End of training

README.md

@@ -3,11 +3,23 @@ license: other
 base_model: Qwen/Qwen1.5-4B
 tags:
 - generated_from_trainer
+datasets:
+- tyzhu/lmind_hotpot_train8000_eval7405_v1_qa
 metrics:
 - accuracy
 model-index:
 - name: lmind_hotpot_train8000_eval7405_v1_qa_5e-5_lora2
-  results: []
+  results:
+  - task:
+      name: Causal Language Modeling
+      type: text-generation
+    dataset:
+      name: tyzhu/lmind_hotpot_train8000_eval7405_v1_qa
+      type: tyzhu/lmind_hotpot_train8000_eval7405_v1_qa
+    metrics:
+    - name: Accuracy
+      type: accuracy
+      value: 0.487015873015873
 library_name: peft
 ---
 
@@ -16,7 +28,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # lmind_hotpot_train8000_eval7405_v1_qa_5e-5_lora2
 
-This model is a fine-tuned version of [Qwen/Qwen1.5-4B](https://huggingface.co/Qwen/Qwen1.5-4B) on an unknown dataset.
+This model is a fine-tuned version of [Qwen/Qwen1.5-4B](https://huggingface.co/Qwen/Qwen1.5-4B) on the tyzhu/lmind_hotpot_train8000_eval7405_v1_qa dataset.
 It achieves the following results on the evaluation set:
 - Loss: 4.3509
 - Accuracy: 0.4870

all_results.json

@@ -1,15 +1,18 @@
 {
-    "epoch": 20.0,
-    "eval_accuracy": 0.5839240506329114,
-    "eval_loss": 3.229841947555542,
-    "eval_runtime": 4.8462,
+    "epoch": 50.0,
+    "eval_accuracy": 0.487015873015873,
+    "eval_exact_match": 13.0,
+    "eval_f1": 19.709444444444447,
+    "eval_loss": 4.3508992195129395,
+    "eval_runtime": 5.4895,
     "eval_samples": 500,
-    "eval_samples_per_second": 103.173,
-    "eval_steps_per_second": 13.0,
-    "perplexity": 25.275661775117126,
-    "train_loss": 0.8756309669494629,
-    "train_runtime": 3519.8357,
+    "eval_samples_per_second": 91.084,
+    "eval_steps_per_second": 11.477,
+    "perplexity": 77.54816440472433,
+    "total_flos": 3.233846492439511e+17,
+    "train_loss": 0.676274776916504,
+    "train_runtime": 26211.3515,
     "train_samples": 8000,
-    "train_samples_per_second": 45.457,
-    "train_steps_per_second": 1.421
+    "train_samples_per_second": 15.261,
+    "train_steps_per_second": 0.477
 }

eval_results.json

@@ -1,10 +1,12 @@
 {
-    "epoch": 20.0,
-    "eval_accuracy": 0.5839240506329114,
-    "eval_loss": 3.229841947555542,
-    "eval_runtime": 4.8462,
+    "epoch": 50.0,
+    "eval_accuracy": 0.487015873015873,
+    "eval_exact_match": 13.0,
+    "eval_f1": 19.709444444444447,
+    "eval_loss": 4.3508992195129395,
+    "eval_runtime": 5.4895,
     "eval_samples": 500,
-    "eval_samples_per_second": 103.173,
-    "eval_steps_per_second": 13.0,
-    "perplexity": 25.275661775117126
+    "eval_samples_per_second": 91.084,
+    "eval_steps_per_second": 11.477,
+    "perplexity": 77.54816440472433
 }

train_results.json

@@ -1,8 +1,9 @@
 {
-    "epoch": 20.0,
-    "train_loss": 0.8756309669494629,
-    "train_runtime": 3519.8357,
+    "epoch": 50.0,
+    "total_flos": 3.233846492439511e+17,
+    "train_loss": 0.676274776916504,
+    "train_runtime": 26211.3515,
     "train_samples": 8000,
-    "train_samples_per_second": 45.457,
-    "train_steps_per_second": 1.421
+    "train_samples_per_second": 15.261,
+    "train_steps_per_second": 0.477
 }

trainer_state.json

@@ -1,508 +1,1667 @@
 {
   "best_metric": null,
   "best_model_checkpoint": null,
-  "epoch": 20.0,
+  "epoch": 50.0,
   "eval_steps": 500,
-  "global_step": 5000,
+  "global_step": 12500,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
       "epoch": 0.4,
+      "grad_norm": 0.6251392960548401,
       "learning_rate": 5e-05,
-      "loss": 2.1914,
+      "loss": 2.427,
       "step": 100
     },
     {
       "epoch": 0.8,
+      "grad_norm": 0.6590011715888977,
       "learning_rate": 5e-05,
-      "loss": 1.798,
+      "loss": 2.2748,
       "step": 200
     },
     {
       "epoch": 1.0,
-      "eval_accuracy": 0.6067341772151899,
-      "eval_loss": 1.8213441371917725,
-      "eval_runtime": 4.582,
-      "eval_samples_per_second": 109.122,
-      "eval_steps_per_second": 13.749,
+      "eval_accuracy": 0.5142539682539683,
+      "eval_loss": 2.336690902709961,
+      "eval_runtime": 5.4733,
+      "eval_samples_per_second": 91.353,
+      "eval_steps_per_second": 11.51,
+      "step": 250
+    },
+    {
+      "epoch": 1.0,
+      "eval_exact_match": 16.8,
+      "eval_f1": 24.850246583481884,
       "step": 250
     },
     {
       "epoch": 1.2,
+      "grad_norm": 0.8194239139556885,
       "learning_rate": 5e-05,
-      "loss": 1.7534,
+      "loss": 2.2322,
       "step": 300
     },
     {
       "epoch": 1.6,
+      "grad_norm": 0.9867294430732727,
       "learning_rate": 5e-05,
-      "loss": 1.7163,
+      "loss": 2.1904,
       "step": 400
     },
     {
       "epoch": 2.0,
+      "grad_norm": 0.9660400152206421,
       "learning_rate": 5e-05,
-      "loss": 1.7,
+      "loss": 2.1657,
+      "step": 500
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.5168253968253969,
+      "eval_loss": 2.323688507080078,
+      "eval_runtime": 5.1683,
+      "eval_samples_per_second": 96.744,
+      "eval_steps_per_second": 12.19,
       "step": 500
     },
     {
       "epoch": 2.0,
-      "eval_accuracy": 0.6077215189873417,
-      "eval_loss": 1.8046445846557617,
-      "eval_runtime": 5.1239,
-      "eval_samples_per_second": 97.582,
-      "eval_steps_per_second": 12.295,
+      "eval_exact_match": 17.2,
+      "eval_f1": 24.620036075036083,
       "step": 500
     },
     {
       "epoch": 2.4,
+      "grad_norm": 1.281128168106079,
       "learning_rate": 5e-05,
-      "loss": 1.5854,
+      "loss": 2.0251,
       "step": 600
     },
     {
       "epoch": 2.8,
+      "grad_norm": 1.3365116119384766,
       "learning_rate": 5e-05,
-      "loss": 1.5869,
+      "loss": 2.0158,
       "step": 700
     },
     {
       "epoch": 3.0,
-      "eval_accuracy": 0.6071392405063291,
-      "eval_loss": 1.829284906387329,
-      "eval_runtime": 5.0393,
-      "eval_samples_per_second": 99.221,
-      "eval_steps_per_second": 12.502,
+      "eval_accuracy": 0.5156507936507937,
+      "eval_loss": 2.3560824394226074,
+      "eval_runtime": 5.4675,
+      "eval_samples_per_second": 91.449,
+      "eval_steps_per_second": 11.523,
+      "step": 750
+    },
+    {
+      "epoch": 3.0,
+      "eval_exact_match": 19.0,
+      "eval_f1": 27.242790688843332,
       "step": 750
     },
     {
       "epoch": 3.2,
+      "grad_norm": 1.7840828895568848,
       "learning_rate": 5e-05,
-      "loss": 1.5039,
+      "loss": 1.9394,
       "step": 800
     },
     {
       "epoch": 3.6,
+      "grad_norm": 1.8030190467834473,
       "learning_rate": 5e-05,
-      "loss": 1.4165,
+      "loss": 1.8502,
       "step": 900
     },
     {
       "epoch": 4.0,
+      "grad_norm": 1.9682222604751587,
       "learning_rate": 5e-05,
-      "loss": 1.4349,
+      "loss": 1.8682,
       "step": 1000
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.6042784810126582,
-      "eval_loss": 1.89736008644104,
-      "eval_runtime": 4.5584,
-      "eval_samples_per_second": 109.688,
-      "eval_steps_per_second": 13.821,
+      "eval_accuracy": 0.5131746031746032,
+      "eval_loss": 2.4071946144104004,
+      "eval_runtime": 5.2825,
+      "eval_samples_per_second": 94.653,
+      "eval_steps_per_second": 11.926,
+      "step": 1000
+    },
+    {
+      "epoch": 4.0,
+      "eval_exact_match": 17.6,
+      "eval_f1": 26.010988455988468,
       "step": 1000
     },
     {
       "epoch": 4.4,
+      "grad_norm": 2.138505220413208,
       "learning_rate": 5e-05,
-      "loss": 1.2624,
+      "loss": 1.6795,
       "step": 1100
     },
     {
       "epoch": 4.8,
+      "grad_norm": 2.3212926387786865,
       "learning_rate": 5e-05,
-      "loss": 1.3111,
+      "loss": 1.7036,
       "step": 1200
     },
     {
       "epoch": 5.0,
-      "eval_accuracy": 0.601493670886076,
-      "eval_loss": 1.9769378900527954,
-      "eval_runtime": 4.4063,
-      "eval_samples_per_second": 113.474,
-      "eval_steps_per_second": 14.298,
+      "eval_accuracy": 0.5099365079365079,
+      "eval_loss": 2.494770050048828,
+      "eval_runtime": 5.1799,
+      "eval_samples_per_second": 96.528,
+      "eval_steps_per_second": 12.162,
+      "step": 1250
+    },
+    {
+      "epoch": 5.0,
+      "eval_exact_match": 17.2,
+      "eval_f1": 25.84752525252526,
       "step": 1250
     },
     {
       "epoch": 5.2,
+      "grad_norm": 2.4878485202789307,
       "learning_rate": 5e-05,
-      "loss": 1.226,
+      "loss": 1.6305,
       "step": 1300
     },
     {
       "epoch": 5.6,
+      "grad_norm": 2.599329710006714,
       "learning_rate": 5e-05,
-      "loss": 1.1608,
+      "loss": 1.5477,
       "step": 1400
     },
     {
       "epoch": 6.0,
+      "grad_norm": 2.5542759895324707,
       "learning_rate": 5e-05,
-      "loss": 1.197,
+      "loss": 1.5708,
       "step": 1500
     },
     {
       "epoch": 6.0,
-      "eval_accuracy": 0.599240506329114,
-      "eval_loss": 2.0634872913360596,
-      "eval_runtime": 4.8833,
-      "eval_samples_per_second": 102.389,
-      "eval_steps_per_second": 12.901,
+      "eval_accuracy": 0.5065079365079365,
+      "eval_loss": 2.5797038078308105,
+      "eval_runtime": 5.6827,
+      "eval_samples_per_second": 87.986,
+      "eval_steps_per_second": 11.086,
+      "step": 1500
+    },
+    {
+      "epoch": 6.0,
+      "eval_exact_match": 17.0,
+      "eval_f1": 25.310858585858586,
       "step": 1500
     },
     {
       "epoch": 6.4,
+      "grad_norm": 3.1968846321105957,
       "learning_rate": 5e-05,
-      "loss": 1.0336,
+      "loss": 1.3896,
       "step": 1600
     },
     {
       "epoch": 6.8,
+      "grad_norm": 3.8637375831604004,
       "learning_rate": 5e-05,
-      "loss": 1.0729,
+      "loss": 1.421,
       "step": 1700
     },
     {
       "epoch": 7.0,
-      "eval_accuracy": 0.5975443037974684,
-      "eval_loss": 2.1523237228393555,
-      "eval_runtime": 4.8877,
-      "eval_samples_per_second": 102.297,
-      "eval_steps_per_second": 12.889,
+      "eval_accuracy": 0.5038412698412699,
+      "eval_loss": 2.6977317333221436,
+      "eval_runtime": 5.3312,
+      "eval_samples_per_second": 93.787,
+      "eval_steps_per_second": 11.817,
+      "step": 1750
+    },
+    {
+      "epoch": 7.0,
+      "eval_exact_match": 16.0,
+      "eval_f1": 23.99441558441559,
       "step": 1750
     },
     {
       "epoch": 7.2,
+      "grad_norm": 3.468459367752075,
       "learning_rate": 5e-05,
-      "loss": 1.0005,
+      "loss": 1.3366,
       "step": 1800
     },
     {
       "epoch": 7.6,
+      "grad_norm": 3.5727667808532715,
       "learning_rate": 5e-05,
-      "loss": 0.9437,
+      "loss": 1.2585,
       "step": 1900
     },
     {
       "epoch": 8.0,
+      "grad_norm": 3.508258819580078,
       "learning_rate": 5e-05,
-      "loss": 0.9833,
+      "loss": 1.296,
       "step": 2000
     },
     {
       "epoch": 8.0,
-      "eval_accuracy": 0.5947341772151898,
-      "eval_loss": 2.2639756202697754,
-      "eval_runtime": 4.3964,
-      "eval_samples_per_second": 113.73,
-      "eval_steps_per_second": 14.33,
+      "eval_accuracy": 0.5013333333333333,
+      "eval_loss": 2.786928653717041,
+      "eval_runtime": 5.8398,
+      "eval_samples_per_second": 85.62,
+      "eval_steps_per_second": 10.788,
+      "step": 2000
+    },
+    {
+      "epoch": 8.0,
+      "eval_exact_match": 15.4,
+      "eval_f1": 22.542871572871572,
       "step": 2000
     },
     {
       "epoch": 8.4,
+      "grad_norm": 4.042112350463867,
       "learning_rate": 5e-05,
-      "loss": 0.823,
+      "loss": 1.1084,
       "step": 2100
     },
     {
       "epoch": 8.8,
+      "grad_norm": 3.7548155784606934,
       "learning_rate": 5e-05,
-      "loss": 0.8672,
+      "loss": 1.1457,
       "step": 2200
     },
     {
       "epoch": 9.0,
-      "eval_accuracy": 0.5924050632911393,
-      "eval_loss": 2.3642578125,
-      "eval_runtime": 4.6579,
-      "eval_samples_per_second": 107.345,
-      "eval_steps_per_second": 13.526,
+      "eval_accuracy": 0.49692063492063493,
+      "eval_loss": 2.9215197563171387,
+      "eval_runtime": 5.1412,
+      "eval_samples_per_second": 97.254,
+      "eval_steps_per_second": 12.254,
+      "step": 2250
+    },
+    {
+      "epoch": 9.0,
+      "eval_exact_match": 15.0,
+      "eval_f1": 21.59087301587302,
       "step": 2250
     },
     {
       "epoch": 9.2,
+      "grad_norm": 4.0875773429870605,
       "learning_rate": 5e-05,
-      "loss": 0.8032,
+      "loss": 1.0677,
       "step": 2300
     },
     {
       "epoch": 9.6,
+      "grad_norm": 4.143901348114014,
       "learning_rate": 5e-05,
-      "loss": 0.7566,
+      "loss": 1.0028,
       "step": 2400
     },
     {
       "epoch": 10.0,
+      "grad_norm": 5.0514445304870605,
       "learning_rate": 5e-05,
-      "loss": 0.7883,
+      "loss": 1.0401,
       "step": 2500
     },
     {
       "epoch": 10.0,
-      "eval_accuracy": 0.5908101265822785,
-      "eval_loss": 2.4598007202148438,
-      "eval_runtime": 4.6476,
-      "eval_samples_per_second": 107.582,
-      "eval_steps_per_second": 13.555,
+      "eval_accuracy": 0.4953015873015873,
+      "eval_loss": 3.054328203201294,
+      "eval_runtime": 5.4195,
+      "eval_samples_per_second": 92.259,
+      "eval_steps_per_second": 11.625,
+      "step": 2500
+    },
+    {
+      "epoch": 10.0,
+      "eval_exact_match": 15.4,
+      "eval_f1": 21.505119880119878,
       "step": 2500
     },
     {
       "epoch": 10.4,
+      "grad_norm": 3.8740394115448,
       "learning_rate": 5e-05,
-      "loss": 0.6488,
+      "loss": 0.8595,
       "step": 2600
     },
     {
       "epoch": 10.8,
+      "grad_norm": 4.170004844665527,
       "learning_rate": 5e-05,
-      "loss": 0.6879,
+      "loss": 0.9186,
       "step": 2700
     },
     {
       "epoch": 11.0,
-      "eval_accuracy": 0.588987341772152,
-      "eval_loss": 2.5668857097625732,
-      "eval_runtime": 4.4085,
-      "eval_samples_per_second": 113.416,
-      "eval_steps_per_second": 14.29,
+      "eval_accuracy": 0.49298412698412697,
+      "eval_loss": 3.177030324935913,
+      "eval_runtime": 5.9501,
+      "eval_samples_per_second": 84.033,
+      "eval_steps_per_second": 10.588,
+      "step": 2750
+    },
+    {
+      "epoch": 11.0,
+      "eval_exact_match": 15.6,
+      "eval_f1": 20.912753357753356,
       "step": 2750
     },
     {
       "epoch": 11.2,
+      "grad_norm": 4.2533392906188965,
       "learning_rate": 5e-05,
-      "loss": 0.6502,
+      "loss": 0.8411,
       "step": 2800
     },
     {
       "epoch": 11.6,
+      "grad_norm": 4.165424346923828,
       "learning_rate": 5e-05,
-      "loss": 0.602,
+      "loss": 0.7934,
       "step": 2900
     },
     {
       "epoch": 12.0,
+      "grad_norm": 4.517414093017578,
       "learning_rate": 5e-05,
-      "loss": 0.6295,
+      "loss": 0.8386,
+      "step": 3000
+    },
+    {
+      "epoch": 12.0,
+      "eval_accuracy": 0.4916190476190476,
+      "eval_loss": 3.2663722038269043,
+      "eval_runtime": 5.1518,
+      "eval_samples_per_second": 97.053,
+      "eval_steps_per_second": 12.229,
       "step": 3000
     },
     {
       "epoch": 12.0,
-      "eval_accuracy": 0.5885316455696202,
-      "eval_loss": 2.700011730194092,
-      "eval_runtime": 4.8719,
-      "eval_samples_per_second": 102.63,
-      "eval_steps_per_second": 12.931,
+      "eval_exact_match": 15.2,
+      "eval_f1": 20.465463425463433,
       "step": 3000
     },
     {
       "epoch": 12.4,
+      "grad_norm": 4.059298515319824,
       "learning_rate": 5e-05,
-      "loss": 0.5228,
+      "loss": 0.6908,
       "step": 3100
     },
     {
       "epoch": 12.8,
+      "grad_norm": 3.4847490787506104,
       "learning_rate": 5e-05,
-      "loss": 0.5545,
+      "loss": 0.7326,
       "step": 3200
     },
     {
       "epoch": 13.0,
-      "eval_accuracy": 0.5850886075949367,
-      "eval_loss": 2.8281185626983643,
-      "eval_runtime": 4.4033,
-      "eval_samples_per_second": 113.551,
-      "eval_steps_per_second": 14.307,
+      "eval_accuracy": 0.49047619047619045,
+      "eval_loss": 3.38826584815979,
+      "eval_runtime": 5.2085,
+      "eval_samples_per_second": 95.997,
+      "eval_steps_per_second": 12.096,
+      "step": 3250
+    },
+    {
+      "epoch": 13.0,
+      "eval_exact_match": 14.0,
+      "eval_f1": 20.144243697478988,
       "step": 3250
     },
     {
       "epoch": 13.2,
+      "grad_norm": 4.07555627822876,
       "learning_rate": 5e-05,
-      "loss": 0.5244,
+      "loss": 0.6901,
       "step": 3300
     },
     {
       "epoch": 13.6,
+      "grad_norm": 3.918684482574463,
       "learning_rate": 5e-05,
-      "loss": 0.4924,
+      "loss": 0.6432,
       "step": 3400
     },
     {
       "epoch": 14.0,
+      "grad_norm": 4.3187384605407715,
       "learning_rate": 5e-05,
-      "loss": 0.5208,
+      "loss": 0.6762,
       "step": 3500
     },
     {
       "epoch": 14.0,
-      "eval_accuracy": 0.5852911392405064,
-      "eval_loss": 2.879397392272949,
-      "eval_runtime": 4.703,
-      "eval_samples_per_second": 106.315,
-      "eval_steps_per_second": 13.396,
+      "eval_accuracy": 0.4887936507936508,
+      "eval_loss": 3.466782808303833,
+      "eval_runtime": 5.5309,
+      "eval_samples_per_second": 90.402,
+      "eval_steps_per_second": 11.391,
+      "step": 3500
+    },
+    {
+      "epoch": 14.0,
+      "eval_exact_match": 14.6,
+      "eval_f1": 20.548629148629157,
       "step": 3500
     },
     {
       "epoch": 14.4,
+      "grad_norm": 3.270817995071411,
       "learning_rate": 5e-05,
-      "loss": 0.4405,
+      "loss": 0.5785,
       "step": 3600
     },
     {
       "epoch": 14.8,
+      "grad_norm": 3.882906436920166,
       "learning_rate": 5e-05,
-      "loss": 0.4679,
+      "loss": 0.6117,
       "step": 3700
     },
     {
       "epoch": 15.0,
-      "eval_accuracy": 0.586253164556962,
-      "eval_loss": 2.9183950424194336,
-      "eval_runtime": 4.5629,
-      "eval_samples_per_second": 109.579,
-      "eval_steps_per_second": 13.807,
+      "eval_accuracy": 0.48904761904761906,
+      "eval_loss": 3.5602962970733643,
+      "eval_runtime": 5.4424,
+      "eval_samples_per_second": 91.871,
+      "eval_steps_per_second": 11.576,
+      "step": 3750
+    },
+    {
+      "epoch": 15.0,
+      "eval_exact_match": 14.8,
+      "eval_f1": 20.941592997645635,
       "step": 3750
     },
     {
       "epoch": 15.2,
+      "grad_norm": 4.002102851867676,
       "learning_rate": 5e-05,
-      "loss": 0.4443,
+      "loss": 0.5725,
       "step": 3800
     },
     {
       "epoch": 15.6,
+      "grad_norm": 2.8693177700042725,
       "learning_rate": 5e-05,
-      "loss": 0.4235,
+      "loss": 0.5581,
       "step": 3900
     },
     {
       "epoch": 16.0,
+      "grad_norm": 4.3405256271362305,
       "learning_rate": 5e-05,
-      "loss": 0.4464,
+      "loss": 0.5729,
+      "step": 4000
+    },
+    {
+      "epoch": 16.0,
+      "eval_accuracy": 0.48742857142857143,
+      "eval_loss": 3.6583478450775146,
+      "eval_runtime": 5.1711,
+      "eval_samples_per_second": 96.692,
+      "eval_steps_per_second": 12.183,
       "step": 4000
     },
     {
       "epoch": 16.0,
-      "eval_accuracy": 0.5852151898734177,
-      "eval_loss": 3.0790698528289795,
-      "eval_runtime": 4.6021,
-      "eval_samples_per_second": 108.647,
-      "eval_steps_per_second": 13.689,
+      "eval_exact_match": 14.8,
+      "eval_f1": 20.59385852709383,
       "step": 4000
     },
     {
       "epoch": 16.4,
+      "grad_norm": 4.715936183929443,
       "learning_rate": 5e-05,
-      "loss": 0.3919,
+      "loss": 0.5028,
       "step": 4100
     },
     {
       "epoch": 16.8,
+      "grad_norm": 3.7355644702911377,
       "learning_rate": 5e-05,
-      "loss": 0.4136,
+      "loss": 0.5317,
       "step": 4200
     },
     {
       "epoch": 17.0,
-      "eval_accuracy": 0.5855696202531645,
-      "eval_loss": 3.0832109451293945,
-      "eval_runtime": 5.1254,
-      "eval_samples_per_second": 97.553,
-      "eval_steps_per_second": 12.292,
+      "eval_accuracy": 0.4873968253968254,
+      "eval_loss": 3.6984667778015137,
+      "eval_runtime": 5.4593,
+      "eval_samples_per_second": 91.587,
+      "eval_steps_per_second": 11.54,
+      "step": 4250
+    },
+    {
+      "epoch": 17.0,
+      "eval_exact_match": 13.8,
+      "eval_f1": 20.517505228159344,
       "step": 4250
     },
     {
       "epoch": 17.2,
+      "grad_norm": 2.7771668434143066,
       "learning_rate": 5e-05,
-      "loss": 0.3902,
+      "loss": 0.505,
       "step": 4300
     },
     {
       "epoch": 17.6,
+      "grad_norm": 3.6721012592315674,
       "learning_rate": 5e-05,
-      "loss": 0.3841,
+      "loss": 0.496,
       "step": 4400
     },
     {
       "epoch": 18.0,
+      "grad_norm": 4.477765083312988,
       "learning_rate": 5e-05,
-      "loss": 0.4021,
+      "loss": 0.5109,
       "step": 4500
     },
     {
       "epoch": 18.0,
-      "eval_accuracy": 0.5846835443037974,
-      "eval_loss": 3.0944228172302246,
-      "eval_runtime": 4.5552,
-      "eval_samples_per_second": 109.765,
-      "eval_steps_per_second": 13.83,
+      "eval_accuracy": 0.4871111111111111,
+      "eval_loss": 3.7672173976898193,
+      "eval_runtime": 5.1624,
+      "eval_samples_per_second": 96.855,
+      "eval_steps_per_second": 12.204,
+      "step": 4500
+    },
+    {
+      "epoch": 18.0,
+      "eval_exact_match": 13.2,
+      "eval_f1": 19.935008488243795,
       "step": 4500
     },
     {
       "epoch": 18.4,
+      "grad_norm": 2.88653302192688,
       "learning_rate": 5e-05,
-      "loss": 0.3577,
+      "loss": 0.4564,
       "step": 4600
     },
     {
       "epoch": 18.8,
+      "grad_norm": 2.8045480251312256,
       "learning_rate": 5e-05,
-      "loss": 0.3776,
+      "loss": 0.479,
       "step": 4700
     },
     {
       "epoch": 19.0,
-      "eval_accuracy": 0.5828354430379746,
-      "eval_loss": 3.2120306491851807,
-      "eval_runtime": 4.7023,
-      "eval_samples_per_second": 106.332,
-      "eval_steps_per_second": 13.398,
+      "eval_accuracy": 0.4865714285714286,
+      "eval_loss": 3.847074270248413,
+      "eval_runtime": 5.4643,
+      "eval_samples_per_second": 91.503,
+      "eval_steps_per_second": 11.529,
+      "step": 4750
+    },
+    {
+      "epoch": 19.0,
+      "eval_exact_match": 14.2,
+      "eval_f1": 20.744217407979033,
       "step": 4750
     },
     {
       "epoch": 19.2,
+      "grad_norm": 2.717168092727661,
       "learning_rate": 5e-05,
-      "loss": 0.3617,
+      "loss": 0.4612,
       "step": 4800
     },
     {
       "epoch": 19.6,
+      "grad_norm": 3.0051259994506836,
       "learning_rate": 5e-05,
-      "loss": 0.3575,
+      "loss": 0.453,
       "step": 4900
     },
     {
       "epoch": 20.0,
+      "grad_norm": 2.753119468688965,
       "learning_rate": 5e-05,
-      "loss": 0.373,
+      "loss": 0.4675,
       "step": 5000
     },
     {
       "epoch": 20.0,
-      "eval_accuracy": 0.5839240506329114,
-      "eval_loss": 3.229841947555542,
-      "eval_runtime": 4.7155,
-      "eval_samples_per_second": 106.033,
-      "eval_steps_per_second": 13.36,
+      "eval_accuracy": 0.4867936507936508,
+      "eval_loss": 3.914645195007324,
+      "eval_runtime": 5.9419,
+      "eval_samples_per_second": 84.148,
+      "eval_steps_per_second": 10.603,
       "step": 5000
     },
     {
       "epoch": 20.0,
-      "step": 5000,
-      "total_flos": 3.1967425075347456e+17,
-      "train_loss": 0.8756309669494629,
-      "train_runtime": 3519.8357,
-      "train_samples_per_second": 45.457,
-      "train_steps_per_second": 1.421
+      "eval_exact_match": 13.2,
+      "eval_f1": 20.16965067612128,
+      "step": 5000
+    },
+    {
+      "epoch": 20.4,
+      "grad_norm": 2.736196517944336,
+      "learning_rate": 5e-05,
+      "loss": 0.425,
+      "step": 5100
+    },
+    {
+      "epoch": 20.8,
+      "grad_norm": 3.410940647125244,
+      "learning_rate": 5e-05,
+      "loss": 0.448,
+      "step": 5200
+    },
+    {
+      "epoch": 21.0,
+      "eval_accuracy": 0.485968253968254,
+      "eval_loss": 3.955155849456787,
+      "eval_runtime": 5.1881,
+      "eval_samples_per_second": 96.375,
+      "eval_steps_per_second": 12.143,
+      "step": 5250
+    },
+    {
+      "epoch": 21.0,
+      "eval_exact_match": 13.4,
+      "eval_f1": 19.219269782505087,
+      "step": 5250
+    },
+    {
+      "epoch": 21.2,
+      "grad_norm": 3.2622647285461426,
+      "learning_rate": 5e-05,
+      "loss": 0.4306,
+      "step": 5300
+    },
+    {
+      "epoch": 21.6,
+      "grad_norm": 2.9486541748046875,
+      "learning_rate": 5e-05,
+      "loss": 0.4291,
+      "step": 5400
+    },
+    {
+      "epoch": 22.0,
+      "grad_norm": 3.2296924591064453,
+      "learning_rate": 5e-05,
+      "loss": 0.4389,
+      "step": 5500
+    },
+    {
+      "epoch": 22.0,
+      "eval_accuracy": 0.4855238095238095,
+      "eval_loss": 3.967400312423706,
+      "eval_runtime": 5.4481,
+      "eval_samples_per_second": 91.776,
+      "eval_steps_per_second": 11.564,
+      "step": 5500
+    },
+    {
+      "epoch": 22.0,
+      "eval_exact_match": 14.6,
+      "eval_f1": 20.702058565191567,
+      "step": 5500
+    },
+    {
+      "epoch": 22.4,
+      "grad_norm": 2.210437059402466,
+      "learning_rate": 5e-05,
+      "loss": 0.4047,
+      "step": 5600
+    },
+    {
+      "epoch": 22.8,
+      "grad_norm": 2.7681708335876465,
+      "learning_rate": 5e-05,
+      "loss": 0.4227,
+      "step": 5700
+    },
+    {
+      "epoch": 23.0,
+      "eval_accuracy": 0.48682539682539683,
+      "eval_loss": 4.001584053039551,
+      "eval_runtime": 5.9051,
+      "eval_samples_per_second": 84.673,
+      "eval_steps_per_second": 10.669,
+      "step": 5750
+    },
+    {
+      "epoch": 23.0,
+      "eval_exact_match": 14.6,
+      "eval_f1": 22.022229437229445,
+      "step": 5750
+    },
+    {
+      "epoch": 23.2,
+      "grad_norm": 2.4847607612609863,
+      "learning_rate": 5e-05,
+      "loss": 0.4088,
+      "step": 5800
+    },
+    {
+      "epoch": 23.6,
+      "grad_norm": 2.022418260574341,
+      "learning_rate": 5e-05,
+      "loss": 0.4045,
+      "step": 5900
+    },
+    {
+      "epoch": 24.0,
+      "grad_norm": 1.8691329956054688,
+      "learning_rate": 5e-05,
+      "loss": 0.4234,
+      "step": 6000
+    },
+    {
+      "epoch": 24.0,
+      "eval_accuracy": 0.4853650793650794,
+      "eval_loss": 4.042004585266113,
+      "eval_runtime": 5.5095,
+      "eval_samples_per_second": 90.752,
+      "eval_steps_per_second": 11.435,
+      "step": 6000
+    },
+    {
+      "epoch": 24.0,
+      "eval_exact_match": 13.8,
+      "eval_f1": 21.028606654783133,
+      "step": 6000
+    },
+    {
+      "epoch": 24.4,
+      "grad_norm": 2.199568748474121,
+      "learning_rate": 5e-05,
+      "loss": 0.3915,
+      "step": 6100
+    },
+    {
+      "epoch": 24.8,
+      "grad_norm": 2.636310338973999,
+      "learning_rate": 5e-05,
+      "loss": 0.4051,
+      "step": 6200
+    },
+    {
+      "epoch": 25.0,
+      "eval_accuracy": 0.4865714285714286,
+      "eval_loss": 4.044601917266846,
+      "eval_runtime": 5.1578,
+      "eval_samples_per_second": 96.941,
+      "eval_steps_per_second": 12.215,
+      "step": 6250
+    },
+    {
+      "epoch": 25.0,
+      "eval_exact_match": 12.8,
+      "eval_f1": 20.273670088081865,
+      "step": 6250
+    },
+    {
+      "epoch": 25.2,
+      "grad_norm": 2.0600600242614746,
+      "learning_rate": 5e-05,
+      "loss": 0.3952,
+      "step": 6300
+    },
+    {
+      "epoch": 25.6,
+      "grad_norm": 2.2660067081451416,
+      "learning_rate": 5e-05,
+      "loss": 0.3939,
+      "step": 6400
+    },
+    {
+      "epoch": 26.0,
+      "grad_norm": 2.225553512573242,
+      "learning_rate": 5e-05,
+      "loss": 0.4043,
+      "step": 6500
+    },
+    {
+      "epoch": 26.0,
+      "eval_accuracy": 0.4855238095238095,
+      "eval_loss": 4.121946334838867,
+      "eval_runtime": 6.1465,
+      "eval_samples_per_second": 81.347,
+      "eval_steps_per_second": 10.25,
+      "step": 6500
+    },
+    {
+      "epoch": 26.0,
+      "eval_exact_match": 12.8,
+      "eval_f1": 18.98795730413378,
+      "step": 6500
+    },
+    {
+      "epoch": 26.4,
+      "grad_norm": 1.9317594766616821,
+      "learning_rate": 5e-05,
+      "loss": 0.3786,
+      "step": 6600
+    },
+    {
+      "epoch": 26.8,
+      "grad_norm": 2.19458270072937,
+      "learning_rate": 5e-05,
+      "loss": 0.396,
+      "step": 6700
+    },
+    {
+      "epoch": 27.0,
+      "eval_accuracy": 0.48542857142857143,
+      "eval_loss": 4.126931190490723,
+      "eval_runtime": 5.1829,
+      "eval_samples_per_second": 96.472,
+      "eval_steps_per_second": 12.155,
+      "step": 6750
+    },
+    {
+      "epoch": 27.0,
+      "eval_exact_match": 13.6,
+      "eval_f1": 20.55248196248197,
+      "step": 6750
+    },
+    {
+      "epoch": 27.2,
+      "grad_norm": 1.6904568672180176,
+      "learning_rate": 5e-05,
+      "loss": 0.3843,
+      "step": 6800
+    },
+    {
+      "epoch": 27.6,
+      "grad_norm": 2.3401496410369873,
+      "learning_rate": 5e-05,
+      "loss": 0.3841,
+      "step": 6900
+    },
+    {
+      "epoch": 28.0,
+      "grad_norm": 2.264298439025879,
+      "learning_rate": 5e-05,
+      "loss": 0.3947,
+      "step": 7000
+    },
+    {
+      "epoch": 28.0,
+      "eval_accuracy": 0.48533333333333334,
+      "eval_loss": 4.117328643798828,
+      "eval_runtime": 5.4257,
+      "eval_samples_per_second": 92.154,
+      "eval_steps_per_second": 11.611,
+      "step": 7000
+    },
+    {
+      "epoch": 28.0,
+      "eval_exact_match": 13.2,
+      "eval_f1": 19.55831890331891,
+      "step": 7000
+    },
+    {
+      "epoch": 28.4,
+      "grad_norm": 1.7389506101608276,
+      "learning_rate": 5e-05,
+      "loss": 0.3732,
+      "step": 7100
+    },
+    {
+      "epoch": 28.8,
+      "grad_norm": 2.5369668006896973,
+      "learning_rate": 5e-05,
+      "loss": 0.383,
+      "step": 7200
+    },
+    {
+      "epoch": 29.0,
+      "eval_accuracy": 0.4855555555555556,
+      "eval_loss": 4.185734272003174,
+      "eval_runtime": 5.5139,
+      "eval_samples_per_second": 90.679,
+      "eval_steps_per_second": 11.426,
+      "step": 7250
+    },
+    {
+      "epoch": 29.0,
+      "eval_exact_match": 14.6,
+      "eval_f1": 20.956385281385288,
+      "step": 7250
+    },
+    {
+      "epoch": 29.2,
+      "grad_norm": 1.586300253868103,
+      "learning_rate": 5e-05,
+      "loss": 0.3743,
+      "step": 7300
+    },
+    {
+      "epoch": 29.6,
+      "grad_norm": 1.5583587884902954,
+      "learning_rate": 5e-05,
+      "loss": 0.3777,
+      "step": 7400
+    },
+    {
+      "epoch": 30.0,
+      "grad_norm": 1.5072301626205444,
+      "learning_rate": 5e-05,
+      "loss": 0.3866,
+      "step": 7500
+    },
+    {
+      "epoch": 30.0,
+      "eval_accuracy": 0.48784126984126985,
+      "eval_loss": 4.153143405914307,
+      "eval_runtime": 5.6131,
+      "eval_samples_per_second": 89.078,
+      "eval_steps_per_second": 11.224,
+      "step": 7500
+    },
+    {
+      "epoch": 30.0,
+      "eval_exact_match": 13.0,
+      "eval_f1": 19.762439945675247,
+      "step": 7500
+    },
+    {
+      "epoch": 30.4,
+      "grad_norm": 2.850844144821167,
+      "learning_rate": 5e-05,
+      "loss": 0.3636,
+      "step": 7600
+    },
+    {
+      "epoch": 30.8,
+      "grad_norm": 3.9435479640960693,
+      "learning_rate": 5e-05,
+      "loss": 0.3779,
+      "step": 7700
+    },
+    {
+      "epoch": 31.0,
+      "eval_accuracy": 0.48568253968253966,
+      "eval_loss": 4.214567184448242,
+      "eval_runtime": 5.4963,
+      "eval_samples_per_second": 90.971,
+      "eval_steps_per_second": 11.462,
+      "step": 7750
+    },
+    {
+      "epoch": 31.0,
+      "eval_exact_match": 13.6,
+      "eval_f1": 20.160829725829732,
+      "step": 7750
+    },
+    {
+      "epoch": 31.2,
+      "grad_norm": 1.4926341772079468,
+      "learning_rate": 5e-05,
+      "loss": 0.3721,
+      "step": 7800
+    },
+    {
+      "epoch": 31.6,
+      "grad_norm": 1.510084629058838,
+      "learning_rate": 5e-05,
+      "loss": 0.3731,
+      "step": 7900
+    },
+    {
+      "epoch": 32.0,
+      "grad_norm": 3.654505729675293,
+      "learning_rate": 5e-05,
+      "loss": 0.3785,
+      "step": 8000
+    },
+    {
+      "epoch": 32.0,
+      "eval_accuracy": 0.48565079365079367,
+      "eval_loss": 4.271883964538574,
+      "eval_runtime": 5.5273,
+      "eval_samples_per_second": 90.46,
+      "eval_steps_per_second": 11.398,
+      "step": 8000
+    },
+    {
+      "epoch": 32.0,
+      "eval_exact_match": 13.4,
+      "eval_f1": 19.94316450216451,
+      "step": 8000
+    },
+    {
+      "epoch": 32.4,
+      "grad_norm": 1.457706332206726,
+      "learning_rate": 5e-05,
+      "loss": 0.3574,
+      "step": 8100
+    },
+    {
+      "epoch": 32.8,
+      "grad_norm": 2.00523042678833,
+      "learning_rate": 5e-05,
+      "loss": 0.3764,
+      "step": 8200
+    },
+    {
+      "epoch": 33.0,
+      "eval_accuracy": 0.48765079365079367,
+      "eval_loss": 4.250087738037109,
+      "eval_runtime": 5.3196,
+      "eval_samples_per_second": 93.992,
+      "eval_steps_per_second": 11.843,
+      "step": 8250
+    },
+    {
+      "epoch": 33.0,
+      "eval_exact_match": 14.4,
+      "eval_f1": 20.652601842601847,
+      "step": 8250
+    },
+    {
+      "epoch": 33.2,
+      "grad_norm": 2.976921319961548,
+      "learning_rate": 5e-05,
+      "loss": 0.3656,
+      "step": 8300
+    },
+    {
+      "epoch": 33.6,
+      "grad_norm": 2.1466617584228516,
+      "learning_rate": 5e-05,
+      "loss": 0.3646,
+      "step": 8400
+    },
+    {
+      "epoch": 34.0,
+      "grad_norm": 1.3529678583145142,
+      "learning_rate": 5e-05,
+      "loss": 0.3756,
+      "step": 8500
+    },
+    {
+      "epoch": 34.0,
+      "eval_accuracy": 0.4862222222222222,
+      "eval_loss": 4.241517543792725,
+      "eval_runtime": 5.3044,
+      "eval_samples_per_second": 94.262,
+      "eval_steps_per_second": 11.877,
+      "step": 8500
+    },
+    {
+      "epoch": 34.0,
+      "eval_exact_match": 14.0,
+      "eval_f1": 20.943053613053618,
+      "step": 8500
+    },
+    {
+      "epoch": 34.4,
+      "grad_norm": 1.9960662126541138,
+      "learning_rate": 5e-05,
+      "loss": 0.3545,
+      "step": 8600
+    },
+    {
+      "epoch": 34.8,
+      "grad_norm": 1.7610321044921875,
+      "learning_rate": 5e-05,
+      "loss": 0.3687,
+      "step": 8700
+    },
+    {
+      "epoch": 35.0,
+      "eval_accuracy": 0.4872063492063492,
+      "eval_loss": 4.275779724121094,
+      "eval_runtime": 5.1886,
+      "eval_samples_per_second": 96.364,
+      "eval_steps_per_second": 12.142,
+      "step": 8750
+    },
+    {
+      "epoch": 35.0,
+      "eval_exact_match": 13.4,
+      "eval_f1": 19.964098287333588,
+      "step": 8750
+    },
+    {
+      "epoch": 35.2,
+      "grad_norm": 1.8763376474380493,
+      "learning_rate": 5e-05,
+      "loss": 0.3597,
+      "step": 8800
+    },
+    {
+      "epoch": 35.6,
+      "grad_norm": 3.3443212509155273,
+      "learning_rate": 5e-05,
+      "loss": 0.3611,
+      "step": 8900
+    },
+    {
+      "epoch": 36.0,
+      "grad_norm": 1.9623003005981445,
+      "learning_rate": 5e-05,
+      "loss": 0.3707,
+      "step": 9000
+    },
+    {
+      "epoch": 36.0,
+      "eval_accuracy": 0.48558730158730157,
+      "eval_loss": 4.269599437713623,
+      "eval_runtime": 5.3187,
+      "eval_samples_per_second": 94.007,
+      "eval_steps_per_second": 11.845,
+      "step": 9000
+    },
+    {
+      "epoch": 36.0,
+      "eval_exact_match": 14.0,
+      "eval_f1": 20.398275613275622,
+      "step": 9000
+    },
+    {
+      "epoch": 36.4,
+      "grad_norm": 1.66078782081604,
+      "learning_rate": 5e-05,
+      "loss": 0.3523,
+      "step": 9100
+    },
+    {
+      "epoch": 36.8,
+      "grad_norm": 1.3792146444320679,
+      "learning_rate": 5e-05,
+      "loss": 0.3657,
+      "step": 9200
+    },
+    {
+      "epoch": 37.0,
+      "eval_accuracy": 0.4855238095238095,
+      "eval_loss": 4.312060356140137,
+      "eval_runtime": 5.8808,
+      "eval_samples_per_second": 85.023,
+      "eval_steps_per_second": 10.713,
+      "step": 9250
+    },
+    {
+      "epoch": 37.0,
+      "eval_exact_match": 14.0,
+      "eval_f1": 20.292691197691205,
+      "step": 9250
+    },
+    {
+      "epoch": 37.2,
+      "grad_norm": 1.2422215938568115,
+      "learning_rate": 5e-05,
+      "loss": 0.3569,
+      "step": 9300
+    },
+    {
+      "epoch": 37.6,
+      "grad_norm": 2.2910890579223633,
+      "learning_rate": 5e-05,
+      "loss": 0.3581,
+      "step": 9400
+    },
+    {
+      "epoch": 38.0,
+      "grad_norm": 1.8451393842697144,
+      "learning_rate": 5e-05,
+      "loss": 0.3682,
+      "step": 9500
+    },
+    {
+      "epoch": 38.0,
+      "eval_accuracy": 0.48593650793650794,
+      "eval_loss": 4.3264689445495605,
+      "eval_runtime": 5.866,
+      "eval_samples_per_second": 85.237,
+      "eval_steps_per_second": 10.74,
+      "step": 9500
+    },
+    {
+      "epoch": 38.0,
+      "eval_exact_match": 13.8,
+      "eval_f1": 19.786422629657935,
+      "step": 9500
+    },
+    {
+      "epoch": 38.4,
+      "grad_norm": 1.5858269929885864,
+      "learning_rate": 5e-05,
+      "loss": 0.3487,
+      "step": 9600
+    },
+    {
+      "epoch": 38.8,
+      "grad_norm": 1.2035001516342163,
+      "learning_rate": 5e-05,
+      "loss": 0.3596,
+      "step": 9700
+    },
+    {
+      "epoch": 39.0,
+      "eval_accuracy": 0.4847619047619048,
+      "eval_loss": 4.394227027893066,
+      "eval_runtime": 5.543,
+      "eval_samples_per_second": 90.204,
+      "eval_steps_per_second": 11.366,
+      "step": 9750
+    },
+    {
+      "epoch": 39.0,
+      "eval_exact_match": 14.2,
+      "eval_f1": 20.00341991341992,
+      "step": 9750
+    },
+    {
+      "epoch": 39.2,
+      "grad_norm": 2.2688498497009277,
+      "learning_rate": 5e-05,
+      "loss": 0.3564,
+      "step": 9800
+    },
+    {
+      "epoch": 39.6,
+      "grad_norm": 1.2643145322799683,
+      "learning_rate": 5e-05,
+      "loss": 0.3545,
+      "step": 9900
+    },
+    {
+      "epoch": 40.0,
+      "grad_norm": 2.0884294509887695,
+      "learning_rate": 5e-05,
+      "loss": 0.3629,
+      "step": 10000
+    },
+    {
+      "epoch": 40.0,
+      "eval_accuracy": 0.4852063492063492,
+      "eval_loss": 4.303957462310791,
+      "eval_runtime": 5.1423,
+      "eval_samples_per_second": 97.233,
+      "eval_steps_per_second": 12.251,
+      "step": 10000
+    },
+    {
+      "epoch": 40.0,
+      "eval_exact_match": 13.0,
+      "eval_f1": 18.57473686444276,
+      "step": 10000
+    },
+    {
+      "epoch": 40.4,
+      "grad_norm": 1.535133719444275,
+      "learning_rate": 5e-05,
+      "loss": 0.3451,
+      "step": 10100
+    },
+    {
+      "epoch": 40.8,
+      "grad_norm": 1.328994870185852,
+      "learning_rate": 5e-05,
+      "loss": 0.3596,
+      "step": 10200
+    },
+    {
+      "epoch": 41.0,
+      "eval_accuracy": 0.48707936507936506,
+      "eval_loss": 4.302526473999023,
+      "eval_runtime": 5.2959,
+      "eval_samples_per_second": 94.412,
+      "eval_steps_per_second": 11.896,
+      "step": 10250
+    },
+    {
+      "epoch": 41.0,
+      "eval_exact_match": 12.4,
+      "eval_f1": 18.663619158619166,
+      "step": 10250
+    },
+    {
+      "epoch": 41.2,
+      "grad_norm": 1.0413349866867065,
+      "learning_rate": 5e-05,
+      "loss": 0.3504,
+      "step": 10300
+    },
+    {
+      "epoch": 41.6,
+      "grad_norm": 1.6248804330825806,
+      "learning_rate": 5e-05,
+      "loss": 0.3522,
+      "step": 10400
+    },
+    {
+      "epoch": 42.0,
+      "grad_norm": 1.2970424890518188,
+      "learning_rate": 5e-05,
+      "loss": 0.3611,
+      "step": 10500
+    },
+    {
+      "epoch": 42.0,
+      "eval_accuracy": 0.48568253968253966,
+      "eval_loss": 4.312370300292969,
+      "eval_runtime": 5.3427,
+      "eval_samples_per_second": 93.585,
+      "eval_steps_per_second": 11.792,
+      "step": 10500
+    },
+    {
+      "epoch": 42.0,
+      "eval_exact_match": 13.6,
+      "eval_f1": 19.535006390006394,
+      "step": 10500
+    },
+    {
+      "epoch": 42.4,
+      "grad_norm": 1.0228374004364014,
+      "learning_rate": 5e-05,
+      "loss": 0.3424,
+      "step": 10600
+    },
+    {
+      "epoch": 42.8,
+      "grad_norm": 1.2492530345916748,
+      "learning_rate": 5e-05,
+      "loss": 0.3552,
+      "step": 10700
+    },
+    {
+      "epoch": 43.0,
+      "eval_accuracy": 0.4858095238095238,
+      "eval_loss": 4.322521209716797,
+      "eval_runtime": 5.4428,
+      "eval_samples_per_second": 91.865,
+      "eval_steps_per_second": 11.575,
+      "step": 10750
+    },
+    {
+      "epoch": 43.0,
+      "eval_exact_match": 13.2,
+      "eval_f1": 19.179316934316944,
+      "step": 10750
+    },
+    {
+      "epoch": 43.2,
+      "grad_norm": 1.3973324298858643,
+      "learning_rate": 5e-05,
+      "loss": 0.3531,
+      "step": 10800
+    },
+    {
+      "epoch": 43.6,
+      "grad_norm": 1.3194059133529663,
+      "learning_rate": 5e-05,
+      "loss": 0.3489,
+      "step": 10900
+    },
+    {
+      "epoch": 44.0,
+      "grad_norm": 2.390622615814209,
+      "learning_rate": 5e-05,
+      "loss": 0.3593,
+      "step": 11000
+    },
+    {
+      "epoch": 44.0,
+      "eval_accuracy": 0.48438095238095236,
+      "eval_loss": 4.3125104904174805,
+      "eval_runtime": 5.8599,
+      "eval_samples_per_second": 85.326,
+      "eval_steps_per_second": 10.751,
+      "step": 11000
+    },
+    {
+      "epoch": 44.0,
+      "eval_exact_match": 14.2,
+      "eval_f1": 20.37479271479273,
+      "step": 11000
+    },
+    {
+      "epoch": 44.4,
+      "grad_norm": 2.320546865463257,
+      "learning_rate": 5e-05,
+      "loss": 0.3426,
+      "step": 11100
+    },
+    {
+      "epoch": 44.8,
+      "grad_norm": 1.38444185256958,
+      "learning_rate": 5e-05,
+      "loss": 0.3551,
+      "step": 11200
+    },
+    {
+      "epoch": 45.0,
+      "eval_accuracy": 0.4860952380952381,
+      "eval_loss": 4.3344855308532715,
+      "eval_runtime": 5.152,
+      "eval_samples_per_second": 97.05,
+      "eval_steps_per_second": 12.228,
+      "step": 11250
+    },
+    {
+      "epoch": 45.0,
+      "eval_exact_match": 14.4,
+      "eval_f1": 20.7534892884893,
+      "step": 11250
+    },
+    {
+      "epoch": 45.2,
+      "grad_norm": 2.1086132526397705,
+      "learning_rate": 5e-05,
+      "loss": 0.3451,
+      "step": 11300
+    },
+    {
+      "epoch": 45.6,
+      "grad_norm": 5.763082027435303,
+      "learning_rate": 5e-05,
+      "loss": 0.3512,
+      "step": 11400
+    },
+    {
+      "epoch": 46.0,
+      "grad_norm": 1.1683377027511597,
+      "learning_rate": 5e-05,
+      "loss": 0.3538,
+      "step": 11500
+    },
+    {
+      "epoch": 46.0,
+      "eval_accuracy": 0.48634920634920636,
+      "eval_loss": 4.335656642913818,
+      "eval_runtime": 5.1561,
+      "eval_samples_per_second": 96.973,
+      "eval_steps_per_second": 12.219,
+      "step": 11500
+    },
+    {
+      "epoch": 46.0,
+      "eval_exact_match": 12.6,
+      "eval_f1": 19.21051393051394,
+      "step": 11500
+    },
+    {
+      "epoch": 46.4,
+      "grad_norm": 1.80911123752594,
+      "learning_rate": 5e-05,
+      "loss": 0.3426,
+      "step": 11600
+    },
+    {
+      "epoch": 46.8,
+      "grad_norm": 1.8256689310073853,
+      "learning_rate": 5e-05,
+      "loss": 0.3497,
+      "step": 11700
+    },
+    {
+      "epoch": 47.0,
+      "eval_accuracy": 0.4864126984126984,
+      "eval_loss": 4.379979610443115,
+      "eval_runtime": 5.3692,
+      "eval_samples_per_second": 93.123,
+      "eval_steps_per_second": 11.733,
+      "step": 11750
+    },
+    {
+      "epoch": 47.0,
+      "eval_exact_match": 13.8,
+      "eval_f1": 20.139399652634953,
+      "step": 11750
+    },
+    {
+      "epoch": 47.2,
+      "grad_norm": 0.8401724696159363,
+      "learning_rate": 5e-05,
+      "loss": 0.3448,
+      "step": 11800
+    },
+    {
+      "epoch": 47.6,
+      "grad_norm": 1.1846015453338623,
+      "learning_rate": 5e-05,
+      "loss": 0.3485,
+      "step": 11900
+    },
+    {
+      "epoch": 48.0,
+      "grad_norm": 1.4425411224365234,
+      "learning_rate": 5e-05,
+      "loss": 0.3535,
+      "step": 12000
+    },
+    {
+      "epoch": 48.0,
+      "eval_accuracy": 0.4866984126984127,
+      "eval_loss": 4.32601261138916,
+      "eval_runtime": 5.1934,
+      "eval_samples_per_second": 96.276,
+      "eval_steps_per_second": 12.131,
+      "step": 12000
+    },
+    {
+      "epoch": 48.0,
+      "eval_exact_match": 12.8,
+      "eval_f1": 20.51256981580512,
+      "step": 12000
+    },
+    {
+      "epoch": 48.4,
+      "grad_norm": 2.4570260047912598,
+      "learning_rate": 5e-05,
+      "loss": 0.3369,
+      "step": 12100
+    },
+    {
+      "epoch": 48.8,
+      "grad_norm": 3.1862845420837402,
+      "learning_rate": 5e-05,
+      "loss": 0.3527,
+      "step": 12200
+    },
+    {
+      "epoch": 49.0,
+      "eval_accuracy": 0.4864126984126984,
+      "eval_loss": 4.437545299530029,
+      "eval_runtime": 5.4674,
+      "eval_samples_per_second": 91.452,
+      "eval_steps_per_second": 11.523,
+      "step": 12250
+    },
+    {
+      "epoch": 49.0,
+      "eval_exact_match": 13.0,
+      "eval_f1": 19.741555666555666,
+      "step": 12250
+    },
+    {
+      "epoch": 49.2,
+      "grad_norm": 2.3995940685272217,
+      "learning_rate": 5e-05,
+      "loss": 0.3447,
+      "step": 12300
+    },
+    {
+      "epoch": 49.6,
+      "grad_norm": 1.9548571109771729,
+      "learning_rate": 5e-05,
+      "loss": 0.3444,
+      "step": 12400
+    },
+    {
+      "epoch": 50.0,
+      "grad_norm": 0.9694028496742249,
+      "learning_rate": 5e-05,
+      "loss": 0.3514,
+      "step": 12500
+    },
+    {
+      "epoch": 50.0,
+      "eval_accuracy": 0.487015873015873,
+      "eval_loss": 4.3508992195129395,
+      "eval_runtime": 5.5789,
+      "eval_samples_per_second": 89.623,
+      "eval_steps_per_second": 11.292,
+      "step": 12500
+    },
+    {
+      "epoch": 50.0,
+      "eval_exact_match": 13.0,
+      "eval_f1": 19.709444444444447,
+      "step": 12500
+    },
+    {
+      "epoch": 50.0,
+      "step": 12500,
+      "total_flos": 3.233846492439511e+17,
+      "train_loss": 0.676274776916504,
+      "train_runtime": 26211.3515,
+      "train_samples_per_second": 15.261,
+      "train_steps_per_second": 0.477
     }
   ],
   "logging_steps": 100,
-  "max_steps": 5000,
-  "num_train_epochs": 20,
+  "max_steps": 12500,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 50,
   "save_steps": 500,
-  "total_flos": 3.1967425075347456e+17,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.233846492439511e+17,
+  "train_batch_size": 1,
   "trial_name": null,
   "trial_params": null
 }