "logging_steps": 500,
-  "max_steps": 4175,
+  "max_steps": 1670,
   "num_input_tokens_seen": 0,
-  "num_train_epochs": 5,
+  "num_train_epochs": 2,
   "save_steps": 500,
-  "total_flos": 3026276636534160.0,
-  "train_batch_size": 8,
+  "total_flos": 2178221399563560.0,
+  "train_batch_size": 4,
   "trial_name": null,
   "trial_params": {
-    "learning_rate": 8.70242612246794e-06,
-    "num_train_epochs": 5,
-    "per_device_train_batch_size": 8,
-    "seed": 20
+    "learning_rate": 5.536075657928461e-06,
+    "num_train_epochs": 2,
+    "per_device_train_batch_size": 4,
+    "seed": 30
   }
 }

run-0/checkpoint-1670/training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e099818249ddecf442df98b949821a6e1bc96f54b7f97f17dac7a7067b75e3b4
+oid sha256:e3cf64e30d88e54cc8c23684bd03a07683ebeb538447855e2dd3ae4132912700
 size 4920

runs/Apr07_21-55-44_2e7e2ed6349a/events.out.tfevents.1712526953.2e7e2ed6349a.1594.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d1c754fb390b048cd7352780ed6b5b6add8708074d691787374dabd5ce6af6a1
-size 14275
+oid sha256:549c8a8af752aa8d5c0a86a6c84a5173428b48aa013beaa85cea0f62ec42daa0
+size 15374