Training in progress, step 6800

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:acaf6023a66bfe9eb2692b7a2d808e465670c18e7d11d715f587f3b14ec28504
+oid sha256:52bc52a3ae980b15a5d966c90b9fddede4efc3a1dac10dcc3f21f2381be89d73
 size 319977229

{checkpoint-4600 → checkpoint-6600/adapter_model/adapter_model}/adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:61fdb1233c27fd4122f604cf922df45c2733289bf2ddd2460e0c6270562fac97
+oid sha256:acaf6023a66bfe9eb2692b7a2d808e465670c18e7d11d715f587f3b14ec28504
 size 319977229

checkpoint-6800/README.md

@@ -0,0 +1,20 @@
+---
+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
+### Framework versions
+
+
+- PEFT 0.4.0

checkpoint-6800/adapter_config.json

@@ -0,0 +1,26 @@
+{
+  "auto_mapping": null,
+  "base_model_name_or_path": "pankajmathur/orca_mini_v3_7b",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "lora_alpha": 16.0,
+  "lora_dropout": 0.1,
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 64,
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "k_proj",
+    "down_proj",
+    "up_proj",
+    "o_proj",
+    "v_proj",
+    "q_proj"
+  ],
+  "task_type": "CAUSAL_LM"
+}

checkpoint-6800/adapter_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:52bc52a3ae980b15a5d966c90b9fddede4efc3a1dac10dcc3f21f2381be89d73
+size 319977229

{checkpoint-4600 → checkpoint-6800}/optimizer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:29e9284f0052845ae2622c488aeff7f078d5615dbdd2b47e8d6320c0b60e48c3
+oid sha256:bbb16b20d7d40f377de3eb71ea4b4212ad28e688681ea7355e46fe26813bd0e0
 size 1279539973

{checkpoint-4600 → checkpoint-6800}/rng_state.pth

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c2dec82a1ac0107dea8cf3d6084aadb26d6b32ec931066589c2fbcbf5613e28f
+oid sha256:a8d9a0678923710b2485d11ec408e8498503d6dab1d31edec8084361895bdea5
 size 14511

{checkpoint-4600 → checkpoint-6800}/scheduler.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:793c554376f8e7f2de92b146fd20e985a7a5b03ad4002c60e108f3677dee0ff3
+oid sha256:97c7e7304629a0d5b32d77774ba8769b4783d1dab8e735fe47e12df568c740c2
 size 627

{checkpoint-4600 → checkpoint-6800}/trainer_state.json

@@ -1,8 +1,8 @@
 {
-  "best_metric": 0.7327941656112671,
-  "best_model_checkpoint": "experts/expert-4/checkpoint-4400",
-  "epoch": 0.5142250293443631,
-  "global_step": 4600,
+  "best_metric": 0.7229225039482117,
+  "best_model_checkpoint": "experts/expert-4/checkpoint-6800",
+  "epoch": 0.7601587390307976,
+  "global_step": 6800,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -4399,11 +4399,2112 @@
       "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
       "mmlu_loss": 1.159490028031501,
       "step": 4600
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002,
+      "loss": 0.7769,
+      "step": 4610
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002,
+      "loss": 0.7641,
+      "step": 4620
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002,
+      "loss": 0.7344,
+      "step": 4630
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002,
+      "loss": 0.6414,
+      "step": 4640
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002,
+      "loss": 0.7067,
+      "step": 4650
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002,
+      "loss": 0.8444,
+      "step": 4660
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002,
+      "loss": 0.7274,
+      "step": 4670
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002,
+      "loss": 0.7244,
+      "step": 4680
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0002,
+      "loss": 0.6626,
+      "step": 4690
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002,
+      "loss": 0.8,
+      "step": 4700
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002,
+      "loss": 0.8715,
+      "step": 4710
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002,
+      "loss": 0.6898,
+      "step": 4720
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002,
+      "loss": 0.7025,
+      "step": 4730
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002,
+      "loss": 0.7526,
+      "step": 4740
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002,
+      "loss": 0.7053,
+      "step": 4750
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002,
+      "loss": 0.7028,
+      "step": 4760
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002,
+      "loss": 0.6821,
+      "step": 4770
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0002,
+      "loss": 0.7686,
+      "step": 4780
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0002,
+      "loss": 0.7203,
+      "step": 4790
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0002,
+      "loss": 0.6715,
+      "step": 4800
+    },
+    {
+      "epoch": 0.54,
+      "eval_loss": 0.7351219058036804,
+      "eval_runtime": 133.719,
+      "eval_samples_per_second": 7.478,
+      "eval_steps_per_second": 3.739,
+      "step": 4800
+    },
+    {
+      "epoch": 0.54,
+      "mmlu_eval_accuracy": 0.5029711403549669,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.7142857142857143,
+      "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": 0.125,
+      "mmlu_eval_accuracy_college_computer_science": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_mathematics": 0.09090909090909091,
+      "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
+      "mmlu_eval_accuracy_conceptual_physics": 0.5384615384615384,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.25,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.4146341463414634,
+      "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
+      "mmlu_eval_accuracy_global_facts": 0.6,
+      "mmlu_eval_accuracy_high_school_biology": 0.46875,
+      "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": 0.8636363636363636,
+      "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": 0.5769230769230769,
+      "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8666666666666667,
+      "mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
+      "mmlu_eval_accuracy_high_school_us_history": 0.5909090909090909,
+      "mmlu_eval_accuracy_high_school_world_history": 0.5384615384615384,
+      "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": 0.45454545454545453,
+      "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": 0.47368421052631576,
+      "mmlu_eval_accuracy_moral_scenarios": 0.24,
+      "mmlu_eval_accuracy_nutrition": 0.696969696969697,
+      "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
+      "mmlu_eval_accuracy_prehistory": 0.5142857142857142,
+      "mmlu_eval_accuracy_professional_accounting": 0.41935483870967744,
+      "mmlu_eval_accuracy_professional_law": 0.35294117647058826,
+      "mmlu_eval_accuracy_professional_medicine": 0.5483870967741935,
+      "mmlu_eval_accuracy_professional_psychology": 0.5072463768115942,
+      "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
+      "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": 0.7368421052631579,
+      "mmlu_loss": 1.030532050965349,
+      "step": 4800
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0002,
+      "loss": 0.7474,
+      "step": 4810
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0002,
+      "loss": 0.7798,
+      "step": 4820
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0002,
+      "loss": 0.7953,
+      "step": 4830
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0002,
+      "loss": 0.7642,
+      "step": 4840
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0002,
+      "loss": 0.7026,
+      "step": 4850
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0002,
+      "loss": 0.6345,
+      "step": 4860
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0002,
+      "loss": 0.7413,
+      "step": 4870
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002,
+      "loss": 0.7181,
+      "step": 4880
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002,
+      "loss": 0.7847,
+      "step": 4890
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002,
+      "loss": 0.6536,
+      "step": 4900
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002,
+      "loss": 0.7542,
+      "step": 4910
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002,
+      "loss": 0.6633,
+      "step": 4920
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002,
+      "loss": 0.8086,
+      "step": 4930
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002,
+      "loss": 0.6857,
+      "step": 4940
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002,
+      "loss": 0.7208,
+      "step": 4950
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002,
+      "loss": 0.7221,
+      "step": 4960
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0002,
+      "loss": 0.7655,
+      "step": 4970
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0002,
+      "loss": 0.7159,
+      "step": 4980
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0002,
+      "loss": 0.7949,
+      "step": 4990
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0002,
+      "loss": 0.7852,
+      "step": 5000
+    },
+    {
+      "epoch": 0.56,
+      "eval_loss": 0.7331032752990723,
+      "eval_runtime": 133.7651,
+      "eval_samples_per_second": 7.476,
+      "eval_steps_per_second": 3.738,
+      "step": 5000
+    },
+    {
+      "epoch": 0.56,
+      "mmlu_eval_accuracy": 0.5192070831352871,
+      "mmlu_eval_accuracy_abstract_algebra": 0.45454545454545453,
+      "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
+      "mmlu_eval_accuracy_astronomy": 0.5,
+      "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5862068965517241,
+      "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": 0.09090909090909091,
+      "mmlu_eval_accuracy_college_medicine": 0.5,
+      "mmlu_eval_accuracy_college_physics": 0.5454545454545454,
+      "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
+      "mmlu_eval_accuracy_conceptual_physics": 0.6153846153846154,
+      "mmlu_eval_accuracy_econometrics": 0.25,
+      "mmlu_eval_accuracy_electrical_engineering": 0.375,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.2926829268292683,
+      "mmlu_eval_accuracy_formal_logic": 0.14285714285714285,
+      "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": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_geography": 0.9090909090909091,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.4418604651162791,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.3103448275862069,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
+      "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": 0.6363636363636364,
+      "mmlu_eval_accuracy_high_school_world_history": 0.6538461538461539,
+      "mmlu_eval_accuracy_human_aging": 0.6086956521739131,
+      "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": 0.8181818181818182,
+      "mmlu_eval_accuracy_miscellaneous": 0.686046511627907,
+      "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
+      "mmlu_eval_accuracy_moral_scenarios": 0.25,
+      "mmlu_eval_accuracy_nutrition": 0.7575757575757576,
+      "mmlu_eval_accuracy_philosophy": 0.5588235294117647,
+      "mmlu_eval_accuracy_prehistory": 0.5428571428571428,
+      "mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
+      "mmlu_eval_accuracy_professional_law": 0.3176470588235294,
+      "mmlu_eval_accuracy_professional_medicine": 0.5483870967741935,
+      "mmlu_eval_accuracy_professional_psychology": 0.5652173913043478,
+      "mmlu_eval_accuracy_public_relations": 0.5,
+      "mmlu_eval_accuracy_security_studies": 0.48148148148148145,
+      "mmlu_eval_accuracy_sociology": 0.7272727272727273,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
+      "mmlu_eval_accuracy_virology": 0.5555555555555556,
+      "mmlu_eval_accuracy_world_religions": 0.8947368421052632,
+      "mmlu_loss": 0.9450486395751216,
+      "step": 5000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0002,
+      "loss": 0.7456,
+      "step": 5010
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0002,
+      "loss": 0.6729,
+      "step": 5020
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0002,
+      "loss": 0.7849,
+      "step": 5030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0002,
+      "loss": 0.7429,
+      "step": 5040
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0002,
+      "loss": 0.7743,
+      "step": 5050
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0002,
+      "loss": 0.6442,
+      "step": 5060
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0002,
+      "loss": 0.7174,
+      "step": 5070
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0002,
+      "loss": 0.7322,
+      "step": 5080
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0002,
+      "loss": 0.8523,
+      "step": 5090
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0002,
+      "loss": 0.7459,
+      "step": 5100
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0002,
+      "loss": 0.7132,
+      "step": 5110
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0002,
+      "loss": 0.6701,
+      "step": 5120
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0002,
+      "loss": 0.6646,
+      "step": 5130
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0002,
+      "loss": 0.724,
+      "step": 5140
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002,
+      "loss": 0.7932,
+      "step": 5150
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002,
+      "loss": 0.7287,
+      "step": 5160
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002,
+      "loss": 0.7211,
+      "step": 5170
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002,
+      "loss": 0.7424,
+      "step": 5180
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002,
+      "loss": 0.789,
+      "step": 5190
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002,
+      "loss": 0.8518,
+      "step": 5200
+    },
+    {
+      "epoch": 0.58,
+      "eval_loss": 0.7311785817146301,
+      "eval_runtime": 133.7882,
+      "eval_samples_per_second": 7.475,
+      "eval_steps_per_second": 3.737,
+      "step": 5200
+    },
+    {
+      "epoch": 0.58,
+      "mmlu_eval_accuracy": 0.5220292787339565,
+      "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
+      "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
+      "mmlu_eval_accuracy_astronomy": 0.625,
+      "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
+      "mmlu_eval_accuracy_college_biology": 0.375,
+      "mmlu_eval_accuracy_college_chemistry": 0.125,
+      "mmlu_eval_accuracy_college_computer_science": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_medicine": 0.45454545454545453,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
+      "mmlu_eval_accuracy_conceptual_physics": 0.5769230769230769,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.5,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.43902439024390244,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.5625,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.4090909090909091,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.7222222222222222,
+      "mmlu_eval_accuracy_high_school_geography": 0.8636363636363636,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.3488372093023256,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.3448275862068966,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5769230769230769,
+      "mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8333333333333334,
+      "mmlu_eval_accuracy_high_school_statistics": 0.43478260869565216,
+      "mmlu_eval_accuracy_high_school_us_history": 0.7272727272727273,
+      "mmlu_eval_accuracy_high_school_world_history": 0.6153846153846154,
+      "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
+      "mmlu_eval_accuracy_human_sexuality": 0.5833333333333334,
+      "mmlu_eval_accuracy_international_law": 0.6923076923076923,
+      "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": 0.9090909090909091,
+      "mmlu_eval_accuracy_miscellaneous": 0.6976744186046512,
+      "mmlu_eval_accuracy_moral_disputes": 0.42105263157894735,
+      "mmlu_eval_accuracy_moral_scenarios": 0.22,
+      "mmlu_eval_accuracy_nutrition": 0.6666666666666666,
+      "mmlu_eval_accuracy_philosophy": 0.5,
+      "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
+      "mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
+      "mmlu_eval_accuracy_professional_law": 0.34705882352941175,
+      "mmlu_eval_accuracy_professional_medicine": 0.5806451612903226,
+      "mmlu_eval_accuracy_professional_psychology": 0.4782608695652174,
+      "mmlu_eval_accuracy_public_relations": 0.4166666666666667,
+      "mmlu_eval_accuracy_security_studies": 0.5555555555555556,
+      "mmlu_eval_accuracy_sociology": 0.7727272727272727,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.7272727272727273,
+      "mmlu_eval_accuracy_virology": 0.5555555555555556,
+      "mmlu_eval_accuracy_world_religions": 0.8947368421052632,
+      "mmlu_loss": 1.055438042542953,
+      "step": 5200
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002,
+      "loss": 0.78,
+      "step": 5210
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002,
+      "loss": 0.6611,
+      "step": 5220
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.0002,
+      "loss": 0.7485,
+      "step": 5230
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0002,
+      "loss": 0.6969,
+      "step": 5240
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0002,
+      "loss": 0.6866,
+      "step": 5250
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0002,
+      "loss": 0.7333,
+      "step": 5260
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0002,
+      "loss": 0.6648,
+      "step": 5270
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0002,
+      "loss": 0.7576,
+      "step": 5280
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0002,
+      "loss": 0.7222,
+      "step": 5290
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0002,
+      "loss": 0.6631,
+      "step": 5300
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0002,
+      "loss": 0.7768,
+      "step": 5310
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0002,
+      "loss": 0.7548,
+      "step": 5320
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0002,
+      "loss": 0.8303,
+      "step": 5330
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0002,
+      "loss": 0.7654,
+      "step": 5340
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0002,
+      "loss": 0.7272,
+      "step": 5350
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0002,
+      "loss": 0.658,
+      "step": 5360
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0002,
+      "loss": 0.7735,
+      "step": 5370
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0002,
+      "loss": 0.6828,
+      "step": 5380
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0002,
+      "loss": 0.7471,
+      "step": 5390
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0002,
+      "loss": 0.7102,
+      "step": 5400
+    },
+    {
+      "epoch": 0.6,
+      "eval_loss": 0.7356892228126526,
+      "eval_runtime": 133.8949,
+      "eval_samples_per_second": 7.469,
+      "eval_steps_per_second": 3.734,
+      "step": 5400
+    },
+    {
+      "epoch": 0.6,
+      "mmlu_eval_accuracy": 0.492140950643189,
+      "mmlu_eval_accuracy_abstract_algebra": 0.45454545454545453,
+      "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
+      "mmlu_eval_accuracy_astronomy": 0.625,
+      "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.4827586206896552,
+      "mmlu_eval_accuracy_college_biology": 0.375,
+      "mmlu_eval_accuracy_college_chemistry": 0.0,
+      "mmlu_eval_accuracy_college_computer_science": 0.5454545454545454,
+      "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
+      "mmlu_eval_accuracy_college_medicine": 0.3181818181818182,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
+      "mmlu_eval_accuracy_conceptual_physics": 0.6153846153846154,
+      "mmlu_eval_accuracy_econometrics": 0.08333333333333333,
+      "mmlu_eval_accuracy_electrical_engineering": 0.4375,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.3,
+      "mmlu_eval_accuracy_high_school_biology": 0.375,
+      "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": 0.9090909090909091,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.32558139534883723,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.27586206896551724,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
+      "mmlu_eval_accuracy_high_school_physics": 0.17647058823529413,
+      "mmlu_eval_accuracy_high_school_psychology": 0.7166666666666667,
+      "mmlu_eval_accuracy_high_school_statistics": 0.34782608695652173,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
+      "mmlu_eval_accuracy_high_school_world_history": 0.5384615384615384,
+      "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
+      "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+      "mmlu_eval_accuracy_international_law": 0.7692307692307693,
+      "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+      "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": 0.6744186046511628,
+      "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
+      "mmlu_eval_accuracy_moral_scenarios": 0.25,
+      "mmlu_eval_accuracy_nutrition": 0.696969696969697,
+      "mmlu_eval_accuracy_philosophy": 0.5294117647058824,
+      "mmlu_eval_accuracy_prehistory": 0.45714285714285713,
+      "mmlu_eval_accuracy_professional_accounting": 0.45161290322580644,
+      "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": 0.5,
+      "mmlu_eval_accuracy_security_studies": 0.4074074074074074,
+      "mmlu_eval_accuracy_sociology": 0.8181818181818182,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.5555555555555556,
+      "mmlu_eval_accuracy_world_religions": 0.8421052631578947,
+      "mmlu_loss": 0.9140872581236978,
+      "step": 5400
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0002,
+      "loss": 0.7793,
+      "step": 5410
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0002,
+      "loss": 0.8121,
+      "step": 5420
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0002,
+      "loss": 0.6735,
+      "step": 5430
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0002,
+      "loss": 0.7642,
+      "step": 5440
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0002,
+      "loss": 0.639,
+      "step": 5450
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0002,
+      "loss": 0.7216,
+      "step": 5460
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0002,
+      "loss": 0.5886,
+      "step": 5470
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0002,
+      "loss": 0.7149,
+      "step": 5480
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0002,
+      "loss": 0.8108,
+      "step": 5490
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0002,
+      "loss": 0.7307,
+      "step": 5500
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0002,
+      "loss": 0.5537,
+      "step": 5510
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0002,
+      "loss": 0.734,
+      "step": 5520
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0002,
+      "loss": 0.7498,
+      "step": 5530
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0002,
+      "loss": 0.7673,
+      "step": 5540
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0002,
+      "loss": 0.7387,
+      "step": 5550
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0002,
+      "loss": 0.5929,
+      "step": 5560
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0002,
+      "loss": 0.6754,
+      "step": 5570
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0002,
+      "loss": 0.7214,
+      "step": 5580
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0002,
+      "loss": 0.7545,
+      "step": 5590
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0002,
+      "loss": 0.8808,
+      "step": 5600
+    },
+    {
+      "epoch": 0.63,
+      "eval_loss": 0.7327717542648315,
+      "eval_runtime": 133.9533,
+      "eval_samples_per_second": 7.465,
+      "eval_steps_per_second": 3.733,
+      "step": 5600
+    },
+    {
+      "epoch": 0.63,
+      "mmlu_eval_accuracy": 0.4815276776532176,
+      "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": 0.5862068965517241,
+      "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": 0.4090909090909091,
+      "mmlu_eval_accuracy_college_physics": 0.5454545454545454,
+      "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
+      "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": 0.3170731707317073,
+      "mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
+      "mmlu_eval_accuracy_global_facts": 0.3,
+      "mmlu_eval_accuracy_high_school_biology": 0.34375,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.22727272727272727,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.47619047619047616,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.3953488372093023,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5384615384615384,
+      "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
+      "mmlu_eval_accuracy_high_school_psychology": 0.7333333333333333,
+      "mmlu_eval_accuracy_high_school_statistics": 0.2608695652173913,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
+      "mmlu_eval_accuracy_high_school_world_history": 0.6153846153846154,
+      "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": 0.6666666666666666,
+      "mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
+      "mmlu_eval_accuracy_management": 0.6363636363636364,
+      "mmlu_eval_accuracy_marketing": 0.72,
+      "mmlu_eval_accuracy_medical_genetics": 0.9090909090909091,
+      "mmlu_eval_accuracy_miscellaneous": 0.6395348837209303,
+      "mmlu_eval_accuracy_moral_disputes": 0.5526315789473685,
+      "mmlu_eval_accuracy_moral_scenarios": 0.27,
+      "mmlu_eval_accuracy_nutrition": 0.6363636363636364,
+      "mmlu_eval_accuracy_philosophy": 0.4117647058823529,
+      "mmlu_eval_accuracy_prehistory": 0.42857142857142855,
+      "mmlu_eval_accuracy_professional_accounting": 0.3225806451612903,
+      "mmlu_eval_accuracy_professional_law": 0.36470588235294116,
+      "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": 0.7272727272727273,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.5,
+      "mmlu_eval_accuracy_world_religions": 0.631578947368421,
+      "mmlu_loss": 1.0687501477199186,
+      "step": 5600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0002,
+      "loss": 0.8089,
+      "step": 5610
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0002,
+      "loss": 0.7176,
+      "step": 5620
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0002,
+      "loss": 0.7577,
+      "step": 5630
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0002,
+      "loss": 0.639,
+      "step": 5640
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0002,
+      "loss": 0.7752,
+      "step": 5650
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0002,
+      "loss": 0.7445,
+      "step": 5660
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0002,
+      "loss": 0.7355,
+      "step": 5670
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0002,
+      "loss": 0.6133,
+      "step": 5680
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002,
+      "loss": 0.7981,
+      "step": 5690
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002,
+      "loss": 0.7717,
+      "step": 5700
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002,
+      "loss": 0.7383,
+      "step": 5710
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002,
+      "loss": 0.7865,
+      "step": 5720
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002,
+      "loss": 0.7674,
+      "step": 5730
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002,
+      "loss": 0.6665,
+      "step": 5740
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002,
+      "loss": 0.7025,
+      "step": 5750
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002,
+      "loss": 0.8471,
+      "step": 5760
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002,
+      "loss": 0.6446,
+      "step": 5770
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002,
+      "loss": 0.6375,
+      "step": 5780
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002,
+      "loss": 0.8063,
+      "step": 5790
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002,
+      "loss": 0.7628,
+      "step": 5800
+    },
+    {
+      "epoch": 0.65,
+      "eval_loss": 0.7311202883720398,
+      "eval_runtime": 133.8445,
+      "eval_samples_per_second": 7.471,
+      "eval_steps_per_second": 3.736,
+      "step": 5800
+    },
+    {
+      "epoch": 0.65,
+      "mmlu_eval_accuracy": 0.51251895005488,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.625,
+      "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5517241379310345,
+      "mmlu_eval_accuracy_college_biology": 0.5625,
+      "mmlu_eval_accuracy_college_chemistry": 0.0,
+      "mmlu_eval_accuracy_college_computer_science": 0.5454545454545454,
+      "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
+      "mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
+      "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
+      "mmlu_eval_accuracy_conceptual_physics": 0.5769230769230769,
+      "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": 0.14285714285714285,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.46875,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.36363636363636365,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.7222222222222222,
+      "mmlu_eval_accuracy_high_school_geography": 0.8636363636363636,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.5238095238095238,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.32558139534883723,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5384615384615384,
+      "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8166666666666667,
+      "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
+      "mmlu_eval_accuracy_high_school_world_history": 0.5769230769230769,
+      "mmlu_eval_accuracy_human_aging": 0.6086956521739131,
+      "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
+      "mmlu_eval_accuracy_international_law": 0.6923076923076923,
+      "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": 1.0,
+      "mmlu_eval_accuracy_miscellaneous": 0.6511627906976745,
+      "mmlu_eval_accuracy_moral_disputes": 0.42105263157894735,
+      "mmlu_eval_accuracy_moral_scenarios": 0.24,
+      "mmlu_eval_accuracy_nutrition": 0.6060606060606061,
+      "mmlu_eval_accuracy_philosophy": 0.5294117647058824,
+      "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
+      "mmlu_eval_accuracy_professional_accounting": 0.4838709677419355,
+      "mmlu_eval_accuracy_professional_law": 0.35294117647058826,
+      "mmlu_eval_accuracy_professional_medicine": 0.6129032258064516,
+      "mmlu_eval_accuracy_professional_psychology": 0.5362318840579711,
+      "mmlu_eval_accuracy_public_relations": 0.5,
+      "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
+      "mmlu_eval_accuracy_sociology": 0.7727272727272727,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.8181818181818182,
+      "mmlu_eval_accuracy_virology": 0.6111111111111112,
+      "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
+      "mmlu_loss": 0.9438617095234499,
+      "step": 5800
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002,
+      "loss": 0.7504,
+      "step": 5810
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002,
+      "loss": 0.787,
+      "step": 5820
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002,
+      "loss": 0.8835,
+      "step": 5830
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002,
+      "loss": 0.6397,
+      "step": 5840
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0002,
+      "loss": 0.7431,
+      "step": 5850
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002,
+      "loss": 0.7642,
+      "step": 5860
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002,
+      "loss": 0.7145,
+      "step": 5870
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002,
+      "loss": 0.6849,
+      "step": 5880
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002,
+      "loss": 0.7892,
+      "step": 5890
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002,
+      "loss": 0.8894,
+      "step": 5900
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002,
+      "loss": 0.7159,
+      "step": 5910
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002,
+      "loss": 0.7377,
+      "step": 5920
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002,
+      "loss": 0.6979,
+      "step": 5930
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002,
+      "loss": 0.7603,
+      "step": 5940
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002,
+      "loss": 0.7723,
+      "step": 5950
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002,
+      "loss": 0.6461,
+      "step": 5960
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002,
+      "loss": 0.6774,
+      "step": 5970
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002,
+      "loss": 0.8045,
+      "step": 5980
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002,
+      "loss": 0.7093,
+      "step": 5990
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002,
+      "loss": 0.7327,
+      "step": 6000
+    },
+    {
+      "epoch": 0.67,
+      "eval_loss": 0.7304873466491699,
+      "eval_runtime": 133.9409,
+      "eval_samples_per_second": 7.466,
+      "eval_steps_per_second": 3.733,
+      "step": 6000
+    },
+    {
+      "epoch": 0.67,
+      "mmlu_eval_accuracy": 0.5033617036874508,
+      "mmlu_eval_accuracy_abstract_algebra": 0.45454545454545453,
+      "mmlu_eval_accuracy_anatomy": 0.5,
+      "mmlu_eval_accuracy_astronomy": 0.5625,
+      "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": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
+      "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
+      "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+      "mmlu_eval_accuracy_computer_security": 0.6363636363636364,
+      "mmlu_eval_accuracy_conceptual_physics": 0.46153846153846156,
+      "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+      "mmlu_eval_accuracy_electrical_engineering": 0.3125,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.6,
+      "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": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_geography": 0.8636363636363636,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.3953488372093023,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
+      "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8166666666666667,
+      "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
+      "mmlu_eval_accuracy_high_school_world_history": 0.6153846153846154,
+      "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
+      "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
+      "mmlu_eval_accuracy_international_law": 0.6923076923076923,
+      "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+      "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
+      "mmlu_eval_accuracy_management": 0.6363636363636364,
+      "mmlu_eval_accuracy_marketing": 0.84,
+      "mmlu_eval_accuracy_medical_genetics": 1.0,
+      "mmlu_eval_accuracy_miscellaneous": 0.6744186046511628,
+      "mmlu_eval_accuracy_moral_disputes": 0.4473684210526316,
+      "mmlu_eval_accuracy_moral_scenarios": 0.22,
+      "mmlu_eval_accuracy_nutrition": 0.5757575757575758,
+      "mmlu_eval_accuracy_philosophy": 0.5,
+      "mmlu_eval_accuracy_prehistory": 0.5142857142857142,
+      "mmlu_eval_accuracy_professional_accounting": 0.45161290322580644,
+      "mmlu_eval_accuracy_professional_law": 0.38235294117647056,
+      "mmlu_eval_accuracy_professional_medicine": 0.5161290322580645,
+      "mmlu_eval_accuracy_professional_psychology": 0.5362318840579711,
+      "mmlu_eval_accuracy_public_relations": 0.5833333333333334,
+      "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": 0.5555555555555556,
+      "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
+      "mmlu_loss": 0.9632397257861518,
+      "step": 6000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002,
+      "loss": 0.7012,
+      "step": 6010
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002,
+      "loss": 0.7146,
+      "step": 6020
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0002,
+      "loss": 0.6171,
+      "step": 6030
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002,
+      "loss": 0.7061,
+      "step": 6040
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002,
+      "loss": 0.8499,
+      "step": 6050
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002,
+      "loss": 0.8027,
+      "step": 6060
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002,
+      "loss": 0.7858,
+      "step": 6070
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002,
+      "loss": 0.7095,
+      "step": 6080
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002,
+      "loss": 0.7967,
+      "step": 6090
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002,
+      "loss": 0.7728,
+      "step": 6100
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002,
+      "loss": 0.7238,
+      "step": 6110
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002,
+      "loss": 0.6484,
+      "step": 6120
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002,
+      "loss": 0.6462,
+      "step": 6130
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002,
+      "loss": 0.6792,
+      "step": 6140
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002,
+      "loss": 0.6605,
+      "step": 6150
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002,
+      "loss": 0.8883,
+      "step": 6160
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002,
+      "loss": 0.8081,
+      "step": 6170
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002,
+      "loss": 0.8386,
+      "step": 6180
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002,
+      "loss": 0.8036,
+      "step": 6190
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002,
+      "loss": 0.7298,
+      "step": 6200
+    },
+    {
+      "epoch": 0.69,
+      "eval_loss": 0.7294755578041077,
+      "eval_runtime": 133.8367,
+      "eval_samples_per_second": 7.472,
+      "eval_steps_per_second": 3.736,
+      "step": 6200
+    },
+    {
+      "epoch": 0.69,
+      "mmlu_eval_accuracy": 0.4707100974830736,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
+      "mmlu_eval_accuracy_astronomy": 0.5625,
+      "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.4482758620689655,
+      "mmlu_eval_accuracy_college_biology": 0.375,
+      "mmlu_eval_accuracy_college_chemistry": 0.125,
+      "mmlu_eval_accuracy_college_computer_science": 0.09090909090909091,
+      "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
+      "mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
+      "mmlu_eval_accuracy_computer_security": 0.2727272727272727,
+      "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": 0.3170731707317073,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.4,
+      "mmlu_eval_accuracy_high_school_biology": 0.34375,
+      "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": 0.8181818181818182,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.4418604651162791,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.3076923076923077,
+      "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8166666666666667,
+      "mmlu_eval_accuracy_high_school_statistics": 0.2608695652173913,
+      "mmlu_eval_accuracy_high_school_us_history": 0.7272727272727273,
+      "mmlu_eval_accuracy_high_school_world_history": 0.6923076923076923,
+      "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
+      "mmlu_eval_accuracy_human_sexuality": 0.25,
+      "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": 0.36363636363636365,
+      "mmlu_eval_accuracy_management": 0.6363636363636364,
+      "mmlu_eval_accuracy_marketing": 0.88,
+      "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": 0.21,
+      "mmlu_eval_accuracy_nutrition": 0.6666666666666666,
+      "mmlu_eval_accuracy_philosophy": 0.4411764705882353,
+      "mmlu_eval_accuracy_prehistory": 0.45714285714285713,
+      "mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
+      "mmlu_eval_accuracy_professional_law": 0.3176470588235294,
+      "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": 0.48148148148148145,
+      "mmlu_eval_accuracy_sociology": 0.5909090909090909,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
+      "mmlu_eval_accuracy_virology": 0.5,
+      "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
+      "mmlu_loss": 1.1253070027492065,
+      "step": 6200
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.0002,
+      "loss": 0.8025,
+      "step": 6210
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002,
+      "loss": 0.6823,
+      "step": 6220
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002,
+      "loss": 0.7048,
+      "step": 6230
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002,
+      "loss": 0.6192,
+      "step": 6240
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002,
+      "loss": 0.7806,
+      "step": 6250
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002,
+      "loss": 0.7558,
+      "step": 6260
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002,
+      "loss": 0.5937,
+      "step": 6270
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002,
+      "loss": 0.7113,
+      "step": 6280
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002,
+      "loss": 0.6004,
+      "step": 6290
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002,
+      "loss": 0.8532,
+      "step": 6300
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002,
+      "loss": 0.7288,
+      "step": 6310
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002,
+      "loss": 0.7663,
+      "step": 6320
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002,
+      "loss": 0.6625,
+      "step": 6330
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002,
+      "loss": 0.7516,
+      "step": 6340
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002,
+      "loss": 0.6731,
+      "step": 6350
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002,
+      "loss": 0.7451,
+      "step": 6360
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002,
+      "loss": 0.7792,
+      "step": 6370
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002,
+      "loss": 0.6916,
+      "step": 6380
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002,
+      "loss": 0.6542,
+      "step": 6390
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0002,
+      "loss": 0.661,
+      "step": 6400
+    },
+    {
+      "epoch": 0.72,
+      "eval_loss": 0.7265727519989014,
+      "eval_runtime": 133.6855,
+      "eval_samples_per_second": 7.48,
+      "eval_steps_per_second": 3.74,
+      "step": 6400
+    },
+    {
+      "epoch": 0.72,
+      "mmlu_eval_accuracy": 0.49809858794637457,
+      "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.5,
+      "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5862068965517241,
+      "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": 0.5,
+      "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
+      "mmlu_eval_accuracy_computer_security": 0.5454545454545454,
+      "mmlu_eval_accuracy_conceptual_physics": 0.5,
+      "mmlu_eval_accuracy_econometrics": 0.25,
+      "mmlu_eval_accuracy_electrical_engineering": 0.3125,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.3170731707317073,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.5,
+      "mmlu_eval_accuracy_high_school_biology": 0.5625,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.13636363636363635,
+      "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": 0.5238095238095238,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.46511627906976744,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5769230769230769,
+      "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8333333333333334,
+      "mmlu_eval_accuracy_high_school_statistics": 0.4782608695652174,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
+      "mmlu_eval_accuracy_high_school_world_history": 0.6153846153846154,
+      "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": 0.45454545454545453,
+      "mmlu_eval_accuracy_logical_fallacies": 0.6666666666666666,
+      "mmlu_eval_accuracy_machine_learning": 0.36363636363636365,
+      "mmlu_eval_accuracy_management": 0.6363636363636364,
+      "mmlu_eval_accuracy_marketing": 0.84,
+      "mmlu_eval_accuracy_medical_genetics": 0.8181818181818182,
+      "mmlu_eval_accuracy_miscellaneous": 0.6744186046511628,
+      "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
+      "mmlu_eval_accuracy_moral_scenarios": 0.22,
+      "mmlu_eval_accuracy_nutrition": 0.6060606060606061,
+      "mmlu_eval_accuracy_philosophy": 0.4117647058823529,
+      "mmlu_eval_accuracy_prehistory": 0.6857142857142857,
+      "mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
+      "mmlu_eval_accuracy_professional_law": 0.38823529411764707,
+      "mmlu_eval_accuracy_professional_medicine": 0.45161290322580644,
+      "mmlu_eval_accuracy_professional_psychology": 0.5507246376811594,
+      "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": 0.6363636363636364,
+      "mmlu_eval_accuracy_virology": 0.5555555555555556,
+      "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+      "mmlu_loss": 0.9850227478745712,
+      "step": 6400
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0002,
+      "loss": 0.76,
+      "step": 6410
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0002,
+      "loss": 0.777,
+      "step": 6420
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0002,
+      "loss": 0.7474,
+      "step": 6430
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0002,
+      "loss": 0.7083,
+      "step": 6440
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0002,
+      "loss": 0.7153,
+      "step": 6450
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0002,
+      "loss": 0.6572,
+      "step": 6460
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0002,
+      "loss": 0.6472,
+      "step": 6470
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0002,
+      "loss": 0.8041,
+      "step": 6480
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0002,
+      "loss": 0.7238,
+      "step": 6490
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0002,
+      "loss": 0.779,
+      "step": 6500
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0002,
+      "loss": 0.8241,
+      "step": 6510
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0002,
+      "loss": 0.6237,
+      "step": 6520
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0002,
+      "loss": 0.7347,
+      "step": 6530
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0002,
+      "loss": 0.6243,
+      "step": 6540
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0002,
+      "loss": 0.7694,
+      "step": 6550
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0002,
+      "loss": 0.7634,
+      "step": 6560
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0002,
+      "loss": 0.7448,
+      "step": 6570
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0002,
+      "loss": 0.7376,
+      "step": 6580
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0002,
+      "loss": 0.7703,
+      "step": 6590
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0002,
+      "loss": 0.6609,
+      "step": 6600
+    },
+    {
+      "epoch": 0.74,
+      "eval_loss": 0.7248952984809875,
+      "eval_runtime": 133.8066,
+      "eval_samples_per_second": 7.473,
+      "eval_steps_per_second": 3.737,
+      "step": 6600
+    },
+    {
+      "epoch": 0.74,
+      "mmlu_eval_accuracy": 0.5149231107622494,
+      "mmlu_eval_accuracy_abstract_algebra": 0.45454545454545453,
+      "mmlu_eval_accuracy_anatomy": 0.6428571428571429,
+      "mmlu_eval_accuracy_astronomy": 0.5625,
+      "mmlu_eval_accuracy_business_ethics": 0.45454545454545453,
+      "mmlu_eval_accuracy_clinical_knowledge": 0.5862068965517241,
+      "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": 0.09090909090909091,
+      "mmlu_eval_accuracy_college_medicine": 0.5,
+      "mmlu_eval_accuracy_college_physics": 0.2727272727272727,
+      "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": 0.2857142857142857,
+      "mmlu_eval_accuracy_global_facts": 0.6,
+      "mmlu_eval_accuracy_high_school_biology": 0.5,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.18181818181818182,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_european_history": 0.6666666666666666,
+      "mmlu_eval_accuracy_high_school_geography": 0.7272727272727273,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.5238095238095238,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.4186046511627907,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5769230769230769,
+      "mmlu_eval_accuracy_high_school_physics": 0.17647058823529413,
+      "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": 0.6538461538461539,
+      "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
+      "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
+      "mmlu_eval_accuracy_international_law": 0.8461538461538461,
+      "mmlu_eval_accuracy_jurisprudence": 0.5454545454545454,
+      "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": 0.6395348837209303,
+      "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
+      "mmlu_eval_accuracy_moral_scenarios": 0.26,
+      "mmlu_eval_accuracy_nutrition": 0.6060606060606061,
+      "mmlu_eval_accuracy_philosophy": 0.5,
+      "mmlu_eval_accuracy_prehistory": 0.5428571428571428,
+      "mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
+      "mmlu_eval_accuracy_professional_law": 0.38823529411764707,
+      "mmlu_eval_accuracy_professional_medicine": 0.5483870967741935,
+      "mmlu_eval_accuracy_professional_psychology": 0.5072463768115942,
+      "mmlu_eval_accuracy_public_relations": 0.6666666666666666,
+      "mmlu_eval_accuracy_security_studies": 0.5555555555555556,
+      "mmlu_eval_accuracy_sociology": 0.8181818181818182,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.9090909090909091,
+      "mmlu_eval_accuracy_virology": 0.5555555555555556,
+      "mmlu_eval_accuracy_world_religions": 0.631578947368421,
+      "mmlu_loss": 0.8850108603545642,
+      "step": 6600
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0002,
+      "loss": 0.7221,
+      "step": 6610
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0002,
+      "loss": 0.75,
+      "step": 6620
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0002,
+      "loss": 0.8407,
+      "step": 6630
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0002,
+      "loss": 0.6769,
+      "step": 6640
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0002,
+      "loss": 0.8187,
+      "step": 6650
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0002,
+      "loss": 0.8102,
+      "step": 6660
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0002,
+      "loss": 0.7222,
+      "step": 6670
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0002,
+      "loss": 0.7202,
+      "step": 6680
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0002,
+      "loss": 0.8201,
+      "step": 6690
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0002,
+      "loss": 0.5917,
+      "step": 6700
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0002,
+      "loss": 0.6594,
+      "step": 6710
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0002,
+      "loss": 0.6152,
+      "step": 6720
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0002,
+      "loss": 0.8191,
+      "step": 6730
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0002,
+      "loss": 0.7112,
+      "step": 6740
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0002,
+      "loss": 0.6968,
+      "step": 6750
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0002,
+      "loss": 0.651,
+      "step": 6760
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0002,
+      "loss": 0.8289,
+      "step": 6770
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0002,
+      "loss": 0.8347,
+      "step": 6780
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0002,
+      "loss": 0.7234,
+      "step": 6790
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0002,
+      "loss": 0.7227,
+      "step": 6800
+    },
+    {
+      "epoch": 0.76,
+      "eval_loss": 0.7229225039482117,
+      "eval_runtime": 133.823,
+      "eval_samples_per_second": 7.473,
+      "eval_steps_per_second": 3.736,
+      "step": 6800
+    },
+    {
+      "epoch": 0.76,
+      "mmlu_eval_accuracy": 0.5152214474699063,
+      "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+      "mmlu_eval_accuracy_anatomy": 0.7142857142857143,
+      "mmlu_eval_accuracy_astronomy": 0.5625,
+      "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": 0.6363636363636364,
+      "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
+      "mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
+      "mmlu_eval_accuracy_college_physics": 0.36363636363636365,
+      "mmlu_eval_accuracy_computer_security": 0.45454545454545453,
+      "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
+      "mmlu_eval_accuracy_econometrics": 0.3333333333333333,
+      "mmlu_eval_accuracy_electrical_engineering": 0.3125,
+      "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
+      "mmlu_eval_accuracy_formal_logic": 0.21428571428571427,
+      "mmlu_eval_accuracy_global_facts": 0.4,
+      "mmlu_eval_accuracy_high_school_biology": 0.375,
+      "mmlu_eval_accuracy_high_school_chemistry": 0.22727272727272727,
+      "mmlu_eval_accuracy_high_school_computer_science": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
+      "mmlu_eval_accuracy_high_school_geography": 0.8636363636363636,
+      "mmlu_eval_accuracy_high_school_government_and_politics": 0.7142857142857143,
+      "mmlu_eval_accuracy_high_school_macroeconomics": 0.46511627906976744,
+      "mmlu_eval_accuracy_high_school_mathematics": 0.27586206896551724,
+      "mmlu_eval_accuracy_high_school_microeconomics": 0.5384615384615384,
+      "mmlu_eval_accuracy_high_school_physics": 0.23529411764705882,
+      "mmlu_eval_accuracy_high_school_psychology": 0.8166666666666667,
+      "mmlu_eval_accuracy_high_school_statistics": 0.4782608695652174,
+      "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
+      "mmlu_eval_accuracy_high_school_world_history": 0.6538461538461539,
+      "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
+      "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+      "mmlu_eval_accuracy_international_law": 0.9230769230769231,
+      "mmlu_eval_accuracy_jurisprudence": 0.45454545454545453,
+      "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": 0.9090909090909091,
+      "mmlu_eval_accuracy_miscellaneous": 0.627906976744186,
+      "mmlu_eval_accuracy_moral_disputes": 0.4473684210526316,
+      "mmlu_eval_accuracy_moral_scenarios": 0.22,
+      "mmlu_eval_accuracy_nutrition": 0.6363636363636364,
+      "mmlu_eval_accuracy_philosophy": 0.5,
+      "mmlu_eval_accuracy_prehistory": 0.5142857142857142,
+      "mmlu_eval_accuracy_professional_accounting": 0.45161290322580644,
+      "mmlu_eval_accuracy_professional_law": 0.35294117647058826,
+      "mmlu_eval_accuracy_professional_medicine": 0.4838709677419355,
+      "mmlu_eval_accuracy_professional_psychology": 0.5507246376811594,
+      "mmlu_eval_accuracy_public_relations": 0.4166666666666667,
+      "mmlu_eval_accuracy_security_studies": 0.48148148148148145,
+      "mmlu_eval_accuracy_sociology": 0.8181818181818182,
+      "mmlu_eval_accuracy_us_foreign_policy": 0.9090909090909091,
+      "mmlu_eval_accuracy_virology": 0.5555555555555556,
+      "mmlu_eval_accuracy_world_religions": 0.8421052631578947,
+      "mmlu_loss": 1.011317597901541,
+      "step": 6800
     }
   ],
   "max_steps": 10000,
   "num_train_epochs": 2,
-  "total_flos": 4.5715341246745805e+17,
+  "total_flos": 6.746118287016591e+17,
   "trial_name": null,
   "trial_params": null
 }