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-4_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: 0.47844444444444445
 library_name: peft
 ---
 
@@ -16,7 +28,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 # lmind_hotpot_train8000_eval7405_v1_qa_5e-4_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: 4.0366
 - Accuracy: 0.4784

all_results.json

@@ -1,15 +1,18 @@
 {
-    "epoch": 20.0,
-    "eval_accuracy": 0.5813164556962025,
-    "eval_loss": 2.9419918060302734,
-    "eval_runtime": 5.0845,
+    "epoch": 50.0,
+    "eval_accuracy": 0.47844444444444445,
+    "eval_exact_match": 13.2,
+    "eval_f1": 20.598610833610845,
+    "eval_loss": 4.036568641662598,
+    "eval_runtime": 5.1492,
     "eval_samples": 500,
-    "eval_samples_per_second": 98.338,
-    "eval_steps_per_second": 12.391,
-    "perplexity": 18.953560556268226,
-    "train_loss": 0.5721206008911133,
-    "train_runtime": 3939.3187,
+    "eval_samples_per_second": 97.103,
+    "eval_steps_per_second": 12.235,
+    "perplexity": 56.631685420913975,
+    "total_flos": 3.823469760432046e+17,
+    "train_loss": 0.49708512481689454,
+    "train_runtime": 15055.6405,
     "train_samples": 8000,
-    "train_samples_per_second": 40.616,
-    "train_steps_per_second": 1.269
+    "train_samples_per_second": 26.568,
+    "train_steps_per_second": 0.83
 }

eval_results.json

@@ -1,10 +1,12 @@
 {
-    "epoch": 20.0,
-    "eval_accuracy": 0.5813164556962025,
-    "eval_loss": 2.9419918060302734,
-    "eval_runtime": 5.0845,
+    "epoch": 50.0,
+    "eval_accuracy": 0.47844444444444445,
+    "eval_exact_match": 13.2,
+    "eval_f1": 20.598610833610845,
+    "eval_loss": 4.036568641662598,
+    "eval_runtime": 5.1492,
     "eval_samples": 500,
-    "eval_samples_per_second": 98.338,
-    "eval_steps_per_second": 12.391,
-    "perplexity": 18.953560556268226
+    "eval_samples_per_second": 97.103,
+    "eval_steps_per_second": 12.235,
+    "perplexity": 56.631685420913975
 }

train_results.json

@@ -1,8 +1,9 @@
 {
-    "epoch": 20.0,
-    "train_loss": 0.5721206008911133,
-    "train_runtime": 3939.3187,
+    "epoch": 50.0,
+    "total_flos": 3.823469760432046e+17,
+    "train_loss": 0.49708512481689454,
+    "train_runtime": 15055.6405,
     "train_samples": 8000,
-    "train_samples_per_second": 40.616,
-    "train_steps_per_second": 1.269
+    "train_samples_per_second": 26.568,
+    "train_steps_per_second": 0.83
 }

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": 0.5479037761688232,
       "learning_rate": 0.0005,
-      "loss": 2.3712,
+      "loss": 2.3029,
       "step": 100
     },
     {
       "epoch": 0.8,
+      "grad_norm": 0.5442906618118286,
       "learning_rate": 0.0005,
-      "loss": 1.8732,
+      "loss": 2.2398,
       "step": 200
     },
     {
       "epoch": 1.0,
-      "eval_accuracy": 0.593873417721519,
-      "eval_loss": 2.011091947555542,
-      "eval_runtime": 5.3852,
-      "eval_samples_per_second": 92.847,
-      "eval_steps_per_second": 11.699,
+      "eval_accuracy": 0.5162539682539683,
+      "eval_loss": 2.3236427307128906,
+      "eval_runtime": 5.266,
+      "eval_samples_per_second": 94.949,
+      "eval_steps_per_second": 11.964,
+      "step": 250
+    },
+    {
+      "epoch": 1.0,
+      "eval_exact_match": 17.6,
+      "eval_f1": 26.59450937950939,
       "step": 250
     },
     {
       "epoch": 1.2,
+      "grad_norm": 0.6918729543685913,
       "learning_rate": 0.0005,
-      "loss": 1.7329,
+      "loss": 2.0245,
       "step": 300
     },
     {
       "epoch": 1.6,
+      "grad_norm": 0.7950774431228638,
       "learning_rate": 0.0005,
-      "loss": 1.6233,
+      "loss": 1.8286,
       "step": 400
     },
     {
       "epoch": 2.0,
+      "grad_norm": 0.6913579106330872,
       "learning_rate": 0.0005,
-      "loss": 1.6142,
+      "loss": 1.8301,
+      "step": 500
+    },
+    {
+      "epoch": 2.0,
+      "eval_accuracy": 0.5124126984126984,
+      "eval_loss": 2.4219954013824463,
+      "eval_runtime": 5.2763,
+      "eval_samples_per_second": 94.763,
+      "eval_steps_per_second": 11.94,
       "step": 500
     },
     {
       "epoch": 2.0,
-      "eval_accuracy": 0.6051139240506329,
-      "eval_loss": 1.844284176826477,
-      "eval_runtime": 5.3165,
-      "eval_samples_per_second": 94.047,
-      "eval_steps_per_second": 11.85,
+      "eval_exact_match": 17.0,
+      "eval_f1": 24.808081363081367,
       "step": 500
     },
     {
       "epoch": 2.4,
+      "grad_norm": 0.9923084378242493,
       "learning_rate": 0.0005,
-      "loss": 1.1625,
+      "loss": 1.3144,
       "step": 600
     },
     {
       "epoch": 2.8,
+      "grad_norm": 1.0025588274002075,
       "learning_rate": 0.0005,
-      "loss": 1.206,
+      "loss": 1.3626,
       "step": 700
     },
     {
       "epoch": 3.0,
-      "eval_accuracy": 0.6007341772151898,
-      "eval_loss": 1.981817603111267,
-      "eval_runtime": 5.3466,
-      "eval_samples_per_second": 93.517,
-      "eval_steps_per_second": 11.783,
+      "eval_accuracy": 0.5061904761904762,
+      "eval_loss": 2.6152632236480713,
+      "eval_runtime": 5.2745,
+      "eval_samples_per_second": 94.795,
+      "eval_steps_per_second": 11.944,
+      "step": 750
+    },
+    {
+      "epoch": 3.0,
+      "eval_exact_match": 16.0,
+      "eval_f1": 24.234217449217454,
       "step": 750
     },
     {
       "epoch": 3.2,
+      "grad_norm": 1.0352345705032349,
       "learning_rate": 0.0005,
-      "loss": 1.0064,
+      "loss": 1.1681,
       "step": 800
     },
     {
       "epoch": 3.6,
+      "grad_norm": 1.1078394651412964,
       "learning_rate": 0.0005,
-      "loss": 0.8141,
+      "loss": 0.9584,
       "step": 900
     },
     {
       "epoch": 4.0,
+      "grad_norm": 1.14862060546875,
       "learning_rate": 0.0005,
-      "loss": 0.8693,
+      "loss": 1.0112,
       "step": 1000
     },
     {
       "epoch": 4.0,
-      "eval_accuracy": 0.594126582278481,
-      "eval_loss": 2.2100281715393066,
-      "eval_runtime": 4.739,
-      "eval_samples_per_second": 105.508,
-      "eval_steps_per_second": 13.294,
+      "eval_accuracy": 0.49974603174603177,
+      "eval_loss": 2.834880828857422,
+      "eval_runtime": 5.1472,
+      "eval_samples_per_second": 97.141,
+      "eval_steps_per_second": 12.24,
+      "step": 1000
+    },
+    {
+      "epoch": 4.0,
+      "eval_exact_match": 13.6,
+      "eval_f1": 20.6380957930958,
       "step": 1000
     },
     {
       "epoch": 4.4,
+      "grad_norm": 1.2231959104537964,
       "learning_rate": 0.0005,
-      "loss": 0.5477,
+      "loss": 0.6614,
       "step": 1100
     },
     {
       "epoch": 4.8,
+      "grad_norm": 1.1669212579727173,
       "learning_rate": 0.0005,
-      "loss": 0.6023,
+      "loss": 0.7198,
       "step": 1200
     },
     {
       "epoch": 5.0,
-      "eval_accuracy": 0.5910126582278481,
-      "eval_loss": 2.375615119934082,
-      "eval_runtime": 4.6068,
-      "eval_samples_per_second": 108.536,
-      "eval_steps_per_second": 13.676,
+      "eval_accuracy": 0.4963174603174603,
+      "eval_loss": 3.075563907623291,
+      "eval_runtime": 5.1584,
+      "eval_samples_per_second": 96.929,
+      "eval_steps_per_second": 12.213,
+      "step": 1250
+    },
+    {
+      "epoch": 5.0,
+      "eval_exact_match": 14.0,
+      "eval_f1": 20.88632090132091,
       "step": 1250
     },
     {
       "epoch": 5.2,
+      "grad_norm": 0.9650686979293823,
       "learning_rate": 0.0005,
-      "loss": 0.514,
+      "loss": 0.6299,
       "step": 1300
     },
     {
       "epoch": 5.6,
+      "grad_norm": 0.9634979963302612,
       "learning_rate": 0.0005,
-      "loss": 0.4479,
+      "loss": 0.5491,
       "step": 1400
     },
     {
       "epoch": 6.0,
+      "grad_norm": 1.0452007055282593,
       "learning_rate": 0.0005,
-      "loss": 0.4717,
+      "loss": 0.589,
       "step": 1500
     },
     {
       "epoch": 6.0,
-      "eval_accuracy": 0.5895696202531645,
-      "eval_loss": 2.542104482650757,
-      "eval_runtime": 4.662,
-      "eval_samples_per_second": 107.251,
-      "eval_steps_per_second": 13.514,
+      "eval_accuracy": 0.49434920634920637,
+      "eval_loss": 3.2339136600494385,
+      "eval_runtime": 5.1512,
+      "eval_samples_per_second": 97.064,
+      "eval_steps_per_second": 12.23,
+      "step": 1500
+    },
+    {
+      "epoch": 6.0,
+      "eval_exact_match": 15.0,
+      "eval_f1": 21.66983199806731,
       "step": 1500
     },
     {
       "epoch": 6.4,
+      "grad_norm": 1.0834842920303345,
       "learning_rate": 0.0005,
-      "loss": 0.3723,
+      "loss": 0.4679,
       "step": 1600
     },
     {
       "epoch": 6.8,
+      "grad_norm": 0.9649925231933594,
       "learning_rate": 0.0005,
-      "loss": 0.3938,
+      "loss": 0.4969,
       "step": 1700
     },
     {
       "epoch": 7.0,
-      "eval_accuracy": 0.5891139240506329,
-      "eval_loss": 2.658656597137451,
-      "eval_runtime": 4.9363,
-      "eval_samples_per_second": 101.29,
-      "eval_steps_per_second": 12.763,
+      "eval_accuracy": 0.4934920634920635,
+      "eval_loss": 3.34251070022583,
+      "eval_runtime": 5.2951,
+      "eval_samples_per_second": 94.426,
+      "eval_steps_per_second": 11.898,
+      "step": 1750
+    },
+    {
+      "epoch": 7.0,
+      "eval_exact_match": 13.8,
+      "eval_f1": 20.9043167943168,
       "step": 1750
     },
     {
       "epoch": 7.2,
+      "grad_norm": 0.8583429455757141,
       "learning_rate": 0.0005,
-      "loss": 0.3727,
+      "loss": 0.4737,
       "step": 1800
     },
     {
       "epoch": 7.6,
+      "grad_norm": 0.9595094919204712,
       "learning_rate": 0.0005,
-      "loss": 0.3538,
+      "loss": 0.4544,
       "step": 1900
     },
     {
       "epoch": 8.0,
+      "grad_norm": 0.9260773062705994,
       "learning_rate": 0.0005,
-      "loss": 0.3697,
+      "loss": 0.4786,
+      "step": 2000
+    },
+    {
+      "epoch": 8.0,
+      "eval_accuracy": 0.49244444444444446,
+      "eval_loss": 3.4198057651519775,
+      "eval_runtime": 5.8813,
+      "eval_samples_per_second": 85.015,
+      "eval_steps_per_second": 10.712,
       "step": 2000
     },
     {
       "epoch": 8.0,
-      "eval_accuracy": 0.5872911392405064,
-      "eval_loss": 2.7531518936157227,
-      "eval_runtime": 4.7177,
-      "eval_samples_per_second": 105.985,
-      "eval_steps_per_second": 13.354,
+      "eval_exact_match": 13.4,
+      "eval_f1": 20.923260628260632,
       "step": 2000
     },
     {
       "epoch": 8.4,
+      "grad_norm": 1.045862078666687,
       "learning_rate": 0.0005,
-      "loss": 0.3431,
+      "loss": 0.4114,
       "step": 2100
     },
     {
       "epoch": 8.8,
+      "grad_norm": 0.7850492000579834,
       "learning_rate": 0.0005,
-      "loss": 0.3617,
+      "loss": 0.4399,
       "step": 2200
     },
     {
       "epoch": 9.0,
-      "eval_accuracy": 0.5870126582278481,
-      "eval_loss": 2.7664403915405273,
-      "eval_runtime": 4.7999,
-      "eval_samples_per_second": 104.169,
-      "eval_steps_per_second": 13.125,
+      "eval_accuracy": 0.49114285714285716,
+      "eval_loss": 3.4694714546203613,
+      "eval_runtime": 5.945,
+      "eval_samples_per_second": 84.104,
+      "eval_steps_per_second": 10.597,
+      "step": 2250
+    },
+    {
+      "epoch": 9.0,
+      "eval_exact_match": 13.0,
+      "eval_f1": 20.096701018279973,
       "step": 2250
     },
     {
       "epoch": 9.2,
+      "grad_norm": 0.8059192895889282,
       "learning_rate": 0.0005,
-      "loss": 0.3424,
+      "loss": 0.4299,
       "step": 2300
     },
     {
       "epoch": 9.6,
+      "grad_norm": 0.9573262333869934,
       "learning_rate": 0.0005,
-      "loss": 0.3418,
+      "loss": 0.4169,
       "step": 2400
     },
     {
       "epoch": 10.0,
+      "grad_norm": 1.1837812662124634,
       "learning_rate": 0.0005,
-      "loss": 0.3607,
+      "loss": 0.4481,
       "step": 2500
     },
     {
       "epoch": 10.0,
-      "eval_accuracy": 0.5866835443037974,
-      "eval_loss": 2.8513779640197754,
-      "eval_runtime": 4.6699,
-      "eval_samples_per_second": 107.069,
-      "eval_steps_per_second": 13.491,
+      "eval_accuracy": 0.49133333333333334,
+      "eval_loss": 3.5352694988250732,
+      "eval_runtime": 5.1504,
+      "eval_samples_per_second": 97.08,
+      "eval_steps_per_second": 12.232,
+      "step": 2500
+    },
+    {
+      "epoch": 10.0,
+      "eval_exact_match": 14.0,
+      "eval_f1": 20.9187062937063,
       "step": 2500
     },
     {
       "epoch": 10.4,
+      "grad_norm": 1.1431084871292114,
       "learning_rate": 0.0005,
-      "loss": 0.339,
+      "loss": 0.3986,
       "step": 2600
     },
     {
       "epoch": 10.8,
+      "grad_norm": 1.0149098634719849,
       "learning_rate": 0.0005,
-      "loss": 0.3414,
+      "loss": 0.4166,
       "step": 2700
     },
     {
       "epoch": 11.0,
-      "eval_accuracy": 0.5860759493670886,
-      "eval_loss": 2.8931641578674316,
-      "eval_runtime": 4.8648,
-      "eval_samples_per_second": 102.78,
-      "eval_steps_per_second": 12.95,
+      "eval_accuracy": 0.4893968253968254,
+      "eval_loss": 3.493849992752075,
+      "eval_runtime": 5.9644,
+      "eval_samples_per_second": 83.831,
+      "eval_steps_per_second": 10.563,
+      "step": 2750
+    },
+    {
+      "epoch": 11.0,
+      "eval_exact_match": 14.0,
+      "eval_f1": 21.90111555111556,
       "step": 2750
     },
     {
       "epoch": 11.2,
+      "grad_norm": 0.666008710861206,
       "learning_rate": 0.0005,
-      "loss": 0.3302,
+      "loss": 0.4154,
       "step": 2800
     },
     {
       "epoch": 11.6,
+      "grad_norm": 0.8676968812942505,
       "learning_rate": 0.0005,
-      "loss": 0.329,
+      "loss": 0.4081,
       "step": 2900
     },
     {
       "epoch": 12.0,
+      "grad_norm": 0.8371489644050598,
       "learning_rate": 0.0005,
-      "loss": 0.3439,
+      "loss": 0.429,
       "step": 3000
     },
     {
       "epoch": 12.0,
-      "eval_accuracy": 0.5854683544303797,
-      "eval_loss": 2.9544754028320312,
-      "eval_runtime": 5.1699,
-      "eval_samples_per_second": 96.713,
-      "eval_steps_per_second": 12.186,
+      "eval_accuracy": 0.49057142857142855,
+      "eval_loss": 3.544997453689575,
+      "eval_runtime": 5.1712,
+      "eval_samples_per_second": 96.689,
+      "eval_steps_per_second": 12.183,
+      "step": 3000
+    },
+    {
+      "epoch": 12.0,
+      "eval_exact_match": 13.0,
+      "eval_f1": 20.847710622710633,
       "step": 3000
     },
     {
       "epoch": 12.4,
+      "grad_norm": 0.8826413750648499,
       "learning_rate": 0.0005,
-      "loss": 0.3182,
+      "loss": 0.3877,
       "step": 3100
     },
     {
       "epoch": 12.8,
+      "grad_norm": 0.9569880962371826,
       "learning_rate": 0.0005,
-      "loss": 0.335,
+      "loss": 0.4193,
       "step": 3200
     },
     {
       "epoch": 13.0,
-      "eval_accuracy": 0.5842784810126582,
-      "eval_loss": 2.899094343185425,
-      "eval_runtime": 4.8433,
-      "eval_samples_per_second": 103.236,
-      "eval_steps_per_second": 13.008,
+      "eval_accuracy": 0.4881587301587302,
+      "eval_loss": 3.5636065006256104,
+      "eval_runtime": 5.3238,
+      "eval_samples_per_second": 93.917,
+      "eval_steps_per_second": 11.834,
+      "step": 3250
+    },
+    {
+      "epoch": 13.0,
+      "eval_exact_match": 12.4,
+      "eval_f1": 19.792212787212797,
       "step": 3250
     },
     {
       "epoch": 13.2,
+      "grad_norm": 0.8115692734718323,
       "learning_rate": 0.0005,
-      "loss": 0.327,
+      "loss": 0.408,
       "step": 3300
     },
     {
       "epoch": 13.6,
+      "grad_norm": 0.9201633334159851,
       "learning_rate": 0.0005,
-      "loss": 0.3223,
+      "loss": 0.3962,
       "step": 3400
     },
     {
       "epoch": 14.0,
+      "grad_norm": 1.0557641983032227,
       "learning_rate": 0.0005,
-      "loss": 0.3391,
+      "loss": 0.4276,
       "step": 3500
     },
     {
       "epoch": 14.0,
-      "eval_accuracy": 0.5839746835443038,
-      "eval_loss": 2.879316806793213,
-      "eval_runtime": 4.9294,
-      "eval_samples_per_second": 101.433,
-      "eval_steps_per_second": 12.781,
+      "eval_accuracy": 0.48895238095238097,
+      "eval_loss": 3.5625526905059814,
+      "eval_runtime": 5.8685,
+      "eval_samples_per_second": 85.201,
+      "eval_steps_per_second": 10.735,
+      "step": 3500
+    },
+    {
+      "epoch": 14.0,
+      "eval_exact_match": 13.4,
+      "eval_f1": 20.559842934842944,
       "step": 3500
     },
     {
       "epoch": 14.4,
+      "grad_norm": 0.7237015962600708,
       "learning_rate": 0.0005,
-      "loss": 0.3128,
+      "loss": 0.3888,
       "step": 3600
     },
     {
       "epoch": 14.8,
+      "grad_norm": 0.8207970857620239,
       "learning_rate": 0.0005,
-      "loss": 0.328,
+      "loss": 0.4071,
       "step": 3700
     },
     {
       "epoch": 15.0,
-      "eval_accuracy": 0.5851392405063292,
-      "eval_loss": 2.8953680992126465,
-      "eval_runtime": 4.6969,
-      "eval_samples_per_second": 106.452,
-      "eval_steps_per_second": 13.413,
+      "eval_accuracy": 0.4883174603174603,
+      "eval_loss": 3.630866765975952,
+      "eval_runtime": 5.6988,
+      "eval_samples_per_second": 87.738,
+      "eval_steps_per_second": 11.055,
+      "step": 3750
+    },
+    {
+      "epoch": 15.0,
+      "eval_exact_match": 12.8,
+      "eval_f1": 19.927898212898214,
       "step": 3750
     },
     {
       "epoch": 15.2,
+      "grad_norm": 0.9946625232696533,
       "learning_rate": 0.0005,
-      "loss": 0.3233,
+      "loss": 0.3983,
       "step": 3800
     },
     {
       "epoch": 15.6,
+      "grad_norm": 1.109531283378601,
       "learning_rate": 0.0005,
-      "loss": 0.3216,
+      "loss": 0.3938,
       "step": 3900
     },
     {
       "epoch": 16.0,
+      "grad_norm": 0.8117135167121887,
       "learning_rate": 0.0005,
-      "loss": 0.3351,
+      "loss": 0.421,
       "step": 4000
     },
     {
       "epoch": 16.0,
-      "eval_accuracy": 0.5838481012658228,
-      "eval_loss": 2.913999080657959,
-      "eval_runtime": 4.7522,
-      "eval_samples_per_second": 105.214,
-      "eval_steps_per_second": 13.257,
+      "eval_accuracy": 0.48904761904761906,
+      "eval_loss": 3.58181095123291,
+      "eval_runtime": 5.1769,
+      "eval_samples_per_second": 96.582,
+      "eval_steps_per_second": 12.169,
+      "step": 4000
+    },
+    {
+      "epoch": 16.0,
+      "eval_exact_match": 13.2,
+      "eval_f1": 19.838007156242455,
       "step": 4000
     },
     {
       "epoch": 16.4,
+      "grad_norm": 0.8689247369766235,
       "learning_rate": 0.0005,
-      "loss": 0.3087,
+      "loss": 0.3803,
       "step": 4100
     },
     {
       "epoch": 16.8,
+      "grad_norm": 0.8414099812507629,
       "learning_rate": 0.0005,
-      "loss": 0.3252,
+      "loss": 0.4065,
       "step": 4200
     },
     {
       "epoch": 17.0,
-      "eval_accuracy": 0.5825063291139241,
-      "eval_loss": 2.929701089859009,
-      "eval_runtime": 4.8518,
-      "eval_samples_per_second": 103.055,
-      "eval_steps_per_second": 12.985,
+      "eval_accuracy": 0.4869206349206349,
+      "eval_loss": 3.6166887283325195,
+      "eval_runtime": 5.8573,
+      "eval_samples_per_second": 85.364,
+      "eval_steps_per_second": 10.756,
+      "step": 4250
+    },
+    {
+      "epoch": 17.0,
+      "eval_exact_match": 13.6,
+      "eval_f1": 21.07906926406927,
       "step": 4250
     },
     {
       "epoch": 17.2,
+      "grad_norm": 0.8533599972724915,
       "learning_rate": 0.0005,
-      "loss": 0.3152,
+      "loss": 0.402,
       "step": 4300
     },
     {
       "epoch": 17.6,
+      "grad_norm": 1.02438485622406,
       "learning_rate": 0.0005,
-      "loss": 0.3161,
+      "loss": 0.3916,
       "step": 4400
     },
     {
       "epoch": 18.0,
+      "grad_norm": 1.0211039781570435,
       "learning_rate": 0.0005,
-      "loss": 0.332,
+      "loss": 0.4188,
+      "step": 4500
+    },
+    {
+      "epoch": 18.0,
+      "eval_accuracy": 0.48568253968253966,
+      "eval_loss": 3.692577600479126,
+      "eval_runtime": 5.1508,
+      "eval_samples_per_second": 97.071,
+      "eval_steps_per_second": 12.231,
       "step": 4500
     },
     {
       "epoch": 18.0,
-      "eval_accuracy": 0.5833924050632912,
-      "eval_loss": 2.9811997413635254,
-      "eval_runtime": 5.341,
-      "eval_samples_per_second": 93.615,
-      "eval_steps_per_second": 11.796,
+      "eval_exact_match": 12.0,
+      "eval_f1": 19.268169053169053,
       "step": 4500
     },
     {
       "epoch": 18.4,
+      "grad_norm": 0.692489743232727,
       "learning_rate": 0.0005,
-      "loss": 0.3089,
+      "loss": 0.381,
       "step": 4600
     },
     {
       "epoch": 18.8,
+      "grad_norm": 1.098289966583252,
       "learning_rate": 0.0005,
-      "loss": 0.324,
+      "loss": 0.3994,
       "step": 4700
     },
     {
       "epoch": 19.0,
-      "eval_accuracy": 0.5807848101265822,
-      "eval_loss": 2.982343912124634,
-      "eval_runtime": 4.9018,
-      "eval_samples_per_second": 102.002,
-      "eval_steps_per_second": 12.852,
+      "eval_accuracy": 0.48625396825396827,
+      "eval_loss": 3.653291940689087,
+      "eval_runtime": 5.3607,
+      "eval_samples_per_second": 93.271,
+      "eval_steps_per_second": 11.752,
+      "step": 4750
+    },
+    {
+      "epoch": 19.0,
+      "eval_exact_match": 12.6,
+      "eval_f1": 19.967438672438682,
       "step": 4750
     },
     {
       "epoch": 19.2,
+      "grad_norm": 0.7502797842025757,
       "learning_rate": 0.0005,
-      "loss": 0.3169,
+      "loss": 0.3931,
       "step": 4800
     },
     {
       "epoch": 19.6,
+      "grad_norm": 1.0003832578659058,
       "learning_rate": 0.0005,
-      "loss": 0.3116,
+      "loss": 0.3889,
       "step": 4900
     },
     {
       "epoch": 20.0,
+      "grad_norm": 0.9703432321548462,
       "learning_rate": 0.0005,
-      "loss": 0.3329,
+      "loss": 0.4103,
       "step": 5000
     },
     {
       "epoch": 20.0,
-      "eval_accuracy": 0.5813164556962025,
-      "eval_loss": 2.9419918060302734,
-      "eval_runtime": 4.5961,
-      "eval_samples_per_second": 108.787,
-      "eval_steps_per_second": 13.707,
+      "eval_accuracy": 0.48644444444444446,
+      "eval_loss": 3.689058542251587,
+      "eval_runtime": 5.3154,
+      "eval_samples_per_second": 94.066,
+      "eval_steps_per_second": 11.852,
       "step": 5000
     },
     {
       "epoch": 20.0,
-      "step": 5000,
-      "total_flos": 3.1967425075347456e+17,
-      "train_loss": 0.5721206008911133,
-      "train_runtime": 3939.3187,
-      "train_samples_per_second": 40.616,
-      "train_steps_per_second": 1.269
+      "eval_exact_match": 13.2,
+      "eval_f1": 20.261983016983024,
+      "step": 5000
+    },
+    {
+      "epoch": 20.4,
+      "grad_norm": 0.8916403651237488,
+      "learning_rate": 0.0005,
+      "loss": 0.373,
+      "step": 5100
+    },
+    {
+      "epoch": 20.8,
+      "grad_norm": 0.8316965103149414,
+      "learning_rate": 0.0005,
+      "loss": 0.397,
+      "step": 5200
+    },
+    {
+      "epoch": 21.0,
+      "eval_accuracy": 0.48507936507936505,
+      "eval_loss": 3.6972897052764893,
+      "eval_runtime": 5.2392,
+      "eval_samples_per_second": 95.434,
+      "eval_steps_per_second": 12.025,
+      "step": 5250
+    },
+    {
+      "epoch": 21.0,
+      "eval_exact_match": 13.6,
+      "eval_f1": 20.467407592407604,
+      "step": 5250
+    },
+    {
+      "epoch": 21.2,
+      "grad_norm": 0.6775929927825928,
+      "learning_rate": 0.0005,
+      "loss": 0.389,
+      "step": 5300
+    },
+    {
+      "epoch": 21.6,
+      "grad_norm": 0.9392185211181641,
+      "learning_rate": 0.0005,
+      "loss": 0.3838,
+      "step": 5400
+    },
+    {
+      "epoch": 22.0,
+      "grad_norm": 0.7476337552070618,
+      "learning_rate": 0.0005,
+      "loss": 0.4118,
+      "step": 5500
+    },
+    {
+      "epoch": 22.0,
+      "eval_accuracy": 0.4859047619047619,
+      "eval_loss": 3.7213613986968994,
+      "eval_runtime": 5.4332,
+      "eval_samples_per_second": 92.027,
+      "eval_steps_per_second": 11.595,
+      "step": 5500
+    },
+    {
+      "epoch": 22.0,
+      "eval_exact_match": 14.0,
+      "eval_f1": 21.604624819624828,
+      "step": 5500
+    },
+    {
+      "epoch": 22.4,
+      "grad_norm": 0.8096778392791748,
+      "learning_rate": 0.0005,
+      "loss": 0.3739,
+      "step": 5600
+    },
+    {
+      "epoch": 22.8,
+      "grad_norm": 0.882375180721283,
+      "learning_rate": 0.0005,
+      "loss": 0.3944,
+      "step": 5700
+    },
+    {
+      "epoch": 23.0,
+      "eval_accuracy": 0.48514285714285715,
+      "eval_loss": 3.7193007469177246,
+      "eval_runtime": 5.1774,
+      "eval_samples_per_second": 96.574,
+      "eval_steps_per_second": 12.168,
+      "step": 5750
+    },
+    {
+      "epoch": 23.0,
+      "eval_exact_match": 14.2,
+      "eval_f1": 20.586290942287846,
+      "step": 5750
+    },
+    {
+      "epoch": 23.2,
+      "grad_norm": 0.7671297192573547,
+      "learning_rate": 0.0005,
+      "loss": 0.3844,
+      "step": 5800
+    },
+    {
+      "epoch": 23.6,
+      "grad_norm": 0.6806452870368958,
+      "learning_rate": 0.0005,
+      "loss": 0.3801,
+      "step": 5900
+    },
+    {
+      "epoch": 24.0,
+      "grad_norm": 0.9788285493850708,
+      "learning_rate": 0.0005,
+      "loss": 0.4036,
+      "step": 6000
+    },
+    {
+      "epoch": 24.0,
+      "eval_accuracy": 0.48453968253968255,
+      "eval_loss": 3.756743907928467,
+      "eval_runtime": 5.3317,
+      "eval_samples_per_second": 93.778,
+      "eval_steps_per_second": 11.816,
+      "step": 6000
+    },
+    {
+      "epoch": 24.0,
+      "eval_exact_match": 13.0,
+      "eval_f1": 20.329698242933542,
+      "step": 6000
+    },
+    {
+      "epoch": 24.4,
+      "grad_norm": 0.964042067527771,
+      "learning_rate": 0.0005,
+      "loss": 0.3682,
+      "step": 6100
+    },
+    {
+      "epoch": 24.8,
+      "grad_norm": 0.7888486981391907,
+      "learning_rate": 0.0005,
+      "loss": 0.3939,
+      "step": 6200
+    },
+    {
+      "epoch": 25.0,
+      "eval_accuracy": 0.4840634920634921,
+      "eval_loss": 3.7891385555267334,
+      "eval_runtime": 5.6106,
+      "eval_samples_per_second": 89.117,
+      "eval_steps_per_second": 11.229,
+      "step": 6250
+    },
+    {
+      "epoch": 25.0,
+      "eval_exact_match": 12.6,
+      "eval_f1": 19.69095659242718,
+      "step": 6250
+    },
+    {
+      "epoch": 25.2,
+      "grad_norm": 0.6044737100601196,
+      "learning_rate": 0.0005,
+      "loss": 0.3825,
+      "step": 6300
+    },
+    {
+      "epoch": 25.6,
+      "grad_norm": 0.7787346243858337,
+      "learning_rate": 0.0005,
+      "loss": 0.3788,
+      "step": 6400
+    },
+    {
+      "epoch": 26.0,
+      "grad_norm": 1.076758623123169,
+      "learning_rate": 0.0005,
+      "loss": 0.401,
+      "step": 6500
+    },
+    {
+      "epoch": 26.0,
+      "eval_accuracy": 0.4827936507936508,
+      "eval_loss": 3.7671151161193848,
+      "eval_runtime": 5.5024,
+      "eval_samples_per_second": 90.869,
+      "eval_steps_per_second": 11.449,
+      "step": 6500
+    },
+    {
+      "epoch": 26.0,
+      "eval_exact_match": 13.8,
+      "eval_f1": 20.179704347939637,
+      "step": 6500
+    },
+    {
+      "epoch": 26.4,
+      "grad_norm": 0.6967647075653076,
+      "learning_rate": 0.0005,
+      "loss": 0.3655,
+      "step": 6600
+    },
+    {
+      "epoch": 26.8,
+      "grad_norm": 0.8000121116638184,
+      "learning_rate": 0.0005,
+      "loss": 0.3871,
+      "step": 6700
+    },
+    {
+      "epoch": 27.0,
+      "eval_accuracy": 0.4834603174603175,
+      "eval_loss": 3.783790111541748,
+      "eval_runtime": 5.4864,
+      "eval_samples_per_second": 91.134,
+      "eval_steps_per_second": 11.483,
+      "step": 6750
+    },
+    {
+      "epoch": 27.0,
+      "eval_exact_match": 13.8,
+      "eval_f1": 20.906587301587308,
+      "step": 6750
+    },
+    {
+      "epoch": 27.2,
+      "grad_norm": 0.7237639427185059,
+      "learning_rate": 0.0005,
+      "loss": 0.3844,
+      "step": 6800
+    },
+    {
+      "epoch": 27.6,
+      "grad_norm": 0.8594484925270081,
+      "learning_rate": 0.0005,
+      "loss": 0.3789,
+      "step": 6900
+    },
+    {
+      "epoch": 28.0,
+      "grad_norm": 0.9502065777778625,
+      "learning_rate": 0.0005,
+      "loss": 0.4005,
+      "step": 7000
+    },
+    {
+      "epoch": 28.0,
+      "eval_accuracy": 0.48314285714285715,
+      "eval_loss": 3.8041446208953857,
+      "eval_runtime": 5.5728,
+      "eval_samples_per_second": 89.721,
+      "eval_steps_per_second": 11.305,
+      "step": 7000
+    },
+    {
+      "epoch": 28.0,
+      "eval_exact_match": 11.8,
+      "eval_f1": 19.040579518520694,
+      "step": 7000
+    },
+    {
+      "epoch": 28.4,
+      "grad_norm": 0.6732666492462158,
+      "learning_rate": 0.0005,
+      "loss": 0.3716,
+      "step": 7100
+    },
+    {
+      "epoch": 28.8,
+      "grad_norm": 1.0644075870513916,
+      "learning_rate": 0.0005,
+      "loss": 0.3854,
+      "step": 7200
+    },
+    {
+      "epoch": 29.0,
+      "eval_accuracy": 0.48298412698412696,
+      "eval_loss": 3.860269069671631,
+      "eval_runtime": 5.4184,
+      "eval_samples_per_second": 92.278,
+      "eval_steps_per_second": 11.627,
+      "step": 7250
+    },
+    {
+      "epoch": 29.0,
+      "eval_exact_match": 13.2,
+      "eval_f1": 20.254610389610388,
+      "step": 7250
+    },
+    {
+      "epoch": 29.2,
+      "grad_norm": 0.7752814292907715,
+      "learning_rate": 0.0005,
+      "loss": 0.3778,
+      "step": 7300
+    },
+    {
+      "epoch": 29.6,
+      "grad_norm": 1.0244282484054565,
+      "learning_rate": 0.0005,
+      "loss": 0.3747,
+      "step": 7400
+    },
+    {
+      "epoch": 30.0,
+      "grad_norm": 0.8350189328193665,
+      "learning_rate": 0.0005,
+      "loss": 0.3942,
+      "step": 7500
+    },
+    {
+      "epoch": 30.0,
+      "eval_accuracy": 0.48123809523809524,
+      "eval_loss": 3.8247478008270264,
+      "eval_runtime": 5.9809,
+      "eval_samples_per_second": 83.599,
+      "eval_steps_per_second": 10.533,
+      "step": 7500
+    },
+    {
+      "epoch": 30.0,
+      "eval_exact_match": 12.2,
+      "eval_f1": 19.892261627261625,
+      "step": 7500
+    },
+    {
+      "epoch": 30.4,
+      "grad_norm": 0.7081766724586487,
+      "learning_rate": 0.0005,
+      "loss": 0.3599,
+      "step": 7600
+    },
+    {
+      "epoch": 30.8,
+      "grad_norm": 0.8704819083213806,
+      "learning_rate": 0.0005,
+      "loss": 0.3837,
+      "step": 7700
+    },
+    {
+      "epoch": 31.0,
+      "eval_accuracy": 0.48149206349206347,
+      "eval_loss": 3.8496508598327637,
+      "eval_runtime": 5.1689,
+      "eval_samples_per_second": 96.733,
+      "eval_steps_per_second": 12.188,
+      "step": 7750
+    },
+    {
+      "epoch": 31.0,
+      "eval_exact_match": 11.6,
+      "eval_f1": 18.579704184704187,
+      "step": 7750
+    },
+    {
+      "epoch": 31.2,
+      "grad_norm": 0.9046170115470886,
+      "learning_rate": 0.0005,
+      "loss": 0.3845,
+      "step": 7800
+    },
+    {
+      "epoch": 31.6,
+      "grad_norm": 0.8757615685462952,
+      "learning_rate": 0.0005,
+      "loss": 0.378,
+      "step": 7900
+    },
+    {
+      "epoch": 32.0,
+      "grad_norm": 0.9979965090751648,
+      "learning_rate": 0.0005,
+      "loss": 0.3896,
+      "step": 8000
+    },
+    {
+      "epoch": 32.0,
+      "eval_accuracy": 0.4836190476190476,
+      "eval_loss": 3.870502471923828,
+      "eval_runtime": 5.4758,
+      "eval_samples_per_second": 91.31,
+      "eval_steps_per_second": 11.505,
+      "step": 8000
+    },
+    {
+      "epoch": 32.0,
+      "eval_exact_match": 13.2,
+      "eval_f1": 19.941825396825404,
+      "step": 8000
+    },
+    {
+      "epoch": 32.4,
+      "grad_norm": 0.8767104744911194,
+      "learning_rate": 0.0005,
+      "loss": 0.358,
+      "step": 8100
+    },
+    {
+      "epoch": 32.8,
+      "grad_norm": 0.7482908368110657,
+      "learning_rate": 0.0005,
+      "loss": 0.3817,
+      "step": 8200
+    },
+    {
+      "epoch": 33.0,
+      "eval_accuracy": 0.4818095238095238,
+      "eval_loss": 3.8642616271972656,
+      "eval_runtime": 5.1588,
+      "eval_samples_per_second": 96.922,
+      "eval_steps_per_second": 12.212,
+      "step": 8250
+    },
+    {
+      "epoch": 33.0,
+      "eval_exact_match": 13.6,
+      "eval_f1": 20.228078033078035,
+      "step": 8250
+    },
+    {
+      "epoch": 33.2,
+      "grad_norm": 0.738953709602356,
+      "learning_rate": 0.0005,
+      "loss": 0.3753,
+      "step": 8300
+    },
+    {
+      "epoch": 33.6,
+      "grad_norm": 0.80599045753479,
+      "learning_rate": 0.0005,
+      "loss": 0.37,
+      "step": 8400
+    },
+    {
+      "epoch": 34.0,
+      "grad_norm": 1.0467863082885742,
+      "learning_rate": 0.0005,
+      "loss": 0.3928,
+      "step": 8500
+    },
+    {
+      "epoch": 34.0,
+      "eval_accuracy": 0.4806666666666667,
+      "eval_loss": 3.9377706050872803,
+      "eval_runtime": 5.4713,
+      "eval_samples_per_second": 91.386,
+      "eval_steps_per_second": 11.515,
+      "step": 8500
+    },
+    {
+      "epoch": 34.0,
+      "eval_exact_match": 13.0,
+      "eval_f1": 20.646601731601738,
+      "step": 8500
+    },
+    {
+      "epoch": 34.4,
+      "grad_norm": 0.7660771608352661,
+      "learning_rate": 0.0005,
+      "loss": 0.3615,
+      "step": 8600
+    },
+    {
+      "epoch": 34.8,
+      "grad_norm": 0.9810505509376526,
+      "learning_rate": 0.0005,
+      "loss": 0.3839,
+      "step": 8700
+    },
+    {
+      "epoch": 35.0,
+      "eval_accuracy": 0.48095238095238096,
+      "eval_loss": 3.95415997505188,
+      "eval_runtime": 5.2713,
+      "eval_samples_per_second": 94.854,
+      "eval_steps_per_second": 11.952,
+      "step": 8750
+    },
+    {
+      "epoch": 35.0,
+      "eval_exact_match": 12.8,
+      "eval_f1": 19.906587301587308,
+      "step": 8750
+    },
+    {
+      "epoch": 35.2,
+      "grad_norm": 0.8827441930770874,
+      "learning_rate": 0.0005,
+      "loss": 0.377,
+      "step": 8800
+    },
+    {
+      "epoch": 35.6,
+      "grad_norm": 0.8283040523529053,
+      "learning_rate": 0.0005,
+      "loss": 0.3727,
+      "step": 8900
+    },
+    {
+      "epoch": 36.0,
+      "grad_norm": 0.9295199513435364,
+      "learning_rate": 0.0005,
+      "loss": 0.3942,
+      "step": 9000
+    },
+    {
+      "epoch": 36.0,
+      "eval_accuracy": 0.4806031746031746,
+      "eval_loss": 3.9250476360321045,
+      "eval_runtime": 5.2946,
+      "eval_samples_per_second": 94.436,
+      "eval_steps_per_second": 11.899,
+      "step": 9000
+    },
+    {
+      "epoch": 36.0,
+      "eval_exact_match": 12.4,
+      "eval_f1": 19.917929292929298,
+      "step": 9000
+    },
+    {
+      "epoch": 36.4,
+      "grad_norm": 0.7359641194343567,
+      "learning_rate": 0.0005,
+      "loss": 0.3621,
+      "step": 9100
+    },
+    {
+      "epoch": 36.8,
+      "grad_norm": 0.8008917570114136,
+      "learning_rate": 0.0005,
+      "loss": 0.381,
+      "step": 9200
+    },
+    {
+      "epoch": 37.0,
+      "eval_accuracy": 0.4791746031746032,
+      "eval_loss": 3.921963930130005,
+      "eval_runtime": 5.3326,
+      "eval_samples_per_second": 93.762,
+      "eval_steps_per_second": 11.814,
+      "step": 9250
+    },
+    {
+      "epoch": 37.0,
+      "eval_exact_match": 12.0,
+      "eval_f1": 18.638997113997128,
+      "step": 9250
+    },
+    {
+      "epoch": 37.2,
+      "grad_norm": 0.7444690465927124,
+      "learning_rate": 0.0005,
+      "loss": 0.3771,
+      "step": 9300
+    },
+    {
+      "epoch": 37.6,
+      "grad_norm": 0.7617835998535156,
+      "learning_rate": 0.0005,
+      "loss": 0.3739,
+      "step": 9400
+    },
+    {
+      "epoch": 38.0,
+      "grad_norm": 0.8619845509529114,
+      "learning_rate": 0.0005,
+      "loss": 0.3918,
+      "step": 9500
+    },
+    {
+      "epoch": 38.0,
+      "eval_accuracy": 0.4780952380952381,
+      "eval_loss": 3.9583866596221924,
+      "eval_runtime": 5.169,
+      "eval_samples_per_second": 96.73,
+      "eval_steps_per_second": 12.188,
+      "step": 9500
+    },
+    {
+      "epoch": 38.0,
+      "eval_exact_match": 12.4,
+      "eval_f1": 20.53709235209236,
+      "step": 9500
+    },
+    {
+      "epoch": 38.4,
+      "grad_norm": 0.7432407140731812,
+      "learning_rate": 0.0005,
+      "loss": 0.3646,
+      "step": 9600
+    },
+    {
+      "epoch": 38.8,
+      "grad_norm": 0.9143626093864441,
+      "learning_rate": 0.0005,
+      "loss": 0.3787,
+      "step": 9700
+    },
+    {
+      "epoch": 39.0,
+      "eval_accuracy": 0.47758730158730156,
+      "eval_loss": 3.9241392612457275,
+      "eval_runtime": 5.8692,
+      "eval_samples_per_second": 85.19,
+      "eval_steps_per_second": 10.734,
+      "step": 9750
+    },
+    {
+      "epoch": 39.0,
+      "eval_exact_match": 12.6,
+      "eval_f1": 20.360458985458987,
+      "step": 9750
+    },
+    {
+      "epoch": 39.2,
+      "grad_norm": 0.5365042090415955,
+      "learning_rate": 0.0005,
+      "loss": 0.375,
+      "step": 9800
+    },
+    {
+      "epoch": 39.6,
+      "grad_norm": 0.6992344856262207,
+      "learning_rate": 0.0005,
+      "loss": 0.3696,
+      "step": 9900
+    },
+    {
+      "epoch": 40.0,
+      "grad_norm": 0.9610856175422668,
+      "learning_rate": 0.0005,
+      "loss": 0.3897,
+      "step": 10000
+    },
+    {
+      "epoch": 40.0,
+      "eval_accuracy": 0.47733333333333333,
+      "eval_loss": 3.9434309005737305,
+      "eval_runtime": 5.8615,
+      "eval_samples_per_second": 85.302,
+      "eval_steps_per_second": 10.748,
+      "step": 10000
+    },
+    {
+      "epoch": 40.0,
+      "eval_exact_match": 12.0,
+      "eval_f1": 18.266425796425803,
+      "step": 10000
+    },
+    {
+      "epoch": 40.4,
+      "grad_norm": 0.822077214717865,
+      "learning_rate": 0.0005,
+      "loss": 0.3561,
+      "step": 10100
+    },
+    {
+      "epoch": 40.8,
+      "grad_norm": 0.9024446606636047,
+      "learning_rate": 0.0005,
+      "loss": 0.3786,
+      "step": 10200
+    },
+    {
+      "epoch": 41.0,
+      "eval_accuracy": 0.47933333333333333,
+      "eval_loss": 3.9410653114318848,
+      "eval_runtime": 5.3296,
+      "eval_samples_per_second": 93.816,
+      "eval_steps_per_second": 11.821,
+      "step": 10250
+    },
+    {
+      "epoch": 41.0,
+      "eval_exact_match": 11.4,
+      "eval_f1": 18.911030636030635,
+      "step": 10250
+    },
+    {
+      "epoch": 41.2,
+      "grad_norm": 0.9134598970413208,
+      "learning_rate": 0.0005,
+      "loss": 0.3703,
+      "step": 10300
+    },
+    {
+      "epoch": 41.6,
+      "grad_norm": 0.7853434085845947,
+      "learning_rate": 0.0005,
+      "loss": 0.3691,
+      "step": 10400
+    },
+    {
+      "epoch": 42.0,
+      "grad_norm": 0.829653799533844,
+      "learning_rate": 0.0005,
+      "loss": 0.3864,
+      "step": 10500
+    },
+    {
+      "epoch": 42.0,
+      "eval_accuracy": 0.4766031746031746,
+      "eval_loss": 3.9932987689971924,
+      "eval_runtime": 6.0241,
+      "eval_samples_per_second": 83.0,
+      "eval_steps_per_second": 10.458,
+      "step": 10500
+    },
+    {
+      "epoch": 42.0,
+      "eval_exact_match": 12.6,
+      "eval_f1": 19.408923462158764,
+      "step": 10500
+    },
+    {
+      "epoch": 42.4,
+      "grad_norm": 0.9247658252716064,
+      "learning_rate": 0.0005,
+      "loss": 0.3585,
+      "step": 10600
+    },
+    {
+      "epoch": 42.8,
+      "grad_norm": 0.9292989373207092,
+      "learning_rate": 0.0005,
+      "loss": 0.377,
+      "step": 10700
+    },
+    {
+      "epoch": 43.0,
+      "eval_accuracy": 0.47873015873015873,
+      "eval_loss": 4.001534461975098,
+      "eval_runtime": 5.5096,
+      "eval_samples_per_second": 90.75,
+      "eval_steps_per_second": 11.435,
+      "step": 10750
+    },
+    {
+      "epoch": 43.0,
+      "eval_exact_match": 11.2,
+      "eval_f1": 18.621556939792242,
+      "step": 10750
+    },
+    {
+      "epoch": 43.2,
+      "grad_norm": 0.746468186378479,
+      "learning_rate": 0.0005,
+      "loss": 0.3726,
+      "step": 10800
+    },
+    {
+      "epoch": 43.6,
+      "grad_norm": 0.8017551898956299,
+      "learning_rate": 0.0005,
+      "loss": 0.3685,
+      "step": 10900
+    },
+    {
+      "epoch": 44.0,
+      "grad_norm": 0.9033043384552002,
+      "learning_rate": 0.0005,
+      "loss": 0.3887,
+      "step": 11000
+    },
+    {
+      "epoch": 44.0,
+      "eval_accuracy": 0.4787619047619048,
+      "eval_loss": 3.9979331493377686,
+      "eval_runtime": 5.3381,
+      "eval_samples_per_second": 93.666,
+      "eval_steps_per_second": 11.802,
+      "step": 11000
+    },
+    {
+      "epoch": 44.0,
+      "eval_exact_match": 13.0,
+      "eval_f1": 19.826702741702753,
+      "step": 11000
+    },
+    {
+      "epoch": 44.4,
+      "grad_norm": 0.7681131958961487,
+      "learning_rate": 0.0005,
+      "loss": 0.3575,
+      "step": 11100
+    },
+    {
+      "epoch": 44.8,
+      "grad_norm": 1.1326820850372314,
+      "learning_rate": 0.0005,
+      "loss": 0.3805,
+      "step": 11200
+    },
+    {
+      "epoch": 45.0,
+      "eval_accuracy": 0.47958730158730156,
+      "eval_loss": 3.9764013290405273,
+      "eval_runtime": 5.4892,
+      "eval_samples_per_second": 91.089,
+      "eval_steps_per_second": 11.477,
+      "step": 11250
+    },
+    {
+      "epoch": 45.0,
+      "eval_exact_match": 13.8,
+      "eval_f1": 21.24646441793502,
+      "step": 11250
+    },
+    {
+      "epoch": 45.2,
+      "grad_norm": 0.871893048286438,
+      "learning_rate": 0.0005,
+      "loss": 0.3683,
+      "step": 11300
+    },
+    {
+      "epoch": 45.6,
+      "grad_norm": 0.915560781955719,
+      "learning_rate": 0.0005,
+      "loss": 0.3693,
+      "step": 11400
+    },
+    {
+      "epoch": 46.0,
+      "grad_norm": 0.8644348382949829,
+      "learning_rate": 0.0005,
+      "loss": 0.3827,
+      "step": 11500
+    },
+    {
+      "epoch": 46.0,
+      "eval_accuracy": 0.4785714285714286,
+      "eval_loss": 3.9989893436431885,
+      "eval_runtime": 5.3059,
+      "eval_samples_per_second": 94.234,
+      "eval_steps_per_second": 11.873,
+      "step": 11500
+    },
+    {
+      "epoch": 46.0,
+      "eval_exact_match": 13.2,
+      "eval_f1": 20.00336940836941,
+      "step": 11500
+    },
+    {
+      "epoch": 46.4,
+      "grad_norm": 0.8179894089698792,
+      "learning_rate": 0.0005,
+      "loss": 0.3601,
+      "step": 11600
+    },
+    {
+      "epoch": 46.8,
+      "grad_norm": 0.7599920034408569,
+      "learning_rate": 0.0005,
+      "loss": 0.3737,
+      "step": 11700
+    },
+    {
+      "epoch": 47.0,
+      "eval_accuracy": 0.4791746031746032,
+      "eval_loss": 4.005940914154053,
+      "eval_runtime": 5.3285,
+      "eval_samples_per_second": 93.835,
+      "eval_steps_per_second": 11.823,
+      "step": 11750
+    },
+    {
+      "epoch": 47.0,
+      "eval_exact_match": 13.2,
+      "eval_f1": 20.48666555666556,
+      "step": 11750
+    },
+    {
+      "epoch": 47.2,
+      "grad_norm": 0.6046980619430542,
+      "learning_rate": 0.0005,
+      "loss": 0.3667,
+      "step": 11800
+    },
+    {
+      "epoch": 47.6,
+      "grad_norm": 0.7555102109909058,
+      "learning_rate": 0.0005,
+      "loss": 0.3681,
+      "step": 11900
+    },
+    {
+      "epoch": 48.0,
+      "grad_norm": 0.9966368675231934,
+      "learning_rate": 0.0005,
+      "loss": 0.3807,
+      "step": 12000
+    },
+    {
+      "epoch": 48.0,
+      "eval_accuracy": 0.47984126984126985,
+      "eval_loss": 4.0745978355407715,
+      "eval_runtime": 5.1625,
+      "eval_samples_per_second": 96.853,
+      "eval_steps_per_second": 12.203,
+      "step": 12000
+    },
+    {
+      "epoch": 48.0,
+      "eval_exact_match": 14.0,
+      "eval_f1": 20.423154231389532,
+      "step": 12000
+    },
+    {
+      "epoch": 48.4,
+      "grad_norm": 0.5641322731971741,
+      "learning_rate": 0.0005,
+      "loss": 0.3545,
+      "step": 12100
+    },
+    {
+      "epoch": 48.8,
+      "grad_norm": 0.752568781375885,
+      "learning_rate": 0.0005,
+      "loss": 0.3772,
+      "step": 12200
+    },
+    {
+      "epoch": 49.0,
+      "eval_accuracy": 0.47755555555555557,
+      "eval_loss": 4.012275695800781,
+      "eval_runtime": 5.9118,
+      "eval_samples_per_second": 84.576,
+      "eval_steps_per_second": 10.657,
+      "step": 12250
+    },
+    {
+      "epoch": 49.0,
+      "eval_exact_match": 12.2,
+      "eval_f1": 18.930353535353543,
+      "step": 12250
+    },
+    {
+      "epoch": 49.2,
+      "grad_norm": 0.8060943484306335,
+      "learning_rate": 0.0005,
+      "loss": 0.3703,
+      "step": 12300
+    },
+    {
+      "epoch": 49.6,
+      "grad_norm": 1.4030524492263794,
+      "learning_rate": 0.0005,
+      "loss": 0.3626,
+      "step": 12400
+    },
+    {
+      "epoch": 50.0,
+      "grad_norm": 0.9272465705871582,
+      "learning_rate": 0.0005,
+      "loss": 0.3808,
+      "step": 12500
+    },
+    {
+      "epoch": 50.0,
+      "eval_accuracy": 0.47844444444444445,
+      "eval_loss": 4.036568641662598,
+      "eval_runtime": 5.8609,
+      "eval_samples_per_second": 85.31,
+      "eval_steps_per_second": 10.749,
+      "step": 12500
+    },
+    {
+      "epoch": 50.0,
+      "eval_exact_match": 13.2,
+      "eval_f1": 20.598610833610845,
+      "step": 12500
+    },
+    {
+      "epoch": 50.0,
+      "step": 12500,
+      "total_flos": 3.823469760432046e+17,
+      "train_loss": 0.49708512481689454,
+      "train_runtime": 15055.6405,
+      "train_samples_per_second": 26.568,
+      "train_steps_per_second": 0.83
     }
   ],
   "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": 3.823469760432046e+17,
+  "train_batch_size": 2,
   "trial_name": null,
   "trial_params": null
 }