Training in progress, step 10000

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7754bb41003fb89bfa30e3352933a73fcd27e921965535e0ee980469396bfc2a
+oid sha256:ca2d642f0f79bc4c657e49b0ffc5e40fdd2d860461f38472b2e70b0ee9627d23
 size 319977229

{checkpoint-8000 → checkpoint-10000}/adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:646ed65259dc5d04c5282bbe5b348c9f1f0cb64b734fd66ea1c6922f99f884ff
+oid sha256:ca2d642f0f79bc4c657e49b0ffc5e40fdd2d860461f38472b2e70b0ee9627d23
 size 319977229

{checkpoint-8000 → checkpoint-10000}/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d7b01a8b56bf99265a76988d8610b9373a259d30880322631c55bbdfa6bc8437
+oid sha256:13b91c1a3830885c38167b2bb1887c4d18368e4fbd2646fb9eb79d7cf35a8344
 size 1279539973

{checkpoint-8000 → checkpoint-10000}/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c952eba50974e6f7b653fe3f67256c91eefd65860fa93b3fdfe43130fd128ba3
+oid sha256:9b74a4f2cac52e5bf2071acc79b46dd7af7a29f3b19955549aea235de7dd1eb3
 size 14511

{checkpoint-8000 → checkpoint-10000}/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ec0a535d2c9c4c62a74336a7f93b6d947a1152f53a6066eccd4123d6b477c15c
+oid sha256:a4d1f3cad6e54148c289328646571ee0d43f9671c93cc919adafe5bd27a33d05
 size 627

{checkpoint-8000 → checkpoint-10000}/trainer_state.json

@@ -1,8 +1,8 @@
 {
   "best_metric": 0.4120824635028839,
   "best_model_checkpoint": "experts/expert-10/checkpoint-5200",
-  "epoch": 3.0191527502594586,
-  "global_step": 8000,
+  "epoch": 3.773940937824323,
+  "global_step": 10000,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -7646,11 +7646,1921 @@
       "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
       "mmlu_loss": 1.7326317684768697,
       "step": 8000
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.0002,
+      "loss": 0.2559,
+      "step": 8010
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.0002,
+      "loss": 0.2461,
+      "step": 8020
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.0002,
+      "loss": 0.2483,
+      "step": 8030
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.0002,
+      "loss": 0.2793,
+      "step": 8040
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.0002,
+      "loss": 0.2798,
+      "step": 8050
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.0002,
+      "loss": 0.2513,
+      "step": 8060
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.0002,
+      "loss": 0.26,
+      "step": 8070
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.0002,
+      "loss": 0.2244,
+      "step": 8080
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.0002,
+      "loss": 0.2245,
+      "step": 8090
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.0002,
+      "loss": 0.2473,
+      "step": 8100
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.0002,
+      "loss": 0.2722,
+      "step": 8110
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.0002,
+      "loss": 0.2332,
+      "step": 8120
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.0002,
+      "loss": 0.2472,
+      "step": 8130
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.0002,
+      "loss": 0.2793,
+      "step": 8140
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.0002,
+      "loss": 0.255,
+      "step": 8150
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.0002,
+      "loss": 0.233,
+      "step": 8160
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.0002,
+      "loss": 0.218,
+      "step": 8170
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.0002,
+      "loss": 0.2363,
+      "step": 8180
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.0002,
+      "loss": 0.2689,
+      "step": 8190
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.0002,
+      "loss": 0.2431,
+      "step": 8200
+    },
+    {
+      "epoch": 3.09,
+      "eval_loss": 0.44677451252937317,
+      "eval_runtime": 103.6908,
+      "eval_samples_per_second": 9.644,
+      "eval_steps_per_second": 4.822,
+      "step": 8200
+    },
+    {
+      "epoch": 3.09,
+      "mmlu_eval_accuracy": 0.49684302016607873,
+      "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
+      "mmlu_eval_accuracy_anatomy": 0.7142857142857143,
+      "mmlu_eval_accuracy_astronomy": 0.5625,
+      "mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5172413793103449,
+      "mmlu_eval_accuracy_college_biology": 0.5,
+      "mmlu_eval_accuracy_college_chemistry": 0.25,
+      "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_mathematics": 0.09090909090909091,
+      "mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_physics": 0.18181818181818182,
+      "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
+      "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.4375,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
+      "mmlu_eval_accuracy_formal_logic": 0.35714285714285715,
+      "mmlu_eval_accuracy_global_facts": 0.4,
+      "mmlu_eval_accuracy_high_school_biology": 0.375,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.4090909090909091,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.4418604651162791,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
+      "mmlu_eval_accuracy_high_school_physics": 0.058823529411764705,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8333333333333334,
+      "mmlu_eval_accuracy_high_school_statistics": 0.34782608695652173,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
+      "mmlu_eval_accuracy_high_school_world_history": 0.7307692307692307,
+      "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
+      "mmlu_eval_accuracy_human_sexuality": 0.5,
+      "mmlu_eval_accuracy_international_law": 0.9230769230769231,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+      "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
+      "mmlu_eval_accuracy_management": 0.7272727272727273,
+      "mmlu_eval_accuracy_marketing": 0.84,
+      "mmlu_eval_accuracy_medical_genetics": 1.0,
+      "mmlu_eval_accuracy_miscellaneous": 0.6976744186046512,
+      "mmlu_eval_accuracy_moral_disputes": 0.5789473684210527,
+      "mmlu_eval_accuracy_moral_scenarios": 0.23,
+      "mmlu_eval_accuracy_nutrition": 0.6060606060606061,
+      "mmlu_eval_accuracy_philosophy": 0.4411764705882353,
+      "mmlu_eval_accuracy_prehistory": 0.42857142857142855,
+      "mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
+      "mmlu_eval_accuracy_professional_law": 0.35294117647058826,
+      "mmlu_eval_accuracy_professional_medicine": 0.5161290322580645,
+      "mmlu_eval_accuracy_professional_psychology": 0.4927536231884058,
+      "mmlu_eval_accuracy_public_relations": 0.5,
+      "mmlu_eval_accuracy_security_studies": 0.4074074074074074,
+      "mmlu_eval_accuracy_sociology": 0.6818181818181818,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.5,
+      "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
+      "mmlu_loss": 1.6755223667185861,
+      "step": 8200
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.0002,
+      "loss": 0.256,
+      "step": 8210
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.0002,
+      "loss": 0.2537,
+      "step": 8220
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.0002,
+      "loss": 0.2487,
+      "step": 8230
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.0002,
+      "loss": 0.2516,
+      "step": 8240
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.0002,
+      "loss": 0.2536,
+      "step": 8250
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.0002,
+      "loss": 0.2443,
+      "step": 8260
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.0002,
+      "loss": 0.2416,
+      "step": 8270
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.0002,
+      "loss": 0.2233,
+      "step": 8280
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.0002,
+      "loss": 0.2477,
+      "step": 8290
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.0002,
+      "loss": 0.2589,
+      "step": 8300
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.0002,
+      "loss": 0.2895,
+      "step": 8310
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.0002,
+      "loss": 0.2549,
+      "step": 8320
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.0002,
+      "loss": 0.2598,
+      "step": 8330
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.0002,
+      "loss": 0.2644,
+      "step": 8340
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.0002,
+      "loss": 0.2634,
+      "step": 8350
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.0002,
+      "loss": 0.2791,
+      "step": 8360
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.0002,
+      "loss": 0.268,
+      "step": 8370
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.0002,
+      "loss": 0.2445,
+      "step": 8380
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.0002,
+      "loss": 0.244,
+      "step": 8390
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.0002,
+      "loss": 0.2543,
+      "step": 8400
+    },
+    {
+      "epoch": 3.17,
+      "eval_loss": 0.44773170351982117,
+      "eval_runtime": 103.8011,
+      "eval_samples_per_second": 9.634,
+      "eval_steps_per_second": 4.817,
+      "step": 8400
+    },
+    {
+      "epoch": 3.17,
+      "mmlu_eval_accuracy": 0.5005608847321679,
+      "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
+      "mmlu_eval_accuracy_anatomy": 0.7142857142857143,
+      "mmlu_eval_accuracy_astronomy": 0.375,
+      "mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5862068965517241,
+      "mmlu_eval_accuracy_college_biology": 0.5,
+      "mmlu_eval_accuracy_college_chemistry": 0.375,
+      "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
+      "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
+      "mmlu_eval_accuracy_college_physics": 0.18181818181818182,
+      "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
+      "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.4375,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
+      "mmlu_eval_accuracy_formal_logic": 0.42857142857142855,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.40625,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.7222222222222222,
+      "mmlu_eval_accuracy_high_school_geography": 0.8636363636363636,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.4418604651162791,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5384615384615384,
+      "mmlu_eval_accuracy_high_school_physics": 0.11764705882352941,
+      "mmlu_eval_accuracy_high_school_psychology": 0.85,
+      "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
+      "mmlu_eval_accuracy_high_school_world_history": 0.6923076923076923,
+      "mmlu_eval_accuracy_human_aging": 0.5652173913043478,
+      "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+      "mmlu_eval_accuracy_international_law": 0.9230769230769231,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+      "mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
+      "mmlu_eval_accuracy_management": 0.7272727272727273,
+      "mmlu_eval_accuracy_marketing": 0.84,
+      "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
+      "mmlu_eval_accuracy_miscellaneous": 0.7093023255813954,
+      "mmlu_eval_accuracy_moral_disputes": 0.5526315789473685,
+      "mmlu_eval_accuracy_moral_scenarios": 0.26,
+      "mmlu_eval_accuracy_nutrition": 0.6060606060606061,
+      "mmlu_eval_accuracy_philosophy": 0.5294117647058824,
+      "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
+      "mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
+      "mmlu_eval_accuracy_professional_law": 0.3411764705882353,
+      "mmlu_eval_accuracy_professional_medicine": 0.5483870967741935,
+      "mmlu_eval_accuracy_professional_psychology": 0.4492753623188406,
+      "mmlu_eval_accuracy_public_relations": 0.5,
+      "mmlu_eval_accuracy_security_studies": 0.37037037037037035,
+      "mmlu_eval_accuracy_sociology": 0.7272727272727273,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.4444444444444444,
+      "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+      "mmlu_loss": 1.913919418155369,
+      "step": 8400
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.0002,
+      "loss": 0.2659,
+      "step": 8410
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.0002,
+      "loss": 0.2569,
+      "step": 8420
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.0002,
+      "loss": 0.2549,
+      "step": 8430
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.0002,
+      "loss": 0.2692,
+      "step": 8440
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.0002,
+      "loss": 0.253,
+      "step": 8450
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.0002,
+      "loss": 0.2554,
+      "step": 8460
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.0002,
+      "loss": 0.2434,
+      "step": 8470
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.0002,
+      "loss": 0.2782,
+      "step": 8480
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.0002,
+      "loss": 0.248,
+      "step": 8490
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.0002,
+      "loss": 0.2421,
+      "step": 8500
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.0002,
+      "loss": 0.2713,
+      "step": 8510
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.0002,
+      "loss": 0.302,
+      "step": 8520
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.0002,
+      "loss": 0.2296,
+      "step": 8530
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.0002,
+      "loss": 0.2504,
+      "step": 8540
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.0002,
+      "loss": 0.2448,
+      "step": 8550
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.0002,
+      "loss": 0.2559,
+      "step": 8560
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.0002,
+      "loss": 0.2477,
+      "step": 8570
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.0002,
+      "loss": 0.2633,
+      "step": 8580
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.0002,
+      "loss": 0.2511,
+      "step": 8590
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.0002,
+      "loss": 0.2375,
+      "step": 8600
+    },
+    {
+      "epoch": 3.25,
+      "eval_loss": 0.4444422721862793,
+      "eval_runtime": 103.8053,
+      "eval_samples_per_second": 9.633,
+      "eval_steps_per_second": 4.817,
+      "step": 8600
+    },
+    {
+      "epoch": 3.25,
+      "mmlu_eval_accuracy": 0.49446869409691946,
+      "mmlu_eval_accuracy_abstract_algebra": 0.18181818181818182,
+      "mmlu_eval_accuracy_anatomy": 0.7142857142857143,
+      "mmlu_eval_accuracy_astronomy": 0.375,
+      "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5172413793103449,
+      "mmlu_eval_accuracy_college_biology": 0.5,
+      "mmlu_eval_accuracy_college_chemistry": 0.25,
+      "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
+      "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
+      "mmlu_eval_accuracy_college_physics": 0.18181818181818182,
+      "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
+      "mmlu_eval_accuracy_conceptual_physics": 0.34615384615384615,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.4375,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.3902439024390244,
+      "mmlu_eval_accuracy_formal_logic": 0.5,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.4375,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.27586206896551724,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5384615384615384,
+      "mmlu_eval_accuracy_high_school_physics": 0.11764705882352941,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8333333333333334,
+      "mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
+      "mmlu_eval_accuracy_high_school_world_history": 0.7307692307692307,
+      "mmlu_eval_accuracy_human_aging": 0.5652173913043478,
+      "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+      "mmlu_eval_accuracy_international_law": 0.9230769230769231,
+      "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
+      "mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
+      "mmlu_eval_accuracy_management": 0.7272727272727273,
+      "mmlu_eval_accuracy_marketing": 0.84,
+      "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
+      "mmlu_eval_accuracy_miscellaneous": 0.7209302325581395,
+      "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
+      "mmlu_eval_accuracy_moral_scenarios": 0.23,
+      "mmlu_eval_accuracy_nutrition": 0.6363636363636364,
+      "mmlu_eval_accuracy_philosophy": 0.5882352941176471,
+      "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
+      "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
+      "mmlu_eval_accuracy_professional_law": 0.3411764705882353,
+      "mmlu_eval_accuracy_professional_medicine": 0.5806451612903226,
+      "mmlu_eval_accuracy_professional_psychology": 0.4492753623188406,
+      "mmlu_eval_accuracy_public_relations": 0.5,
+      "mmlu_eval_accuracy_security_studies": 0.37037037037037035,
+      "mmlu_eval_accuracy_sociology": 0.7272727272727273,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.5,
+      "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
+      "mmlu_loss": 1.8425939399408608,
+      "step": 8600
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.0002,
+      "loss": 0.2548,
+      "step": 8610
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.0002,
+      "loss": 0.2687,
+      "step": 8620
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.0002,
+      "loss": 0.266,
+      "step": 8630
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.0002,
+      "loss": 0.2434,
+      "step": 8640
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.0002,
+      "loss": 0.2515,
+      "step": 8650
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.0002,
+      "loss": 0.2651,
+      "step": 8660
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.0002,
+      "loss": 0.2386,
+      "step": 8670
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.0002,
+      "loss": 0.2915,
+      "step": 8680
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.0002,
+      "loss": 0.2617,
+      "step": 8690
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.0002,
+      "loss": 0.2343,
+      "step": 8700
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.0002,
+      "loss": 0.3014,
+      "step": 8710
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.0002,
+      "loss": 0.2322,
+      "step": 8720
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.0002,
+      "loss": 0.2442,
+      "step": 8730
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.0002,
+      "loss": 0.2935,
+      "step": 8740
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.0002,
+      "loss": 0.267,
+      "step": 8750
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.0002,
+      "loss": 0.2437,
+      "step": 8760
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.0002,
+      "loss": 0.2508,
+      "step": 8770
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.0002,
+      "loss": 0.2499,
+      "step": 8780
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.0002,
+      "loss": 0.2514,
+      "step": 8790
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.0002,
+      "loss": 0.2656,
+      "step": 8800
+    },
+    {
+      "epoch": 3.32,
+      "eval_loss": 0.44510698318481445,
+      "eval_runtime": 103.723,
+      "eval_samples_per_second": 9.641,
+      "eval_steps_per_second": 4.821,
+      "step": 8800
+    },
+    {
+      "epoch": 3.32,
+      "mmlu_eval_accuracy": 0.49959311172053505,
+      "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
+      "mmlu_eval_accuracy_anatomy": 0.7142857142857143,
+      "mmlu_eval_accuracy_astronomy": 0.375,
+      "mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5172413793103449,
+      "mmlu_eval_accuracy_college_biology": 0.5,
+      "mmlu_eval_accuracy_college_chemistry": 0.5,
+      "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
+      "mmlu_eval_accuracy_college_medicine": 0.5,
+      "mmlu_eval_accuracy_college_physics": 0.18181818181818182,
+      "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
+      "mmlu_eval_accuracy_conceptual_physics": 0.38461538461538464,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.4375,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.3902439024390244,
+      "mmlu_eval_accuracy_formal_logic": 0.5,
+      "mmlu_eval_accuracy_global_facts": 0.4,
+      "mmlu_eval_accuracy_high_school_biology": 0.40625,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
+      "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.27586206896551724,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5769230769230769,
+      "mmlu_eval_accuracy_high_school_physics": 0.11764705882352941,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8333333333333334,
+      "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
+      "mmlu_eval_accuracy_high_school_world_history": 0.6923076923076923,
+      "mmlu_eval_accuracy_human_aging": 0.6086956521739131,
+      "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+      "mmlu_eval_accuracy_international_law": 0.9230769230769231,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+      "mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
+      "mmlu_eval_accuracy_management": 0.7272727272727273,
+      "mmlu_eval_accuracy_marketing": 0.84,
+      "mmlu_eval_accuracy_medical_genetics": 1.0,
+      "mmlu_eval_accuracy_miscellaneous": 0.6976744186046512,
+      "mmlu_eval_accuracy_moral_disputes": 0.5263157894736842,
+      "mmlu_eval_accuracy_moral_scenarios": 0.23,
+      "mmlu_eval_accuracy_nutrition": 0.6060606060606061,
+      "mmlu_eval_accuracy_philosophy": 0.5,
+      "mmlu_eval_accuracy_prehistory": 0.45714285714285713,
+      "mmlu_eval_accuracy_professional_accounting": 0.25806451612903225,
+      "mmlu_eval_accuracy_professional_law": 0.3352941176470588,
+      "mmlu_eval_accuracy_professional_medicine": 0.5483870967741935,
+      "mmlu_eval_accuracy_professional_psychology": 0.43478260869565216,
+      "mmlu_eval_accuracy_public_relations": 0.5,
+      "mmlu_eval_accuracy_security_studies": 0.37037037037037035,
+      "mmlu_eval_accuracy_sociology": 0.7272727272727273,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.5,
+      "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
+      "mmlu_loss": 1.966283672160953,
+      "step": 8800
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.0002,
+      "loss": 0.2716,
+      "step": 8810
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.0002,
+      "loss": 0.2498,
+      "step": 8820
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.0002,
+      "loss": 0.271,
+      "step": 8830
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.0002,
+      "loss": 0.2947,
+      "step": 8840
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.0002,
+      "loss": 0.2409,
+      "step": 8850
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.0002,
+      "loss": 0.2452,
+      "step": 8860
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.0002,
+      "loss": 0.2649,
+      "step": 8870
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.0002,
+      "loss": 0.2434,
+      "step": 8880
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.0002,
+      "loss": 0.2709,
+      "step": 8890
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.0002,
+      "loss": 0.2648,
+      "step": 8900
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.0002,
+      "loss": 0.2713,
+      "step": 8910
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.0002,
+      "loss": 0.248,
+      "step": 8920
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.0002,
+      "loss": 0.2569,
+      "step": 8930
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.0002,
+      "loss": 0.2533,
+      "step": 8940
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.0002,
+      "loss": 0.2663,
+      "step": 8950
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.0002,
+      "loss": 0.2252,
+      "step": 8960
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.0002,
+      "loss": 0.2691,
+      "step": 8970
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.0002,
+      "loss": 0.2574,
+      "step": 8980
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.0002,
+      "loss": 0.2727,
+      "step": 8990
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.0002,
+      "loss": 0.2504,
+      "step": 9000
+    },
+    {
+      "epoch": 3.4,
+      "eval_loss": 0.4468487501144409,
+      "eval_runtime": 103.7779,
+      "eval_samples_per_second": 9.636,
+      "eval_steps_per_second": 4.818,
+      "step": 9000
+    },
+    {
+      "epoch": 3.4,
+      "mmlu_eval_accuracy": 0.5008790997186083,
+      "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
+      "mmlu_eval_accuracy_anatomy": 0.7142857142857143,
+      "mmlu_eval_accuracy_astronomy": 0.375,
+      "mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
+      "mmlu_eval_accuracy_college_biology": 0.4375,
+      "mmlu_eval_accuracy_college_chemistry": 0.375,
+      "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
+      "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_physics": 0.18181818181818182,
+      "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
+      "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.4375,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
+      "mmlu_eval_accuracy_formal_logic": 0.5,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.4375,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
+      "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.6153846153846154,
+      "mmlu_eval_accuracy_high_school_physics": 0.11764705882352941,
+      "mmlu_eval_accuracy_high_school_psychology": 0.85,
+      "mmlu_eval_accuracy_high_school_statistics": 0.34782608695652173,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
+      "mmlu_eval_accuracy_high_school_world_history": 0.7307692307692307,
+      "mmlu_eval_accuracy_human_aging": 0.6086956521739131,
+      "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
+      "mmlu_eval_accuracy_international_law": 0.9230769230769231,
+      "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+      "mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
+      "mmlu_eval_accuracy_management": 0.7272727272727273,
+      "mmlu_eval_accuracy_marketing": 0.84,
+      "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
+      "mmlu_eval_accuracy_miscellaneous": 0.7209302325581395,
+      "mmlu_eval_accuracy_moral_disputes": 0.5263157894736842,
+      "mmlu_eval_accuracy_moral_scenarios": 0.23,
+      "mmlu_eval_accuracy_nutrition": 0.5757575757575758,
+      "mmlu_eval_accuracy_philosophy": 0.5,
+      "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
+      "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
+      "mmlu_eval_accuracy_professional_law": 0.36470588235294116,
+      "mmlu_eval_accuracy_professional_medicine": 0.5483870967741935,
+      "mmlu_eval_accuracy_professional_psychology": 0.4492753623188406,
+      "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
+      "mmlu_eval_accuracy_security_studies": 0.37037037037037035,
+      "mmlu_eval_accuracy_sociology": 0.6818181818181818,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.5,
+      "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
+      "mmlu_loss": 1.7695583491039026,
+      "step": 9000
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.0002,
+      "loss": 0.2562,
+      "step": 9010
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.0002,
+      "loss": 0.2719,
+      "step": 9020
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.0002,
+      "loss": 0.2505,
+      "step": 9030
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.0002,
+      "loss": 0.2439,
+      "step": 9040
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.0002,
+      "loss": 0.236,
+      "step": 9050
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.0002,
+      "loss": 0.2481,
+      "step": 9060
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.0002,
+      "loss": 0.2504,
+      "step": 9070
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.0002,
+      "loss": 0.2354,
+      "step": 9080
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.0002,
+      "loss": 0.2311,
+      "step": 9090
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.0002,
+      "loss": 0.2788,
+      "step": 9100
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.0002,
+      "loss": 0.2507,
+      "step": 9110
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.0002,
+      "loss": 0.2632,
+      "step": 9120
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.0002,
+      "loss": 0.2514,
+      "step": 9130
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.0002,
+      "loss": 0.3119,
+      "step": 9140
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.0002,
+      "loss": 0.27,
+      "step": 9150
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.0002,
+      "loss": 0.2565,
+      "step": 9160
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.0002,
+      "loss": 0.2504,
+      "step": 9170
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.0002,
+      "loss": 0.252,
+      "step": 9180
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.0002,
+      "loss": 0.2466,
+      "step": 9190
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.0002,
+      "loss": 0.271,
+      "step": 9200
+    },
+    {
+      "epoch": 3.47,
+      "eval_loss": 0.4423753619194031,
+      "eval_runtime": 103.6623,
+      "eval_samples_per_second": 9.647,
+      "eval_steps_per_second": 4.823,
+      "step": 9200
+    },
+    {
+      "epoch": 3.47,
+      "mmlu_eval_accuracy": 0.4873543767068175,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.7142857142857143,
+      "mmlu_eval_accuracy_astronomy": 0.3125,
+      "mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
+      "mmlu_eval_accuracy_college_biology": 0.5,
+      "mmlu_eval_accuracy_college_chemistry": 0.25,
+      "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_mathematics": 0.09090909090909091,
+      "mmlu_eval_accuracy_college_medicine": 0.5,
+      "mmlu_eval_accuracy_college_physics": 0.18181818181818182,
+      "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
+      "mmlu_eval_accuracy_conceptual_physics": 0.3076923076923077,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.25,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
+      "mmlu_eval_accuracy_formal_logic": 0.35714285714285715,
+      "mmlu_eval_accuracy_global_facts": 0.6,
+      "mmlu_eval_accuracy_high_school_biology": 0.40625,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.4090909090909091,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5384615384615384,
+      "mmlu_eval_accuracy_high_school_physics": 0.11764705882352941,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8333333333333334,
+      "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
+      "mmlu_eval_accuracy_high_school_world_history": 0.6923076923076923,
+      "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
+      "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
+      "mmlu_eval_accuracy_international_law": 0.9230769230769231,
+      "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+      "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
+      "mmlu_eval_accuracy_management": 0.7272727272727273,
+      "mmlu_eval_accuracy_marketing": 0.84,
+      "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
+      "mmlu_eval_accuracy_miscellaneous": 0.7209302325581395,
+      "mmlu_eval_accuracy_moral_disputes": 0.5526315789473685,
+      "mmlu_eval_accuracy_moral_scenarios": 0.29,
+      "mmlu_eval_accuracy_nutrition": 0.6060606060606061,
+      "mmlu_eval_accuracy_philosophy": 0.5588235294117647,
+      "mmlu_eval_accuracy_prehistory": 0.42857142857142855,
+      "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
+      "mmlu_eval_accuracy_professional_law": 0.31176470588235294,
+      "mmlu_eval_accuracy_professional_medicine": 0.5483870967741935,
+      "mmlu_eval_accuracy_professional_psychology": 0.4492753623188406,
+      "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
+      "mmlu_eval_accuracy_security_studies": 0.37037037037037035,
+      "mmlu_eval_accuracy_sociology": 0.6818181818181818,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.5,
+      "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
+      "mmlu_loss": 1.736403314703127,
+      "step": 9200
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.0002,
+      "loss": 0.3062,
+      "step": 9210
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.0002,
+      "loss": 0.2405,
+      "step": 9220
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.0002,
+      "loss": 0.2398,
+      "step": 9230
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.0002,
+      "loss": 0.2919,
+      "step": 9240
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.0002,
+      "loss": 0.279,
+      "step": 9250
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.0002,
+      "loss": 0.2545,
+      "step": 9260
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.0002,
+      "loss": 0.2533,
+      "step": 9270
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.0002,
+      "loss": 0.2638,
+      "step": 9280
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.0002,
+      "loss": 0.2707,
+      "step": 9290
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.0002,
+      "loss": 0.2473,
+      "step": 9300
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.0002,
+      "loss": 0.236,
+      "step": 9310
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.0002,
+      "loss": 0.2767,
+      "step": 9320
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.0002,
+      "loss": 0.273,
+      "step": 9330
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.0002,
+      "loss": 0.279,
+      "step": 9340
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.0002,
+      "loss": 0.2711,
+      "step": 9350
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.0002,
+      "loss": 0.2518,
+      "step": 9360
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.0002,
+      "loss": 0.2497,
+      "step": 9370
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.0002,
+      "loss": 0.261,
+      "step": 9380
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.0002,
+      "loss": 0.2882,
+      "step": 9390
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.0002,
+      "loss": 0.2787,
+      "step": 9400
+    },
+    {
+      "epoch": 3.55,
+      "eval_loss": 0.4377273917198181,
+      "eval_runtime": 103.8201,
+      "eval_samples_per_second": 9.632,
+      "eval_steps_per_second": 4.816,
+      "step": 9400
+    },
+    {
+      "epoch": 3.55,
+      "mmlu_eval_accuracy": 0.4929698947566552,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.7142857142857143,
+      "mmlu_eval_accuracy_astronomy": 0.4375,
+      "mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
+      "mmlu_eval_accuracy_college_biology": 0.4375,
+      "mmlu_eval_accuracy_college_chemistry": 0.5,
+      "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
+      "mmlu_eval_accuracy_college_medicine": 0.5,
+      "mmlu_eval_accuracy_college_physics": 0.09090909090909091,
+      "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
+      "mmlu_eval_accuracy_conceptual_physics": 0.34615384615384615,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.3125,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.3902439024390244,
+      "mmlu_eval_accuracy_formal_logic": 0.35714285714285715,
+      "mmlu_eval_accuracy_global_facts": 0.6,
+      "mmlu_eval_accuracy_high_school_biology": 0.4375,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
+      "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.3953488372093023,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
+      "mmlu_eval_accuracy_high_school_physics": 0.11764705882352941,
+      "mmlu_eval_accuracy_high_school_psychology": 0.85,
+      "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
+      "mmlu_eval_accuracy_high_school_us_history": 0.5909090909090909,
+      "mmlu_eval_accuracy_high_school_world_history": 0.6923076923076923,
+      "mmlu_eval_accuracy_human_aging": 0.6086956521739131,
+      "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+      "mmlu_eval_accuracy_international_law": 0.9230769230769231,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+      "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
+      "mmlu_eval_accuracy_management": 0.7272727272727273,
+      "mmlu_eval_accuracy_marketing": 0.84,
+      "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
+      "mmlu_eval_accuracy_miscellaneous": 0.7325581395348837,
+      "mmlu_eval_accuracy_moral_disputes": 0.5263157894736842,
+      "mmlu_eval_accuracy_moral_scenarios": 0.23,
+      "mmlu_eval_accuracy_nutrition": 0.5757575757575758,
+      "mmlu_eval_accuracy_philosophy": 0.5294117647058824,
+      "mmlu_eval_accuracy_prehistory": 0.45714285714285713,
+      "mmlu_eval_accuracy_professional_accounting": 0.22580645161290322,
+      "mmlu_eval_accuracy_professional_law": 0.3411764705882353,
+      "mmlu_eval_accuracy_professional_medicine": 0.5806451612903226,
+      "mmlu_eval_accuracy_professional_psychology": 0.4492753623188406,
+      "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
+      "mmlu_eval_accuracy_security_studies": 0.37037037037037035,
+      "mmlu_eval_accuracy_sociology": 0.6363636363636364,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.5,
+      "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
+      "mmlu_loss": 1.6837934175452738,
+      "step": 9400
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.0002,
+      "loss": 0.2684,
+      "step": 9410
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.0002,
+      "loss": 0.2639,
+      "step": 9420
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.0002,
+      "loss": 0.2368,
+      "step": 9430
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.0002,
+      "loss": 0.2597,
+      "step": 9440
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.0002,
+      "loss": 0.2633,
+      "step": 9450
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.0002,
+      "loss": 0.2697,
+      "step": 9460
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.0002,
+      "loss": 0.259,
+      "step": 9470
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.0002,
+      "loss": 0.2495,
+      "step": 9480
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.0002,
+      "loss": 0.2398,
+      "step": 9490
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.0002,
+      "loss": 0.299,
+      "step": 9500
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.0002,
+      "loss": 0.2481,
+      "step": 9510
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.0002,
+      "loss": 0.2575,
+      "step": 9520
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.0002,
+      "loss": 0.2617,
+      "step": 9530
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.0002,
+      "loss": 0.2644,
+      "step": 9540
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.0002,
+      "loss": 0.2573,
+      "step": 9550
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.0002,
+      "loss": 0.2737,
+      "step": 9560
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.0002,
+      "loss": 0.2426,
+      "step": 9570
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.0002,
+      "loss": 0.2495,
+      "step": 9580
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.0002,
+      "loss": 0.2438,
+      "step": 9590
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.0002,
+      "loss": 0.2751,
+      "step": 9600
+    },
+    {
+      "epoch": 3.62,
+      "eval_loss": 0.4409026503562927,
+      "eval_runtime": 103.7547,
+      "eval_samples_per_second": 9.638,
+      "eval_steps_per_second": 4.819,
+      "step": 9600
+    },
+    {
+      "epoch": 3.62,
+      "mmlu_eval_accuracy": 0.48665891694110297,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.3125,
+      "mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
+      "mmlu_eval_accuracy_college_biology": 0.375,
+      "mmlu_eval_accuracy_college_chemistry": 0.5,
+      "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
+      "mmlu_eval_accuracy_college_medicine": 0.5,
+      "mmlu_eval_accuracy_college_physics": 0.18181818181818182,
+      "mmlu_eval_accuracy_computer_security": 0.2727272727272727,
+      "mmlu_eval_accuracy_conceptual_physics": 0.34615384615384615,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.375,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
+      "mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.40625,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
+      "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
+      "mmlu_eval_accuracy_high_school_physics": 0.11764705882352941,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8333333333333334,
+      "mmlu_eval_accuracy_high_school_statistics": 0.4782608695652174,
+      "mmlu_eval_accuracy_high_school_us_history": 0.5909090909090909,
+      "mmlu_eval_accuracy_high_school_world_history": 0.6923076923076923,
+      "mmlu_eval_accuracy_human_aging": 0.6086956521739131,
+      "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+      "mmlu_eval_accuracy_international_law": 0.9230769230769231,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
+      "mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
+      "mmlu_eval_accuracy_management": 0.7272727272727273,
+      "mmlu_eval_accuracy_marketing": 0.84,
+      "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
+      "mmlu_eval_accuracy_miscellaneous": 0.7093023255813954,
+      "mmlu_eval_accuracy_moral_disputes": 0.5526315789473685,
+      "mmlu_eval_accuracy_moral_scenarios": 0.26,
+      "mmlu_eval_accuracy_nutrition": 0.5454545454545454,
+      "mmlu_eval_accuracy_philosophy": 0.5882352941176471,
+      "mmlu_eval_accuracy_prehistory": 0.45714285714285713,
+      "mmlu_eval_accuracy_professional_accounting": 0.22580645161290322,
+      "mmlu_eval_accuracy_professional_law": 0.3176470588235294,
+      "mmlu_eval_accuracy_professional_medicine": 0.5161290322580645,
+      "mmlu_eval_accuracy_professional_psychology": 0.4492753623188406,
+      "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
+      "mmlu_eval_accuracy_security_studies": 0.37037037037037035,
+      "mmlu_eval_accuracy_sociology": 0.5909090909090909,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.4444444444444444,
+      "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
+      "mmlu_loss": 1.582305176995753,
+      "step": 9600
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.0002,
+      "loss": 0.2438,
+      "step": 9610
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.0002,
+      "loss": 0.2626,
+      "step": 9620
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.0002,
+      "loss": 0.2264,
+      "step": 9630
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.0002,
+      "loss": 0.243,
+      "step": 9640
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.0002,
+      "loss": 0.2587,
+      "step": 9650
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.0002,
+      "loss": 0.2641,
+      "step": 9660
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.0002,
+      "loss": 0.269,
+      "step": 9670
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.0002,
+      "loss": 0.2666,
+      "step": 9680
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.0002,
+      "loss": 0.2533,
+      "step": 9690
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.0002,
+      "loss": 0.2786,
+      "step": 9700
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.0002,
+      "loss": 0.285,
+      "step": 9710
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.0002,
+      "loss": 0.2312,
+      "step": 9720
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.0002,
+      "loss": 0.2439,
+      "step": 9730
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.0002,
+      "loss": 0.2477,
+      "step": 9740
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.0002,
+      "loss": 0.2635,
+      "step": 9750
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.0002,
+      "loss": 0.2515,
+      "step": 9760
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.0002,
+      "loss": 0.276,
+      "step": 9770
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.0002,
+      "loss": 0.2769,
+      "step": 9780
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.0002,
+      "loss": 0.2931,
+      "step": 9790
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.0002,
+      "loss": 0.266,
+      "step": 9800
+    },
+    {
+      "epoch": 3.7,
+      "eval_loss": 0.43637925386428833,
+      "eval_runtime": 103.707,
+      "eval_samples_per_second": 9.643,
+      "eval_steps_per_second": 4.821,
+      "step": 9800
+    },
+    {
+      "epoch": 3.7,
+      "mmlu_eval_accuracy": 0.49235814106296416,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.7142857142857143,
+      "mmlu_eval_accuracy_astronomy": 0.375,
+      "mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
+      "mmlu_eval_accuracy_college_biology": 0.4375,
+      "mmlu_eval_accuracy_college_chemistry": 0.375,
+      "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
+      "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_physics": 0.18181818181818182,
+      "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
+      "mmlu_eval_accuracy_conceptual_physics": 0.34615384615384615,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.3125,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
+      "mmlu_eval_accuracy_formal_logic": 0.42857142857142855,
+      "mmlu_eval_accuracy_global_facts": 0.6,
+      "mmlu_eval_accuracy_high_school_biology": 0.40625,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
+      "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5384615384615384,
+      "mmlu_eval_accuracy_high_school_physics": 0.11764705882352941,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8333333333333334,
+      "mmlu_eval_accuracy_high_school_statistics": 0.43478260869565216,
+      "mmlu_eval_accuracy_high_school_us_history": 0.5909090909090909,
+      "mmlu_eval_accuracy_high_school_world_history": 0.7307692307692307,
+      "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
+      "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
+      "mmlu_eval_accuracy_international_law": 0.9230769230769231,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+      "mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
+      "mmlu_eval_accuracy_management": 0.7272727272727273,
+      "mmlu_eval_accuracy_marketing": 0.8,
+      "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
+      "mmlu_eval_accuracy_miscellaneous": 0.6976744186046512,
+      "mmlu_eval_accuracy_moral_disputes": 0.5263157894736842,
+      "mmlu_eval_accuracy_moral_scenarios": 0.25,
+      "mmlu_eval_accuracy_nutrition": 0.5757575757575758,
+      "mmlu_eval_accuracy_philosophy": 0.5,
+      "mmlu_eval_accuracy_prehistory": 0.45714285714285713,
+      "mmlu_eval_accuracy_professional_accounting": 0.22580645161290322,
+      "mmlu_eval_accuracy_professional_law": 0.3176470588235294,
+      "mmlu_eval_accuracy_professional_medicine": 0.5806451612903226,
+      "mmlu_eval_accuracy_professional_psychology": 0.463768115942029,
+      "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
+      "mmlu_eval_accuracy_security_studies": 0.4074074074074074,
+      "mmlu_eval_accuracy_sociology": 0.6363636363636364,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.5555555555555556,
+      "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+      "mmlu_loss": 1.6827692933789122,
+      "step": 9800
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.0002,
+      "loss": 0.2515,
+      "step": 9810
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.0002,
+      "loss": 0.2963,
+      "step": 9820
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.0002,
+      "loss": 0.2527,
+      "step": 9830
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.0002,
+      "loss": 0.2947,
+      "step": 9840
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.0002,
+      "loss": 0.2876,
+      "step": 9850
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.0002,
+      "loss": 0.2805,
+      "step": 9860
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.0002,
+      "loss": 0.253,
+      "step": 9870
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.0002,
+      "loss": 0.2665,
+      "step": 9880
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.0002,
+      "loss": 0.2412,
+      "step": 9890
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.0002,
+      "loss": 0.2704,
+      "step": 9900
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.0002,
+      "loss": 0.2702,
+      "step": 9910
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.0002,
+      "loss": 0.2708,
+      "step": 9920
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.0002,
+      "loss": 0.2621,
+      "step": 9930
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.0002,
+      "loss": 0.295,
+      "step": 9940
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.0002,
+      "loss": 0.2596,
+      "step": 9950
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.0002,
+      "loss": 0.2392,
+      "step": 9960
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.0002,
+      "loss": 0.253,
+      "step": 9970
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.0002,
+      "loss": 0.2519,
+      "step": 9980
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.0002,
+      "loss": 0.2662,
+      "step": 9990
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.0002,
+      "loss": 0.2443,
+      "step": 10000
+    },
+    {
+      "epoch": 3.77,
+      "eval_loss": 0.4414416551589966,
+      "eval_runtime": 103.7985,
+      "eval_samples_per_second": 9.634,
+      "eval_steps_per_second": 4.817,
+      "step": 10000
+    },
+    {
+      "epoch": 3.77,
+      "mmlu_eval_accuracy": 0.5035810617973856,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.7142857142857143,
+      "mmlu_eval_accuracy_astronomy": 0.375,
+      "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
+      "mmlu_eval_accuracy_college_biology": 0.5,
+      "mmlu_eval_accuracy_college_chemistry": 0.5,
+      "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_physics": 0.18181818181818182,
+      "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
+      "mmlu_eval_accuracy_conceptual_physics": 0.3076923076923077,
+      "mmlu_eval_accuracy_econometrics": 0.25,
+      "mmlu_eval_accuracy_electrical_engineering": 0.3125,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.3902439024390244,
+      "mmlu_eval_accuracy_formal_logic": 0.42857142857142855,
+      "mmlu_eval_accuracy_global_facts": 0.6,
+      "mmlu_eval_accuracy_high_school_biology": 0.40625,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
+      "mmlu_eval_accuracy_high_school_physics": 0.11764705882352941,
+      "mmlu_eval_accuracy_high_school_psychology": 0.85,
+      "mmlu_eval_accuracy_high_school_statistics": 0.43478260869565216,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
+      "mmlu_eval_accuracy_high_school_world_history": 0.7307692307692307,
+      "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
+      "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
+      "mmlu_eval_accuracy_international_law": 0.9230769230769231,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+      "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
+      "mmlu_eval_accuracy_management": 0.7272727272727273,
+      "mmlu_eval_accuracy_marketing": 0.84,
+      "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
+      "mmlu_eval_accuracy_miscellaneous": 0.7209302325581395,
+      "mmlu_eval_accuracy_moral_disputes": 0.5789473684210527,
+      "mmlu_eval_accuracy_moral_scenarios": 0.24,
+      "mmlu_eval_accuracy_nutrition": 0.6060606060606061,
+      "mmlu_eval_accuracy_philosophy": 0.5588235294117647,
+      "mmlu_eval_accuracy_prehistory": 0.45714285714285713,
+      "mmlu_eval_accuracy_professional_accounting": 0.3225806451612903,
+      "mmlu_eval_accuracy_professional_law": 0.3176470588235294,
+      "mmlu_eval_accuracy_professional_medicine": 0.6451612903225806,
+      "mmlu_eval_accuracy_professional_psychology": 0.43478260869565216,
+      "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
+      "mmlu_eval_accuracy_security_studies": 0.4074074074074074,
+      "mmlu_eval_accuracy_sociology": 0.6818181818181818,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.5555555555555556,
+      "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+      "mmlu_loss": 1.6655967655754587,
+      "step": 10000
     }
   ],
   "max_steps": 10000,
   "num_train_epochs": 4,
-  "total_flos": 5.717073489208443e+17,
+  "total_flos": 7.140498953992765e+17,
   "trial_name": null,
   "trial_params": null
 }

checkpoint-5200/adapter_model/adapter_model/README.md

@@ -246,6 +246,17 @@ The following `bitsandbytes` quantization config was used during training:
 - bnb_4bit_use_double_quant: True
 - bnb_4bit_compute_dtype: bfloat16
 
+The following `bitsandbytes` quantization config was used during training:
+- load_in_8bit: False
+- load_in_4bit: True
+- llm_int8_threshold: 6.0
+- llm_int8_skip_modules: None
+- llm_int8_enable_fp32_cpu_offload: False
+- llm_int8_has_fp16_weight: False
+- bnb_4bit_quant_type: nf4
+- bnb_4bit_use_double_quant: True
+- bnb_4bit_compute_dtype: bfloat16
+
 The following `bitsandbytes` quantization config was used during training:
 - load_in_8bit: False
 - load_in_4bit: True
@@ -280,5 +291,6 @@ The following `bitsandbytes` quantization config was used during training:
 - PEFT 0.4.0
 - PEFT 0.4.0
 - PEFT 0.4.0
+- PEFT 0.4.0
 
 - PEFT 0.4.0

checkpoint-5200/adapter_model/adapter_model/adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7fda114e5453a1b7be85a38f1ff09c348d0c7d129758d415d33b39b2dd2ecd85
+oid sha256:7754bb41003fb89bfa30e3352933a73fcd27e921965535e0ee980469396bfc2a
 size 319977229