Training in progress, step 4200

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:941fcee15938878c1bbc271d655fa527d05a97360b9293b76de97afbf520b878
+oid sha256:7f291bfafb9ef6a2a1ed586008ca33c815141a83aad1477a5a949868e05da6f2
 size 319977229

{checkpoint-2000 → checkpoint-4000/adapter_model/adapter_model}/adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:208b26a6bf48e58b2dd5504ec0f780f75917cc8ed4d5f1d91ae9f2a436030622
+oid sha256:941fcee15938878c1bbc271d655fa527d05a97360b9293b76de97afbf520b878
 size 319977229

{checkpoint-2000/adapter_model/adapter_model → checkpoint-4200}/README.md

@@ -4,28 +4,6 @@ library_name: peft
 ## Training procedure
 
 
-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
-- 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
@@ -38,7 +16,5 @@ The following `bitsandbytes` quantization config was used during training:
 - bnb_4bit_compute_dtype: bfloat16
 ### Framework versions
 
-- PEFT 0.4.0
-- PEFT 0.4.0
 
 - PEFT 0.4.0

{checkpoint-2000/adapter_model/adapter_model → checkpoint-4200}/adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:488ff079cfcce8d1423a4c918348d9b7a286432aabfe02bb957954b57eb75fe3
+oid sha256:7f291bfafb9ef6a2a1ed586008ca33c815141a83aad1477a5a949868e05da6f2
 size 319977229

{checkpoint-2000 → checkpoint-4200}/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:dcfc952d50ab274f579cb7a4d38e012601cbfe54238c2eb7872a6c557ae7118b
+oid sha256:aacdf31e2af638ce3a39687e3793edaf95023d7344f35e14e995bc99aeb69152
 size 1279539973

{checkpoint-2000 → checkpoint-4200}/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0127589819a1e68d43d4f724d6b19218be4d45e525e83205e82ee0875bb59d13
+oid sha256:e0dad2f494d74ce2884fe5b176da88c29f1ceb4addb8a92c248523f518603ff4
 size 14511

{checkpoint-2000 → checkpoint-4200}/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:720bafe9a34b3041027dccef7d9c03bb23064e1fcdd80ae5f27e717d41b7cf31
+oid sha256:d1148c295890aef97427a11a5a53b94b8b59594e8519b33223027fa87b288829
 size 627

{checkpoint-2000 → checkpoint-4200}/trainer_state.json

@@ -1,8 +1,8 @@
 {
-  "best_metric": 0.7063223123550415,
-  "best_model_checkpoint": "experts/expert-9/checkpoint-2000",
-  "epoch": 0.9625797136325351,
-  "global_step": 2000,
+  "best_metric": 0.6935309767723083,
+  "best_model_checkpoint": "experts/expert-9/checkpoint-4000",
+  "epoch": 2.0214173986283237,
+  "global_step": 4200,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -1916,11 +1916,2112 @@
       "mmlu_eval_accuracy_world_religions": 0.631578947368421,
       "mmlu_loss": 1.1613141986548745,
       "step": 2000
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0002,
+      "loss": 0.6937,
+      "step": 2010
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0002,
+      "loss": 0.7293,
+      "step": 2020
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0002,
+      "loss": 0.6768,
+      "step": 2030
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.0002,
+      "loss": 0.6667,
+      "step": 2040
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0002,
+      "loss": 0.6541,
+      "step": 2050
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0002,
+      "loss": 0.6551,
+      "step": 2060
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0002,
+      "loss": 0.695,
+      "step": 2070
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.0002,
+      "loss": 0.7075,
+      "step": 2080
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0002,
+      "loss": 0.6733,
+      "step": 2090
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0002,
+      "loss": 0.6344,
+      "step": 2100
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0002,
+      "loss": 0.6223,
+      "step": 2110
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0002,
+      "loss": 0.6223,
+      "step": 2120
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0002,
+      "loss": 0.6559,
+      "step": 2130
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0002,
+      "loss": 0.6031,
+      "step": 2140
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0002,
+      "loss": 0.6553,
+      "step": 2150
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0002,
+      "loss": 0.6249,
+      "step": 2160
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0002,
+      "loss": 0.6631,
+      "step": 2170
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0002,
+      "loss": 0.6452,
+      "step": 2180
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0002,
+      "loss": 0.6713,
+      "step": 2190
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0002,
+      "loss": 0.642,
+      "step": 2200
+    },
+    {
+      "epoch": 1.06,
+      "eval_loss": 0.7102760672569275,
+      "eval_runtime": 96.7294,
+      "eval_samples_per_second": 10.338,
+      "eval_steps_per_second": 5.169,
+      "step": 2200
+    },
+    {
+      "epoch": 1.06,
+      "mmlu_eval_accuracy": 0.5048075372967952,
+      "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.5,
+      "mmlu_eval_accuracy_business_ethics": 0.6363636363636364,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.4827586206896552,
+      "mmlu_eval_accuracy_college_biology": 0.5,
+      "mmlu_eval_accuracy_college_chemistry": 0.125,
+      "mmlu_eval_accuracy_college_computer_science": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
+      "mmlu_eval_accuracy_college_medicine": 0.5,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
+      "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
+      "mmlu_eval_accuracy_econometrics": 0.3333333333333333,
+      "mmlu_eval_accuracy_electrical_engineering": 0.3125,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.4,
+      "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": 0.6111111111111112,
+      "mmlu_eval_accuracy_high_school_geography": 0.8636363636363636,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.4883720930232558,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
+      "mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8666666666666667,
+      "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": 0.7391304347826086,
+      "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
+      "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+      "mmlu_eval_accuracy_management": 0.5454545454545454,
+      "mmlu_eval_accuracy_marketing": 0.88,
+      "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
+      "mmlu_eval_accuracy_miscellaneous": 0.6627906976744186,
+      "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
+      "mmlu_eval_accuracy_moral_scenarios": 0.26,
+      "mmlu_eval_accuracy_nutrition": 0.6666666666666666,
+      "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
+      "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
+      "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": 0.4444444444444444,
+      "mmlu_eval_accuracy_sociology": 0.6818181818181818,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
+      "mmlu_eval_accuracy_virology": 0.5555555555555556,
+      "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
+      "mmlu_loss": 1.3245754189419685,
+      "step": 2200
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0002,
+      "loss": 0.6058,
+      "step": 2210
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0002,
+      "loss": 0.5853,
+      "step": 2220
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0002,
+      "loss": 0.6362,
+      "step": 2230
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0002,
+      "loss": 0.7074,
+      "step": 2240
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0002,
+      "loss": 0.6176,
+      "step": 2250
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0002,
+      "loss": 0.6252,
+      "step": 2260
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0002,
+      "loss": 0.6239,
+      "step": 2270
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0002,
+      "loss": 0.7087,
+      "step": 2280
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0002,
+      "loss": 0.5969,
+      "step": 2290
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.0002,
+      "loss": 0.6635,
+      "step": 2300
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.0002,
+      "loss": 0.6597,
+      "step": 2310
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0002,
+      "loss": 0.6492,
+      "step": 2320
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.0002,
+      "loss": 0.6499,
+      "step": 2330
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0002,
+      "loss": 0.6339,
+      "step": 2340
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0002,
+      "loss": 0.6691,
+      "step": 2350
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0002,
+      "loss": 0.6186,
+      "step": 2360
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0002,
+      "loss": 0.6209,
+      "step": 2370
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0002,
+      "loss": 0.6214,
+      "step": 2380
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0002,
+      "loss": 0.6427,
+      "step": 2390
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0002,
+      "loss": 0.6316,
+      "step": 2400
+    },
+    {
+      "epoch": 1.16,
+      "eval_loss": 0.7120669484138489,
+      "eval_runtime": 97.0421,
+      "eval_samples_per_second": 10.305,
+      "eval_steps_per_second": 5.152,
+      "step": 2400
+    },
+    {
+      "epoch": 1.16,
+      "mmlu_eval_accuracy": 0.49638584692650584,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.4375,
+      "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5172413793103449,
+      "mmlu_eval_accuracy_college_biology": 0.4375,
+      "mmlu_eval_accuracy_college_chemistry": 0.125,
+      "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
+      "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
+      "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
+      "mmlu_eval_accuracy_computer_security": 0.6363636363636364,
+      "mmlu_eval_accuracy_conceptual_physics": 0.5,
+      "mmlu_eval_accuracy_econometrics": 0.3333333333333333,
+      "mmlu_eval_accuracy_electrical_engineering": 0.3125,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.46875,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.2727272727272727,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_geography": 0.8636363636363636,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.5238095238095238,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.4418604651162791,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.13793103448275862,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
+      "mmlu_eval_accuracy_high_school_physics": 0.47058823529411764,
+      "mmlu_eval_accuracy_high_school_psychology": 0.85,
+      "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": 0.4166666666666667,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
+      "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+      "mmlu_eval_accuracy_management": 0.6363636363636364,
+      "mmlu_eval_accuracy_marketing": 0.8,
+      "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
+      "mmlu_eval_accuracy_miscellaneous": 0.6511627906976745,
+      "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
+      "mmlu_eval_accuracy_moral_scenarios": 0.23,
+      "mmlu_eval_accuracy_nutrition": 0.6060606060606061,
+      "mmlu_eval_accuracy_philosophy": 0.5294117647058824,
+      "mmlu_eval_accuracy_prehistory": 0.5428571428571428,
+      "mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
+      "mmlu_eval_accuracy_professional_law": 0.3352941176470588,
+      "mmlu_eval_accuracy_professional_medicine": 0.5161290322580645,
+      "mmlu_eval_accuracy_professional_psychology": 0.4492753623188406,
+      "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
+      "mmlu_eval_accuracy_security_studies": 0.4444444444444444,
+      "mmlu_eval_accuracy_sociology": 0.7727272727272727,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.4444444444444444,
+      "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
+      "mmlu_loss": 1.026068963382792,
+      "step": 2400
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0002,
+      "loss": 0.6239,
+      "step": 2410
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.0002,
+      "loss": 0.537,
+      "step": 2420
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0002,
+      "loss": 0.5653,
+      "step": 2430
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.0002,
+      "loss": 0.6398,
+      "step": 2440
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0002,
+      "loss": 0.608,
+      "step": 2450
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.0002,
+      "loss": 0.6217,
+      "step": 2460
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0002,
+      "loss": 0.6352,
+      "step": 2470
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0002,
+      "loss": 0.6209,
+      "step": 2480
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0002,
+      "loss": 0.6954,
+      "step": 2490
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0002,
+      "loss": 0.6297,
+      "step": 2500
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0002,
+      "loss": 0.6154,
+      "step": 2510
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0002,
+      "loss": 0.6456,
+      "step": 2520
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0002,
+      "loss": 0.609,
+      "step": 2530
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.0002,
+      "loss": 0.6478,
+      "step": 2540
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0002,
+      "loss": 0.5879,
+      "step": 2550
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0002,
+      "loss": 0.5983,
+      "step": 2560
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0002,
+      "loss": 0.6455,
+      "step": 2570
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0002,
+      "loss": 0.6138,
+      "step": 2580
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0002,
+      "loss": 0.6368,
+      "step": 2590
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0002,
+      "loss": 0.6387,
+      "step": 2600
+    },
+    {
+      "epoch": 1.25,
+      "eval_loss": 0.704391360282898,
+      "eval_runtime": 96.7567,
+      "eval_samples_per_second": 10.335,
+      "eval_steps_per_second": 5.168,
+      "step": 2600
+    },
+    {
+      "epoch": 1.25,
+      "mmlu_eval_accuracy": 0.517209162685207,
+      "mmlu_eval_accuracy_abstract_algebra": 0.45454545454545453,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.5,
+      "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.4827586206896552,
+      "mmlu_eval_accuracy_college_biology": 0.5625,
+      "mmlu_eval_accuracy_college_chemistry": 0.125,
+      "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": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.7272727272727273,
+      "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
+      "mmlu_eval_accuracy_econometrics": 0.3333333333333333,
+      "mmlu_eval_accuracy_electrical_engineering": 0.25,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.3170731707317073,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.6,
+      "mmlu_eval_accuracy_high_school_biology": 0.5,
+      "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": 0.8636363636363636,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.5116279069767442,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
+      "mmlu_eval_accuracy_high_school_physics": 0.5294117647058824,
+      "mmlu_eval_accuracy_high_school_psychology": 0.85,
+      "mmlu_eval_accuracy_high_school_statistics": 0.4782608695652174,
+      "mmlu_eval_accuracy_high_school_us_history": 0.5909090909090909,
+      "mmlu_eval_accuracy_high_school_world_history": 0.7307692307692307,
+      "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
+      "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
+      "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
+      "mmlu_eval_accuracy_management": 0.6363636363636364,
+      "mmlu_eval_accuracy_marketing": 0.84,
+      "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
+      "mmlu_eval_accuracy_miscellaneous": 0.6627906976744186,
+      "mmlu_eval_accuracy_moral_disputes": 0.42105263157894735,
+      "mmlu_eval_accuracy_moral_scenarios": 0.24,
+      "mmlu_eval_accuracy_nutrition": 0.696969696969697,
+      "mmlu_eval_accuracy_philosophy": 0.5588235294117647,
+      "mmlu_eval_accuracy_prehistory": 0.5428571428571428,
+      "mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
+      "mmlu_eval_accuracy_professional_law": 0.3235294117647059,
+      "mmlu_eval_accuracy_professional_medicine": 0.6129032258064516,
+      "mmlu_eval_accuracy_professional_psychology": 0.4927536231884058,
+      "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
+      "mmlu_eval_accuracy_security_studies": 0.48148148148148145,
+      "mmlu_eval_accuracy_sociology": 0.6818181818181818,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.4444444444444444,
+      "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
+      "mmlu_loss": 1.151616911158238,
+      "step": 2600
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0002,
+      "loss": 0.6325,
+      "step": 2610
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0002,
+      "loss": 0.6334,
+      "step": 2620
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0002,
+      "loss": 0.6847,
+      "step": 2630
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0002,
+      "loss": 0.6657,
+      "step": 2640
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0002,
+      "loss": 0.5858,
+      "step": 2650
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0002,
+      "loss": 0.6817,
+      "step": 2660
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0002,
+      "loss": 0.6486,
+      "step": 2670
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0002,
+      "loss": 0.6092,
+      "step": 2680
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0002,
+      "loss": 0.653,
+      "step": 2690
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0002,
+      "loss": 0.6174,
+      "step": 2700
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.0002,
+      "loss": 0.6605,
+      "step": 2710
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0002,
+      "loss": 0.6868,
+      "step": 2720
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0002,
+      "loss": 0.6653,
+      "step": 2730
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0002,
+      "loss": 0.6042,
+      "step": 2740
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0002,
+      "loss": 0.5862,
+      "step": 2750
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0002,
+      "loss": 0.647,
+      "step": 2760
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0002,
+      "loss": 0.6268,
+      "step": 2770
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0002,
+      "loss": 0.6198,
+      "step": 2780
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0002,
+      "loss": 0.6377,
+      "step": 2790
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0002,
+      "loss": 0.6406,
+      "step": 2800
+    },
+    {
+      "epoch": 1.35,
+      "eval_loss": 0.7013482451438904,
+      "eval_runtime": 96.4735,
+      "eval_samples_per_second": 10.366,
+      "eval_steps_per_second": 5.183,
+      "step": 2800
+    },
+    {
+      "epoch": 1.35,
+      "mmlu_eval_accuracy": 0.4925285551555681,
+      "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.4375,
+      "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5172413793103449,
+      "mmlu_eval_accuracy_college_biology": 0.5,
+      "mmlu_eval_accuracy_college_chemistry": 0.125,
+      "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": 0.36363636363636365,
+      "mmlu_eval_accuracy_computer_security": 0.6363636363636364,
+      "mmlu_eval_accuracy_conceptual_physics": 0.34615384615384615,
+      "mmlu_eval_accuracy_econometrics": 0.25,
+      "mmlu_eval_accuracy_electrical_engineering": 0.25,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
+      "mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.5,
+      "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": 0.5714285714285714,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.3953488372093023,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.13793103448275862,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
+      "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8333333333333334,
+      "mmlu_eval_accuracy_high_school_statistics": 0.34782608695652173,
+      "mmlu_eval_accuracy_high_school_us_history": 0.5909090909090909,
+      "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": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
+      "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
+      "mmlu_eval_accuracy_management": 0.6363636363636364,
+      "mmlu_eval_accuracy_marketing": 0.76,
+      "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
+      "mmlu_eval_accuracy_miscellaneous": 0.6511627906976745,
+      "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
+      "mmlu_eval_accuracy_moral_scenarios": 0.22,
+      "mmlu_eval_accuracy_nutrition": 0.7272727272727273,
+      "mmlu_eval_accuracy_philosophy": 0.5294117647058824,
+      "mmlu_eval_accuracy_prehistory": 0.5714285714285714,
+      "mmlu_eval_accuracy_professional_accounting": 0.3225806451612903,
+      "mmlu_eval_accuracy_professional_law": 0.34705882352941175,
+      "mmlu_eval_accuracy_professional_medicine": 0.5806451612903226,
+      "mmlu_eval_accuracy_professional_psychology": 0.463768115942029,
+      "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
+      "mmlu_eval_accuracy_security_studies": 0.48148148148148145,
+      "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": 1.2200810726800102,
+      "step": 2800
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0002,
+      "loss": 0.6398,
+      "step": 2810
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0002,
+      "loss": 0.6551,
+      "step": 2820
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.0002,
+      "loss": 0.6306,
+      "step": 2830
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0002,
+      "loss": 0.5925,
+      "step": 2840
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.0002,
+      "loss": 0.6522,
+      "step": 2850
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0002,
+      "loss": 0.6365,
+      "step": 2860
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0002,
+      "loss": 0.6312,
+      "step": 2870
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0002,
+      "loss": 0.6168,
+      "step": 2880
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0002,
+      "loss": 0.6917,
+      "step": 2890
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.0002,
+      "loss": 0.6114,
+      "step": 2900
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.0002,
+      "loss": 0.6612,
+      "step": 2910
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0002,
+      "loss": 0.5669,
+      "step": 2920
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0002,
+      "loss": 0.6473,
+      "step": 2930
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0002,
+      "loss": 0.6317,
+      "step": 2940
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0002,
+      "loss": 0.6611,
+      "step": 2950
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0002,
+      "loss": 0.6156,
+      "step": 2960
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0002,
+      "loss": 0.6514,
+      "step": 2970
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0002,
+      "loss": 0.6368,
+      "step": 2980
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0002,
+      "loss": 0.6121,
+      "step": 2990
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0002,
+      "loss": 0.6391,
+      "step": 3000
+    },
+    {
+      "epoch": 1.44,
+      "eval_loss": 0.7019394636154175,
+      "eval_runtime": 96.5558,
+      "eval_samples_per_second": 10.357,
+      "eval_steps_per_second": 5.178,
+      "step": 3000
+    },
+    {
+      "epoch": 1.44,
+      "mmlu_eval_accuracy": 0.4853465973418973,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.4375,
+      "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5172413793103449,
+      "mmlu_eval_accuracy_college_biology": 0.4375,
+      "mmlu_eval_accuracy_college_chemistry": 0.125,
+      "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
+      "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
+      "mmlu_eval_accuracy_econometrics": 0.08333333333333333,
+      "mmlu_eval_accuracy_electrical_engineering": 0.25,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.2926829268292683,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.4,
+      "mmlu_eval_accuracy_high_school_biology": 0.40625,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.22727272727272727,
+      "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": 0.5714285714285714,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.4418604651162791,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.3103448275862069,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.34615384615384615,
+      "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
+      "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": 0.6956521739130435,
+      "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
+      "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
+      "mmlu_eval_accuracy_management": 0.6363636363636364,
+      "mmlu_eval_accuracy_marketing": 0.8,
+      "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
+      "mmlu_eval_accuracy_miscellaneous": 0.6511627906976745,
+      "mmlu_eval_accuracy_moral_disputes": 0.5263157894736842,
+      "mmlu_eval_accuracy_moral_scenarios": 0.29,
+      "mmlu_eval_accuracy_nutrition": 0.696969696969697,
+      "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
+      "mmlu_eval_accuracy_prehistory": 0.45714285714285713,
+      "mmlu_eval_accuracy_professional_accounting": 0.41935483870967744,
+      "mmlu_eval_accuracy_professional_law": 0.32941176470588235,
+      "mmlu_eval_accuracy_professional_medicine": 0.6129032258064516,
+      "mmlu_eval_accuracy_professional_psychology": 0.463768115942029,
+      "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
+      "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
+      "mmlu_eval_accuracy_sociology": 0.7272727272727273,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
+      "mmlu_eval_accuracy_virology": 0.5,
+      "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+      "mmlu_loss": 1.1684038531297802,
+      "step": 3000
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0002,
+      "loss": 0.6525,
+      "step": 3010
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0002,
+      "loss": 0.6265,
+      "step": 3020
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0002,
+      "loss": 0.5704,
+      "step": 3030
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.0002,
+      "loss": 0.6634,
+      "step": 3040
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0002,
+      "loss": 0.6899,
+      "step": 3050
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0002,
+      "loss": 0.6469,
+      "step": 3060
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0002,
+      "loss": 0.6063,
+      "step": 3070
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0002,
+      "loss": 0.617,
+      "step": 3080
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0002,
+      "loss": 0.5917,
+      "step": 3090
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0002,
+      "loss": 0.6616,
+      "step": 3100
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0002,
+      "loss": 0.6719,
+      "step": 3110
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.0002,
+      "loss": 0.6399,
+      "step": 3120
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0002,
+      "loss": 0.6856,
+      "step": 3130
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.0002,
+      "loss": 0.5705,
+      "step": 3140
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0002,
+      "loss": 0.6613,
+      "step": 3150
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0002,
+      "loss": 0.6274,
+      "step": 3160
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0002,
+      "loss": 0.6658,
+      "step": 3170
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0002,
+      "loss": 0.6352,
+      "step": 3180
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0002,
+      "loss": 0.6471,
+      "step": 3190
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0002,
+      "loss": 0.632,
+      "step": 3200
+    },
+    {
+      "epoch": 1.54,
+      "eval_loss": 0.7007029056549072,
+      "eval_runtime": 96.7273,
+      "eval_samples_per_second": 10.338,
+      "eval_steps_per_second": 5.169,
+      "step": 3200
+    },
+    {
+      "epoch": 1.54,
+      "mmlu_eval_accuracy": 0.5016841158660759,
+      "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
+      "mmlu_eval_accuracy_anatomy": 0.7142857142857143,
+      "mmlu_eval_accuracy_astronomy": 0.5625,
+      "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.4827586206896552,
+      "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": 0.18181818181818182,
+      "mmlu_eval_accuracy_college_medicine": 0.5,
+      "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
+      "mmlu_eval_accuracy_computer_security": 0.6363636363636364,
+      "mmlu_eval_accuracy_conceptual_physics": 0.34615384615384615,
+      "mmlu_eval_accuracy_econometrics": 0.3333333333333333,
+      "mmlu_eval_accuracy_electrical_engineering": 0.25,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.4146341463414634,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.40625,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.2727272727272727,
+      "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": 0.46511627906976744,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
+      "mmlu_eval_accuracy_high_school_physics": 0.17647058823529413,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8666666666666667,
+      "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": 0.7391304347826086,
+      "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
+      "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
+      "mmlu_eval_accuracy_management": 0.6363636363636364,
+      "mmlu_eval_accuracy_marketing": 0.84,
+      "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
+      "mmlu_eval_accuracy_miscellaneous": 0.6976744186046512,
+      "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
+      "mmlu_eval_accuracy_moral_scenarios": 0.18,
+      "mmlu_eval_accuracy_nutrition": 0.7272727272727273,
+      "mmlu_eval_accuracy_philosophy": 0.5294117647058824,
+      "mmlu_eval_accuracy_prehistory": 0.5428571428571428,
+      "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
+      "mmlu_eval_accuracy_professional_law": 0.3588235294117647,
+      "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": 0.48148148148148145,
+      "mmlu_eval_accuracy_sociology": 0.7727272727272727,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.3333333333333333,
+      "mmlu_eval_accuracy_world_religions": 0.6842105263157895,
+      "mmlu_loss": 1.0545151401359791,
+      "step": 3200
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0002,
+      "loss": 0.6669,
+      "step": 3210
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0002,
+      "loss": 0.631,
+      "step": 3220
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0002,
+      "loss": 0.6271,
+      "step": 3230
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0002,
+      "loss": 0.6661,
+      "step": 3240
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0002,
+      "loss": 0.5809,
+      "step": 3250
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0002,
+      "loss": 0.5527,
+      "step": 3260
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0002,
+      "loss": 0.6357,
+      "step": 3270
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0002,
+      "loss": 0.6659,
+      "step": 3280
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0002,
+      "loss": 0.6133,
+      "step": 3290
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0002,
+      "loss": 0.6268,
+      "step": 3300
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0002,
+      "loss": 0.614,
+      "step": 3310
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.0002,
+      "loss": 0.6681,
+      "step": 3320
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.0002,
+      "loss": 0.6475,
+      "step": 3330
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0002,
+      "loss": 0.6277,
+      "step": 3340
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0002,
+      "loss": 0.6327,
+      "step": 3350
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.0002,
+      "loss": 0.5997,
+      "step": 3360
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.0002,
+      "loss": 0.6252,
+      "step": 3370
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0002,
+      "loss": 0.6061,
+      "step": 3380
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0002,
+      "loss": 0.5826,
+      "step": 3390
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.0002,
+      "loss": 0.6737,
+      "step": 3400
+    },
+    {
+      "epoch": 1.64,
+      "eval_loss": 0.7008338570594788,
+      "eval_runtime": 96.566,
+      "eval_samples_per_second": 10.356,
+      "eval_steps_per_second": 5.178,
+      "step": 3400
+    },
+    {
+      "epoch": 1.64,
+      "mmlu_eval_accuracy": 0.5059968735916802,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.5,
+      "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
+      "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": 0.18181818181818182,
+      "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
+      "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
+      "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
+      "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
+      "mmlu_eval_accuracy_econometrics": 0.4166666666666667,
+      "mmlu_eval_accuracy_electrical_engineering": 0.3125,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.4146341463414634,
+      "mmlu_eval_accuracy_formal_logic": 0.35714285714285715,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.5,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.4883720930232558,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
+      "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8666666666666667,
+      "mmlu_eval_accuracy_high_school_statistics": 0.4782608695652174,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
+      "mmlu_eval_accuracy_high_school_world_history": 0.6923076923076923,
+      "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
+      "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
+      "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+      "mmlu_eval_accuracy_management": 0.6363636363636364,
+      "mmlu_eval_accuracy_marketing": 0.84,
+      "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
+      "mmlu_eval_accuracy_miscellaneous": 0.6976744186046512,
+      "mmlu_eval_accuracy_moral_disputes": 0.4473684210526316,
+      "mmlu_eval_accuracy_moral_scenarios": 0.22,
+      "mmlu_eval_accuracy_nutrition": 0.7272727272727273,
+      "mmlu_eval_accuracy_philosophy": 0.5294117647058824,
+      "mmlu_eval_accuracy_prehistory": 0.5428571428571428,
+      "mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
+      "mmlu_eval_accuracy_professional_law": 0.35294117647058826,
+      "mmlu_eval_accuracy_professional_medicine": 0.5483870967741935,
+      "mmlu_eval_accuracy_professional_psychology": 0.4782608695652174,
+      "mmlu_eval_accuracy_public_relations": 0.5,
+      "mmlu_eval_accuracy_security_studies": 0.48148148148148145,
+      "mmlu_eval_accuracy_sociology": 0.6818181818181818,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.3888888888888889,
+      "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+      "mmlu_loss": 1.1631466560357544,
+      "step": 3400
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.0002,
+      "loss": 0.6199,
+      "step": 3410
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.0002,
+      "loss": 0.6898,
+      "step": 3420
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.0002,
+      "loss": 0.6113,
+      "step": 3430
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0002,
+      "loss": 0.6403,
+      "step": 3440
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0002,
+      "loss": 0.5836,
+      "step": 3450
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0002,
+      "loss": 0.5926,
+      "step": 3460
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0002,
+      "loss": 0.6327,
+      "step": 3470
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0002,
+      "loss": 0.6409,
+      "step": 3480
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.0002,
+      "loss": 0.676,
+      "step": 3490
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.0002,
+      "loss": 0.6035,
+      "step": 3500
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.0002,
+      "loss": 0.5925,
+      "step": 3510
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.0002,
+      "loss": 0.5939,
+      "step": 3520
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0002,
+      "loss": 0.6332,
+      "step": 3530
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0002,
+      "loss": 0.6247,
+      "step": 3540
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0002,
+      "loss": 0.6634,
+      "step": 3550
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.0002,
+      "loss": 0.6296,
+      "step": 3560
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0002,
+      "loss": 0.698,
+      "step": 3570
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0002,
+      "loss": 0.6047,
+      "step": 3580
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.0002,
+      "loss": 0.6075,
+      "step": 3590
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.0002,
+      "loss": 0.6569,
+      "step": 3600
+    },
+    {
+      "epoch": 1.73,
+      "eval_loss": 0.6949359774589539,
+      "eval_runtime": 96.648,
+      "eval_samples_per_second": 10.347,
+      "eval_steps_per_second": 5.173,
+      "step": 3600
+    },
+    {
+      "epoch": 1.73,
+      "mmlu_eval_accuracy": 0.4935483523034825,
+      "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
+      "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
+      "mmlu_eval_accuracy_astronomy": 0.5,
+      "mmlu_eval_accuracy_business_ethics": 0.36363636363636365,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5172413793103449,
+      "mmlu_eval_accuracy_college_biology": 0.4375,
+      "mmlu_eval_accuracy_college_chemistry": 0.375,
+      "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_medicine": 0.5454545454545454,
+      "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
+      "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
+      "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
+      "mmlu_eval_accuracy_econometrics": 0.3333333333333333,
+      "mmlu_eval_accuracy_electrical_engineering": 0.25,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
+      "mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.4375,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.4418604651162791,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
+      "mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8666666666666667,
+      "mmlu_eval_accuracy_high_school_statistics": 0.4782608695652174,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
+      "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": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
+      "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+      "mmlu_eval_accuracy_management": 0.6363636363636364,
+      "mmlu_eval_accuracy_marketing": 0.84,
+      "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
+      "mmlu_eval_accuracy_miscellaneous": 0.6511627906976745,
+      "mmlu_eval_accuracy_moral_disputes": 0.4473684210526316,
+      "mmlu_eval_accuracy_moral_scenarios": 0.23,
+      "mmlu_eval_accuracy_nutrition": 0.6363636363636364,
+      "mmlu_eval_accuracy_philosophy": 0.5588235294117647,
+      "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
+      "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
+      "mmlu_eval_accuracy_professional_law": 0.3588235294117647,
+      "mmlu_eval_accuracy_professional_medicine": 0.5483870967741935,
+      "mmlu_eval_accuracy_professional_psychology": 0.5072463768115942,
+      "mmlu_eval_accuracy_public_relations": 0.5,
+      "mmlu_eval_accuracy_security_studies": 0.48148148148148145,
+      "mmlu_eval_accuracy_sociology": 0.6363636363636364,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
+      "mmlu_eval_accuracy_virology": 0.3333333333333333,
+      "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
+      "mmlu_loss": 1.0990863862026765,
+      "step": 3600
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.0002,
+      "loss": 0.6306,
+      "step": 3610
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.0002,
+      "loss": 0.6212,
+      "step": 3620
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.0002,
+      "loss": 0.6428,
+      "step": 3630
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.0002,
+      "loss": 0.6247,
+      "step": 3640
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.0002,
+      "loss": 0.5999,
+      "step": 3650
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.0002,
+      "loss": 0.6153,
+      "step": 3660
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.0002,
+      "loss": 0.6349,
+      "step": 3670
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.0002,
+      "loss": 0.6589,
+      "step": 3680
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.0002,
+      "loss": 0.6412,
+      "step": 3690
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.0002,
+      "loss": 0.6172,
+      "step": 3700
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.0002,
+      "loss": 0.6902,
+      "step": 3710
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.0002,
+      "loss": 0.6463,
+      "step": 3720
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0002,
+      "loss": 0.5946,
+      "step": 3730
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0002,
+      "loss": 0.6402,
+      "step": 3740
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0002,
+      "loss": 0.6855,
+      "step": 3750
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0002,
+      "loss": 0.6514,
+      "step": 3760
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0002,
+      "loss": 0.6223,
+      "step": 3770
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.0002,
+      "loss": 0.6242,
+      "step": 3780
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.0002,
+      "loss": 0.661,
+      "step": 3790
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.0002,
+      "loss": 0.6234,
+      "step": 3800
+    },
+    {
+      "epoch": 1.83,
+      "eval_loss": 0.6974416971206665,
+      "eval_runtime": 96.5494,
+      "eval_samples_per_second": 10.357,
+      "eval_steps_per_second": 5.179,
+      "step": 3800
+    },
+    {
+      "epoch": 1.83,
+      "mmlu_eval_accuracy": 0.5132864378444664,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.4375,
+      "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.4827586206896552,
+      "mmlu_eval_accuracy_college_biology": 0.4375,
+      "mmlu_eval_accuracy_college_chemistry": 0.375,
+      "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_medicine": 0.5,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
+      "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
+      "mmlu_eval_accuracy_econometrics": 0.3333333333333333,
+      "mmlu_eval_accuracy_electrical_engineering": 0.25,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.3902439024390244,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.4,
+      "mmlu_eval_accuracy_high_school_biology": 0.4375,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.4883720930232558,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
+      "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
+      "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": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
+      "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+      "mmlu_eval_accuracy_management": 0.6363636363636364,
+      "mmlu_eval_accuracy_marketing": 0.88,
+      "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
+      "mmlu_eval_accuracy_miscellaneous": 0.686046511627907,
+      "mmlu_eval_accuracy_moral_disputes": 0.42105263157894735,
+      "mmlu_eval_accuracy_moral_scenarios": 0.24,
+      "mmlu_eval_accuracy_nutrition": 0.7575757575757576,
+      "mmlu_eval_accuracy_philosophy": 0.5588235294117647,
+      "mmlu_eval_accuracy_prehistory": 0.5142857142857142,
+      "mmlu_eval_accuracy_professional_accounting": 0.3870967741935484,
+      "mmlu_eval_accuracy_professional_law": 0.3411764705882353,
+      "mmlu_eval_accuracy_professional_medicine": 0.6129032258064516,
+      "mmlu_eval_accuracy_professional_psychology": 0.5072463768115942,
+      "mmlu_eval_accuracy_public_relations": 0.5,
+      "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
+      "mmlu_eval_accuracy_sociology": 0.6363636363636364,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.8181818181818182,
+      "mmlu_eval_accuracy_virology": 0.4444444444444444,
+      "mmlu_eval_accuracy_world_religions": 0.8421052631578947,
+      "mmlu_loss": 1.197790414094925,
+      "step": 3800
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.0002,
+      "loss": 0.5663,
+      "step": 3810
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.0002,
+      "loss": 0.6275,
+      "step": 3820
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.0002,
+      "loss": 0.6492,
+      "step": 3830
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0002,
+      "loss": 0.6502,
+      "step": 3840
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0002,
+      "loss": 0.6415,
+      "step": 3850
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0002,
+      "loss": 0.6535,
+      "step": 3860
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0002,
+      "loss": 0.6023,
+      "step": 3870
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0002,
+      "loss": 0.6052,
+      "step": 3880
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0002,
+      "loss": 0.6309,
+      "step": 3890
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.0002,
+      "loss": 0.5763,
+      "step": 3900
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.0002,
+      "loss": 0.6168,
+      "step": 3910
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.0002,
+      "loss": 0.6134,
+      "step": 3920
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.0002,
+      "loss": 0.6515,
+      "step": 3930
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0002,
+      "loss": 0.6632,
+      "step": 3940
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0002,
+      "loss": 0.6673,
+      "step": 3950
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.0002,
+      "loss": 0.6382,
+      "step": 3960
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.0002,
+      "loss": 0.6102,
+      "step": 3970
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.0002,
+      "loss": 0.6235,
+      "step": 3980
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.0002,
+      "loss": 0.6015,
+      "step": 3990
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0002,
+      "loss": 0.6513,
+      "step": 4000
+    },
+    {
+      "epoch": 1.93,
+      "eval_loss": 0.6935309767723083,
+      "eval_runtime": 96.5168,
+      "eval_samples_per_second": 10.361,
+      "eval_steps_per_second": 5.18,
+      "step": 4000
+    },
+    {
+      "epoch": 1.93,
+      "mmlu_eval_accuracy": 0.5048661718395844,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.375,
+      "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.4827586206896552,
+      "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": 0.18181818181818182,
+      "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
+      "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
+      "mmlu_eval_accuracy_econometrics": 0.4166666666666667,
+      "mmlu_eval_accuracy_electrical_engineering": 0.25,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.3902439024390244,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.3,
+      "mmlu_eval_accuracy_high_school_biology": 0.40625,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.2727272727272727,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.46511627906976744,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
+      "mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
+      "mmlu_eval_accuracy_high_school_psychology": 0.85,
+      "mmlu_eval_accuracy_high_school_statistics": 0.43478260869565216,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
+      "mmlu_eval_accuracy_high_school_world_history": 0.7307692307692307,
+      "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
+      "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+      "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+      "mmlu_eval_accuracy_management": 0.6363636363636364,
+      "mmlu_eval_accuracy_marketing": 0.8,
+      "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
+      "mmlu_eval_accuracy_miscellaneous": 0.6976744186046512,
+      "mmlu_eval_accuracy_moral_disputes": 0.4473684210526316,
+      "mmlu_eval_accuracy_moral_scenarios": 0.23,
+      "mmlu_eval_accuracy_nutrition": 0.6363636363636364,
+      "mmlu_eval_accuracy_philosophy": 0.4117647058823529,
+      "mmlu_eval_accuracy_prehistory": 0.45714285714285713,
+      "mmlu_eval_accuracy_professional_accounting": 0.45161290322580644,
+      "mmlu_eval_accuracy_professional_law": 0.34705882352941175,
+      "mmlu_eval_accuracy_professional_medicine": 0.6451612903225806,
+      "mmlu_eval_accuracy_professional_psychology": 0.5072463768115942,
+      "mmlu_eval_accuracy_public_relations": 0.5,
+      "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
+      "mmlu_eval_accuracy_sociology": 0.6818181818181818,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
+      "mmlu_eval_accuracy_virology": 0.5,
+      "mmlu_eval_accuracy_world_religions": 0.8421052631578947,
+      "mmlu_loss": 1.0965017748719408,
+      "step": 4000
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0002,
+      "loss": 0.5982,
+      "step": 4010
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0002,
+      "loss": 0.6548,
+      "step": 4020
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.0002,
+      "loss": 0.6693,
+      "step": 4030
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.0002,
+      "loss": 0.5996,
+      "step": 4040
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.0002,
+      "loss": 0.647,
+      "step": 4050
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.0002,
+      "loss": 0.6158,
+      "step": 4060
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.0002,
+      "loss": 0.6132,
+      "step": 4070
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.0002,
+      "loss": 0.6379,
+      "step": 4080
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.0002,
+      "loss": 0.6187,
+      "step": 4090
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.0002,
+      "loss": 0.628,
+      "step": 4100
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.0002,
+      "loss": 0.6404,
+      "step": 4110
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.0002,
+      "loss": 0.6648,
+      "step": 4120
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.0002,
+      "loss": 0.6049,
+      "step": 4130
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.0002,
+      "loss": 0.5937,
+      "step": 4140
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.0002,
+      "loss": 0.635,
+      "step": 4150
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.0002,
+      "loss": 0.5441,
+      "step": 4160
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.0002,
+      "loss": 0.5747,
+      "step": 4170
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.0002,
+      "loss": 0.493,
+      "step": 4180
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.0002,
+      "loss": 0.5512,
+      "step": 4190
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.0002,
+      "loss": 0.5196,
+      "step": 4200
+    },
+    {
+      "epoch": 2.02,
+      "eval_loss": 0.7101156115531921,
+      "eval_runtime": 96.495,
+      "eval_samples_per_second": 10.363,
+      "eval_steps_per_second": 5.182,
+      "step": 4200
+    },
+    {
+      "epoch": 2.02,
+      "mmlu_eval_accuracy": 0.5156976828735957,
+      "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.5,
+      "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.4482758620689655,
+      "mmlu_eval_accuracy_college_biology": 0.5,
+      "mmlu_eval_accuracy_college_chemistry": 0.25,
+      "mmlu_eval_accuracy_college_computer_science": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
+      "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
+      "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
+      "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
+      "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
+      "mmlu_eval_accuracy_econometrics": 0.4166666666666667,
+      "mmlu_eval_accuracy_electrical_engineering": 0.3125,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
+      "mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
+      "mmlu_eval_accuracy_global_facts": 0.4,
+      "mmlu_eval_accuracy_high_school_biology": 0.4375,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_geography": 0.8636363636363636,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.5238095238095238,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.5116279069767442,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.27586206896551724,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.38461538461538464,
+      "mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
+      "mmlu_eval_accuracy_high_school_psychology": 0.7833333333333333,
+      "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": 0.7391304347826086,
+      "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+      "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
+      "mmlu_eval_accuracy_management": 0.6363636363636364,
+      "mmlu_eval_accuracy_marketing": 0.84,
+      "mmlu_eval_accuracy_medical_genetics": 1.0,
+      "mmlu_eval_accuracy_miscellaneous": 0.686046511627907,
+      "mmlu_eval_accuracy_moral_disputes": 0.42105263157894735,
+      "mmlu_eval_accuracy_moral_scenarios": 0.24,
+      "mmlu_eval_accuracy_nutrition": 0.6666666666666666,
+      "mmlu_eval_accuracy_philosophy": 0.5,
+      "mmlu_eval_accuracy_prehistory": 0.5428571428571428,
+      "mmlu_eval_accuracy_professional_accounting": 0.3870967741935484,
+      "mmlu_eval_accuracy_professional_law": 0.3411764705882353,
+      "mmlu_eval_accuracy_professional_medicine": 0.6451612903225806,
+      "mmlu_eval_accuracy_professional_psychology": 0.463768115942029,
+      "mmlu_eval_accuracy_public_relations": 0.5,
+      "mmlu_eval_accuracy_security_studies": 0.6296296296296297,
+      "mmlu_eval_accuracy_sociology": 0.7272727272727273,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.8181818181818182,
+      "mmlu_eval_accuracy_virology": 0.4444444444444444,
+      "mmlu_eval_accuracy_world_religions": 0.8421052631578947,
+      "mmlu_loss": 1.172741430371455,
+      "step": 4200
     }
   ],
   "max_steps": 10000,
   "num_train_epochs": 5,
-  "total_flos": 5.283117051146404e+17,
+  "total_flos": 1.1097720102135398e+18,
   "trial_name": null,
   "trial_params": null
 }