Training in progress, step 1400

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:e344083300bd30eb9777f0a0328191d02ed3d81426dd12bf6e9ae3e704352b5c
+oid sha256:a15f9071fb58e7536acc43bd77f28c5ba1b32f292fb78714878b1d91f0cb7fae
 size 319977229

checkpoint-1400/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-1400/adapter_config.json

@@ -0,0 +1,26 @@
+{
+ "auto_mapping": null,
+ "base_model_name_or_path": "NousResearch/Nous-Hermes-llama-2-7b",
+ "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": [
+ "q_proj",
+ "gate_proj",
+ "up_proj",
+ "k_proj",
+ "down_proj",
+ "o_proj",
+ "v_proj"
+ ],
+ "task_type": "CAUSAL_LM"
+}

checkpoint-1400/adapter_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a15f9071fb58e7536acc43bd77f28c5ba1b32f292fb78714878b1d91f0cb7fae
+size 319977229

checkpoint-1400/added_tokens.json

@@ -0,0 +1,3 @@
+{
+ "<pad>": 32000
+}

checkpoint-1400/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e136fef06aa3f8d2e4ce73515d1e211dd70771172645e5a30b6434a46f1c549f
+size 1279539973

checkpoint-1400/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f908c5be94b2d3e92a4438d6fde316c54345eb4331c1bfbaafd34f1c7b9d8e85
+size 14511

checkpoint-1400/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d1351871b6e069a7a4437616d40b913152ecc2cd046e996fcfeb300a3cf60638
+size 627

checkpoint-1400/special_tokens_map.json

@@ -0,0 +1,6 @@
+{
+ "bos_token": "<s>",
+ "eos_token": "</s>",
+ "pad_token": "<unk>",
+ "unk_token": "<unk>"
+}

checkpoint-1400/tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
+size 499723

checkpoint-1400/tokenizer_config.json

@@ -0,0 +1,35 @@
+{
+ "add_bos_token": true,
+ "add_eos_token": false,
+ "bos_token": {
+ "__type": "AddedToken",
+ "content": "<s>",
+ "lstrip": false,
+ "normalized": true,
+ "rstrip": false,
+ "single_word": false
+ },
+ "clean_up_tokenization_spaces": false,
+ "eos_token": {
+ "__type": "AddedToken",
+ "content": "</s>",
+ "lstrip": false,
+ "normalized": true,
+ "rstrip": false,
+ "single_word": false
+ },
+ "legacy": false,
+ "model_max_length": 1000000000000000019884624838656,
+ "pad_token": null,
+ "padding_side": "right",
+ "sp_model_kwargs": {},
+ "tokenizer_class": "LlamaTokenizer",
+ "unk_token": {
+ "__type": "AddedToken",
+ "content": "<unk>",
+ "lstrip": false,
+ "normalized": true,
+ "rstrip": false,
+ "single_word": false
+ }
+}

checkpoint-1400/trainer_state.json

@@ -0,0 +1,1353 @@
+{
+ "best_metric": 0.7248261570930481,
+ "best_model_checkpoint": "./output_v2/7b_cluster011_Nous-Hermes-llama-2-7b_partitioned_v3_standardized_011/checkpoint-800",
+ "epoch": 3.2388663967611335,
+ "global_step": 1400,
+ "is_hyper_param_search": false,
+ "is_local_process_zero": true,
+ "is_world_process_zero": true,
+ "log_history": [
+ {
+ "epoch": 0.02,
+ "learning_rate": 0.0002,
+ "loss": 0.7957,
+ "step": 10
+ },
+ {
+ "epoch": 0.05,
+ "learning_rate": 0.0002,
+ "loss": 0.718,
+ "step": 20
+ },
+ {
+ "epoch": 0.07,
+ "learning_rate": 0.0002,
+ "loss": 0.73,
+ "step": 30
+ },
+ {
+ "epoch": 0.09,
+ "learning_rate": 0.0002,
+ "loss": 0.7826,
+ "step": 40
+ },
+ {
+ "epoch": 0.12,
+ "learning_rate": 0.0002,
+ "loss": 0.7013,
+ "step": 50
+ },
+ {
+ "epoch": 0.14,
+ "learning_rate": 0.0002,
+ "loss": 0.7353,
+ "step": 60
+ },
+ {
+ "epoch": 0.16,
+ "learning_rate": 0.0002,
+ "loss": 0.6666,
+ "step": 70
+ },
+ {
+ "epoch": 0.19,
+ "learning_rate": 0.0002,
+ "loss": 0.7649,
+ "step": 80
+ },
+ {
+ "epoch": 0.21,
+ "learning_rate": 0.0002,
+ "loss": 0.7018,
+ "step": 90
+ },
+ {
+ "epoch": 0.23,
+ "learning_rate": 0.0002,
+ "loss": 0.7173,
+ "step": 100
+ },
+ {
+ "epoch": 0.25,
+ "learning_rate": 0.0002,
+ "loss": 0.7857,
+ "step": 110
+ },
+ {
+ "epoch": 0.28,
+ "learning_rate": 0.0002,
+ "loss": 0.7154,
+ "step": 120
+ },
+ {
+ "epoch": 0.3,
+ "learning_rate": 0.0002,
+ "loss": 0.7485,
+ "step": 130
+ },
+ {
+ "epoch": 0.32,
+ "learning_rate": 0.0002,
+ "loss": 0.7114,
+ "step": 140
+ },
+ {
+ "epoch": 0.35,
+ "learning_rate": 0.0002,
+ "loss": 0.7333,
+ "step": 150
+ },
+ {
+ "epoch": 0.37,
+ "learning_rate": 0.0002,
+ "loss": 0.6549,
+ "step": 160
+ },
+ {
+ "epoch": 0.39,
+ "learning_rate": 0.0002,
+ "loss": 0.6765,
+ "step": 170
+ },
+ {
+ "epoch": 0.42,
+ "learning_rate": 0.0002,
+ "loss": 0.677,
+ "step": 180
+ },
+ {
+ "epoch": 0.44,
+ "learning_rate": 0.0002,
+ "loss": 0.6763,
+ "step": 190
+ },
+ {
+ "epoch": 0.46,
+ "learning_rate": 0.0002,
+ "loss": 0.6638,
+ "step": 200
+ },
+ {
+ "epoch": 0.46,
+ "eval_loss": 0.7327473163604736,
+ "eval_runtime": 246.3779,
+ "eval_samples_per_second": 4.059,
+ "eval_steps_per_second": 2.029,
+ "step": 200
+ },
+ {
+ "epoch": 0.46,
+ "mmlu_eval_accuracy": 0.4592376175825003,
+ "mmlu_eval_accuracy_abstract_algebra": 0.2727272727272727,
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
+ "mmlu_eval_accuracy_astronomy": 0.375,
+ "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
+ "mmlu_eval_accuracy_clinical_knowledge": 0.4482758620689655,
+ "mmlu_eval_accuracy_college_biology": 0.4375,
+ "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": 0.2727272727272727,
+ "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+ "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
+ "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+ "mmlu_eval_accuracy_electrical_engineering": 0.4375,
+ "mmlu_eval_accuracy_elementary_mathematics": 0.2926829268292683,
+ "mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
+ "mmlu_eval_accuracy_global_facts": 0.4,
+ "mmlu_eval_accuracy_high_school_biology": 0.34375,
+ "mmlu_eval_accuracy_high_school_chemistry": 0.4090909090909091,
+ "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
+ "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
+ "mmlu_eval_accuracy_high_school_geography": 0.7272727272727273,
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.6666666666666666,
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.32558139534883723,
+ "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.38461538461538464,
+ "mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
+ "mmlu_eval_accuracy_high_school_psychology": 0.7333333333333333,
+ "mmlu_eval_accuracy_high_school_statistics": 0.391304347826087,
+ "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
+ "mmlu_eval_accuracy_high_school_world_history": 0.5384615384615384,
+ "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
+ "mmlu_eval_accuracy_international_law": 0.6923076923076923,
+ "mmlu_eval_accuracy_jurisprudence": 0.5454545454545454,
+ "mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
+ "mmlu_eval_accuracy_machine_learning": 0.2727272727272727,
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
+ "mmlu_eval_accuracy_marketing": 0.72,
+ "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+ "mmlu_eval_accuracy_miscellaneous": 0.7093023255813954,
+ "mmlu_eval_accuracy_moral_disputes": 0.39473684210526316,
+ "mmlu_eval_accuracy_moral_scenarios": 0.24,
+ "mmlu_eval_accuracy_nutrition": 0.5757575757575758,
+ "mmlu_eval_accuracy_philosophy": 0.38235294117647056,
+ "mmlu_eval_accuracy_prehistory": 0.45714285714285713,
+ "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
+ "mmlu_eval_accuracy_professional_law": 0.34705882352941175,
+ "mmlu_eval_accuracy_professional_medicine": 0.41935483870967744,
+ "mmlu_eval_accuracy_professional_psychology": 0.391304347826087,
+ "mmlu_eval_accuracy_public_relations": 0.4166666666666667,
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
+ "mmlu_eval_accuracy_sociology": 0.7272727272727273,
+ "mmlu_eval_accuracy_us_foreign_policy": 0.45454545454545453,
+ "mmlu_eval_accuracy_virology": 0.4444444444444444,
+ "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+ "mmlu_loss": 1.173365458184683,
+ "step": 200
+ },
+ {
+ "epoch": 0.49,
+ "learning_rate": 0.0002,
+ "loss": 0.6849,
+ "step": 210
+ },
+ {
+ "epoch": 0.51,
+ "learning_rate": 0.0002,
+ "loss": 0.7275,
+ "step": 220
+ },
+ {
+ "epoch": 0.53,
+ "learning_rate": 0.0002,
+ "loss": 0.6976,
+ "step": 230
+ },
+ {
+ "epoch": 0.56,
+ "learning_rate": 0.0002,
+ "loss": 0.6896,
+ "step": 240
+ },
+ {
+ "epoch": 0.58,
+ "learning_rate": 0.0002,
+ "loss": 0.6831,
+ "step": 250
+ },
+ {
+ "epoch": 0.6,
+ "learning_rate": 0.0002,
+ "loss": 0.8049,
+ "step": 260
+ },
+ {
+ "epoch": 0.62,
+ "learning_rate": 0.0002,
+ "loss": 0.6878,
+ "step": 270
+ },
+ {
+ "epoch": 0.65,
+ "learning_rate": 0.0002,
+ "loss": 0.6679,
+ "step": 280
+ },
+ {
+ "epoch": 0.67,
+ "learning_rate": 0.0002,
+ "loss": 0.6808,
+ "step": 290
+ },
+ {
+ "epoch": 0.69,
+ "learning_rate": 0.0002,
+ "loss": 0.7648,
+ "step": 300
+ },
+ {
+ "epoch": 0.72,
+ "learning_rate": 0.0002,
+ "loss": 0.7605,
+ "step": 310
+ },
+ {
+ "epoch": 0.74,
+ "learning_rate": 0.0002,
+ "loss": 0.7504,
+ "step": 320
+ },
+ {
+ "epoch": 0.76,
+ "learning_rate": 0.0002,
+ "loss": 0.7853,
+ "step": 330
+ },
+ {
+ "epoch": 0.79,
+ "learning_rate": 0.0002,
+ "loss": 0.7272,
+ "step": 340
+ },
+ {
+ "epoch": 0.81,
+ "learning_rate": 0.0002,
+ "loss": 0.6934,
+ "step": 350
+ },
+ {
+ "epoch": 0.83,
+ "learning_rate": 0.0002,
+ "loss": 0.7053,
+ "step": 360
+ },
+ {
+ "epoch": 0.86,
+ "learning_rate": 0.0002,
+ "loss": 0.7487,
+ "step": 370
+ },
+ {
+ "epoch": 0.88,
+ "learning_rate": 0.0002,
+ "loss": 0.668,
+ "step": 380
+ },
+ {
+ "epoch": 0.9,
+ "learning_rate": 0.0002,
+ "loss": 0.6899,
+ "step": 390
+ },
+ {
+ "epoch": 0.93,
+ "learning_rate": 0.0002,
+ "loss": 0.684,
+ "step": 400
+ },
+ {
+ "epoch": 0.93,
+ "eval_loss": 0.7251922488212585,
+ "eval_runtime": 246.5381,
+ "eval_samples_per_second": 4.056,
+ "eval_steps_per_second": 2.028,
+ "step": 400
+ },
+ {
+ "epoch": 0.93,
+ "mmlu_eval_accuracy": 0.46184696708834644,
+ "mmlu_eval_accuracy_abstract_algebra": 0.18181818181818182,
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
+ "mmlu_eval_accuracy_astronomy": 0.375,
+ "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
+ "mmlu_eval_accuracy_clinical_knowledge": 0.4482758620689655,
+ "mmlu_eval_accuracy_college_biology": 0.4375,
+ "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": 0.2727272727272727,
+ "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+ "mmlu_eval_accuracy_computer_security": 0.2727272727272727,
+ "mmlu_eval_accuracy_conceptual_physics": 0.38461538461538464,
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+ "mmlu_eval_accuracy_electrical_engineering": 0.5,
+ "mmlu_eval_accuracy_elementary_mathematics": 0.36585365853658536,
+ "mmlu_eval_accuracy_formal_logic": 0.35714285714285715,
+ "mmlu_eval_accuracy_global_facts": 0.5,
+ "mmlu_eval_accuracy_high_school_biology": 0.34375,
+ "mmlu_eval_accuracy_high_school_chemistry": 0.4090909090909091,
+ "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
+ "mmlu_eval_accuracy_high_school_european_history": 0.7222222222222222,
+ "mmlu_eval_accuracy_high_school_geography": 0.8181818181818182,
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.6190476190476191,
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.32558139534883723,
+ "mmlu_eval_accuracy_high_school_mathematics": 0.13793103448275862,
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.4230769230769231,
+ "mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
+ "mmlu_eval_accuracy_high_school_psychology": 0.75,
+ "mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
+ "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
+ "mmlu_eval_accuracy_high_school_world_history": 0.5769230769230769,
+ "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
+ "mmlu_eval_accuracy_international_law": 0.7692307692307693,
+ "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
+ "mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
+ "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+ "mmlu_eval_accuracy_management": 0.5454545454545454,
+ "mmlu_eval_accuracy_marketing": 0.72,
+ "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+ "mmlu_eval_accuracy_miscellaneous": 0.6511627906976745,
+ "mmlu_eval_accuracy_moral_disputes": 0.5,
+ "mmlu_eval_accuracy_moral_scenarios": 0.23,
+ "mmlu_eval_accuracy_nutrition": 0.5454545454545454,
+ "mmlu_eval_accuracy_philosophy": 0.4411764705882353,
+ "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
+ "mmlu_eval_accuracy_professional_accounting": 0.3225806451612903,
+ "mmlu_eval_accuracy_professional_law": 0.34705882352941175,
+ "mmlu_eval_accuracy_professional_medicine": 0.45161290322580644,
+ "mmlu_eval_accuracy_professional_psychology": 0.4057971014492754,
+ "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": 0.5454545454545454,
+ "mmlu_eval_accuracy_virology": 0.3888888888888889,
+ "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+ "mmlu_loss": 1.1016935974982016,
+ "step": 400
+ },
+ {
+ "epoch": 0.95,
+ "learning_rate": 0.0002,
+ "loss": 0.7535,
+ "step": 410
+ },
+ {
+ "epoch": 0.97,
+ "learning_rate": 0.0002,
+ "loss": 0.707,
+ "step": 420
+ },
+ {
+ "epoch": 0.99,
+ "learning_rate": 0.0002,
+ "loss": 0.7077,
+ "step": 430
+ },
+ {
+ "epoch": 1.02,
+ "learning_rate": 0.0002,
+ "loss": 0.6389,
+ "step": 440
+ },
+ {
+ "epoch": 1.04,
+ "learning_rate": 0.0002,
+ "loss": 0.6701,
+ "step": 450
+ },
+ {
+ "epoch": 1.06,
+ "learning_rate": 0.0002,
+ "loss": 0.6462,
+ "step": 460
+ },
+ {
+ "epoch": 1.09,
+ "learning_rate": 0.0002,
+ "loss": 0.6421,
+ "step": 470
+ },
+ {
+ "epoch": 1.11,
+ "learning_rate": 0.0002,
+ "loss": 0.6822,
+ "step": 480
+ },
+ {
+ "epoch": 1.13,
+ "learning_rate": 0.0002,
+ "loss": 0.5916,
+ "step": 490
+ },
+ {
+ "epoch": 1.16,
+ "learning_rate": 0.0002,
+ "loss": 0.7141,
+ "step": 500
+ },
+ {
+ "epoch": 1.18,
+ "learning_rate": 0.0002,
+ "loss": 0.679,
+ "step": 510
+ },
+ {
+ "epoch": 1.2,
+ "learning_rate": 0.0002,
+ "loss": 0.5723,
+ "step": 520
+ },
+ {
+ "epoch": 1.23,
+ "learning_rate": 0.0002,
+ "loss": 0.6451,
+ "step": 530
+ },
+ {
+ "epoch": 1.25,
+ "learning_rate": 0.0002,
+ "loss": 0.6802,
+ "step": 540
+ },
+ {
+ "epoch": 1.27,
+ "learning_rate": 0.0002,
+ "loss": 0.5868,
+ "step": 550
+ },
+ {
+ "epoch": 1.3,
+ "learning_rate": 0.0002,
+ "loss": 0.6386,
+ "step": 560
+ },
+ {
+ "epoch": 1.32,
+ "learning_rate": 0.0002,
+ "loss": 0.5967,
+ "step": 570
+ },
+ {
+ "epoch": 1.34,
+ "learning_rate": 0.0002,
+ "loss": 0.618,
+ "step": 580
+ },
+ {
+ "epoch": 1.36,
+ "learning_rate": 0.0002,
+ "loss": 0.6294,
+ "step": 590
+ },
+ {
+ "epoch": 1.39,
+ "learning_rate": 0.0002,
+ "loss": 0.6417,
+ "step": 600
+ },
+ {
+ "epoch": 1.39,
+ "eval_loss": 0.7272388935089111,
+ "eval_runtime": 247.55,
+ "eval_samples_per_second": 4.04,
+ "eval_steps_per_second": 2.02,
+ "step": 600
+ },
+ {
+ "epoch": 1.39,
+ "mmlu_eval_accuracy": 0.4580576219014606,
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
+ "mmlu_eval_accuracy_astronomy": 0.375,
+ "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": 0.36363636363636365,
+ "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": 0.45454545454545453,
+ "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+ "mmlu_eval_accuracy_electrical_engineering": 0.4375,
+ "mmlu_eval_accuracy_elementary_mathematics": 0.2926829268292683,
+ "mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
+ "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": 0.7222222222222222,
+ "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.5238095238095238,
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.3488372093023256,
+ "mmlu_eval_accuracy_high_school_mathematics": 0.13793103448275862,
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
+ "mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
+ "mmlu_eval_accuracy_high_school_psychology": 0.7,
+ "mmlu_eval_accuracy_high_school_statistics": 0.2608695652173913,
+ "mmlu_eval_accuracy_high_school_us_history": 0.6818181818181818,
+ "mmlu_eval_accuracy_high_school_world_history": 0.5384615384615384,
+ "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
+ "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+ "mmlu_eval_accuracy_international_law": 0.7692307692307693,
+ "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+ "mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
+ "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+ "mmlu_eval_accuracy_management": 0.45454545454545453,
+ "mmlu_eval_accuracy_marketing": 0.72,
+ "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+ "mmlu_eval_accuracy_miscellaneous": 0.6511627906976745,
+ "mmlu_eval_accuracy_moral_disputes": 0.4473684210526316,
+ "mmlu_eval_accuracy_moral_scenarios": 0.23,
+ "mmlu_eval_accuracy_nutrition": 0.5757575757575758,
+ "mmlu_eval_accuracy_philosophy": 0.4117647058823529,
+ "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
+ "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
+ "mmlu_eval_accuracy_professional_law": 0.3411764705882353,
+ "mmlu_eval_accuracy_professional_medicine": 0.4838709677419355,
+ "mmlu_eval_accuracy_professional_psychology": 0.42028985507246375,
+ "mmlu_eval_accuracy_public_relations": 0.5,
+ "mmlu_eval_accuracy_security_studies": 0.5555555555555556,
+ "mmlu_eval_accuracy_sociology": 0.5454545454545454,
+ "mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
+ "mmlu_eval_accuracy_virology": 0.3888888888888889,
+ "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+ "mmlu_loss": 1.069228487308125,
+ "step": 600
+ },
+ {
+ "epoch": 1.41,
+ "learning_rate": 0.0002,
+ "loss": 0.6904,
+ "step": 610
+ },
+ {
+ "epoch": 1.43,
+ "learning_rate": 0.0002,
+ "loss": 0.6646,
+ "step": 620
+ },
+ {
+ "epoch": 1.46,
+ "learning_rate": 0.0002,
+ "loss": 0.6827,
+ "step": 630
+ },
+ {
+ "epoch": 1.48,
+ "learning_rate": 0.0002,
+ "loss": 0.6825,
+ "step": 640
+ },
+ {
+ "epoch": 1.5,
+ "learning_rate": 0.0002,
+ "loss": 0.6406,
+ "step": 650
+ },
+ {
+ "epoch": 1.53,
+ "learning_rate": 0.0002,
+ "loss": 0.6767,
+ "step": 660
+ },
+ {
+ "epoch": 1.55,
+ "learning_rate": 0.0002,
+ "loss": 0.6222,
+ "step": 670
+ },
+ {
+ "epoch": 1.57,
+ "learning_rate": 0.0002,
+ "loss": 0.5656,
+ "step": 680
+ },
+ {
+ "epoch": 1.6,
+ "learning_rate": 0.0002,
+ "loss": 0.6281,
+ "step": 690
+ },
+ {
+ "epoch": 1.62,
+ "learning_rate": 0.0002,
+ "loss": 0.6449,
+ "step": 700
+ },
+ {
+ "epoch": 1.64,
+ "learning_rate": 0.0002,
+ "loss": 0.6087,
+ "step": 710
+ },
+ {
+ "epoch": 1.67,
+ "learning_rate": 0.0002,
+ "loss": 0.6179,
+ "step": 720
+ },
+ {
+ "epoch": 1.69,
+ "learning_rate": 0.0002,
+ "loss": 0.596,
+ "step": 730
+ },
+ {
+ "epoch": 1.71,
+ "learning_rate": 0.0002,
+ "loss": 0.6577,
+ "step": 740
+ },
+ {
+ "epoch": 1.74,
+ "learning_rate": 0.0002,
+ "loss": 0.65,
+ "step": 750
+ },
+ {
+ "epoch": 1.76,
+ "learning_rate": 0.0002,
+ "loss": 0.5936,
+ "step": 760
+ },
+ {
+ "epoch": 1.78,
+ "learning_rate": 0.0002,
+ "loss": 0.65,
+ "step": 770
+ },
+ {
+ "epoch": 1.8,
+ "learning_rate": 0.0002,
+ "loss": 0.5787,
+ "step": 780
+ },
+ {
+ "epoch": 1.83,
+ "learning_rate": 0.0002,
+ "loss": 0.6082,
+ "step": 790
+ },
+ {
+ "epoch": 1.85,
+ "learning_rate": 0.0002,
+ "loss": 0.6129,
+ "step": 800
+ },
+ {
+ "epoch": 1.85,
+ "eval_loss": 0.7248261570930481,
+ "eval_runtime": 246.8878,
+ "eval_samples_per_second": 4.05,
+ "eval_steps_per_second": 2.025,
+ "step": 800
+ },
+ {
+ "epoch": 1.85,
+ "mmlu_eval_accuracy": 0.4610550473708825,
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
+ "mmlu_eval_accuracy_astronomy": 0.375,
+ "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": 0.36363636363636365,
+ "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": 0.18181818181818182,
+ "mmlu_eval_accuracy_conceptual_physics": 0.38461538461538464,
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+ "mmlu_eval_accuracy_electrical_engineering": 0.4375,
+ "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
+ "mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
+ "mmlu_eval_accuracy_global_facts": 0.5,
+ "mmlu_eval_accuracy_high_school_biology": 0.375,
+ "mmlu_eval_accuracy_high_school_chemistry": 0.4090909090909091,
+ "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
+ "mmlu_eval_accuracy_high_school_european_history": 0.7222222222222222,
+ "mmlu_eval_accuracy_high_school_geography": 0.7272727272727273,
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.32558139534883723,
+ "mmlu_eval_accuracy_high_school_mathematics": 0.1724137931034483,
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
+ "mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
+ "mmlu_eval_accuracy_high_school_psychology": 0.6833333333333333,
+ "mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
+ "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": 0.4166666666666667,
+ "mmlu_eval_accuracy_international_law": 0.7692307692307693,
+ "mmlu_eval_accuracy_jurisprudence": 0.36363636363636365,
+ "mmlu_eval_accuracy_logical_fallacies": 0.5555555555555556,
+ "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+ "mmlu_eval_accuracy_management": 0.45454545454545453,
+ "mmlu_eval_accuracy_marketing": 0.76,
+ "mmlu_eval_accuracy_medical_genetics": 0.7272727272727273,
+ "mmlu_eval_accuracy_miscellaneous": 0.6744186046511628,
+ "mmlu_eval_accuracy_moral_disputes": 0.42105263157894735,
+ "mmlu_eval_accuracy_moral_scenarios": 0.24,
+ "mmlu_eval_accuracy_nutrition": 0.5151515151515151,
+ "mmlu_eval_accuracy_philosophy": 0.4117647058823529,
+ "mmlu_eval_accuracy_prehistory": 0.42857142857142855,
+ "mmlu_eval_accuracy_professional_accounting": 0.3225806451612903,
+ "mmlu_eval_accuracy_professional_law": 0.3411764705882353,
+ "mmlu_eval_accuracy_professional_medicine": 0.4838709677419355,
+ "mmlu_eval_accuracy_professional_psychology": 0.42028985507246375,
+ "mmlu_eval_accuracy_public_relations": 0.5,
+ "mmlu_eval_accuracy_security_studies": 0.5555555555555556,
+ "mmlu_eval_accuracy_sociology": 0.5454545454545454,
+ "mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
+ "mmlu_eval_accuracy_virology": 0.3888888888888889,
+ "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+ "mmlu_loss": 1.0681800356538427,
+ "step": 800
+ },
+ {
+ "epoch": 1.87,
+ "learning_rate": 0.0002,
+ "loss": 0.6516,
+ "step": 810
+ },
+ {
+ "epoch": 1.9,
+ "learning_rate": 0.0002,
+ "loss": 0.6268,
+ "step": 820
+ },
+ {
+ "epoch": 1.92,
+ "learning_rate": 0.0002,
+ "loss": 0.6463,
+ "step": 830
+ },
+ {
+ "epoch": 1.94,
+ "learning_rate": 0.0002,
+ "loss": 0.6401,
+ "step": 840
+ },
+ {
+ "epoch": 1.97,
+ "learning_rate": 0.0002,
+ "loss": 0.6529,
+ "step": 850
+ },
+ {
+ "epoch": 1.99,
+ "learning_rate": 0.0002,
+ "loss": 0.6089,
+ "step": 860
+ },
+ {
+ "epoch": 2.01,
+ "learning_rate": 0.0002,
+ "loss": 0.5814,
+ "step": 870
+ },
+ {
+ "epoch": 2.04,
+ "learning_rate": 0.0002,
+ "loss": 0.4863,
+ "step": 880
+ },
+ {
+ "epoch": 2.06,
+ "learning_rate": 0.0002,
+ "loss": 0.5209,
+ "step": 890
+ },
+ {
+ "epoch": 2.08,
+ "learning_rate": 0.0002,
+ "loss": 0.4846,
+ "step": 900
+ },
+ {
+ "epoch": 2.11,
+ "learning_rate": 0.0002,
+ "loss": 0.5142,
+ "step": 910
+ },
+ {
+ "epoch": 2.13,
+ "learning_rate": 0.0002,
+ "loss": 0.5423,
+ "step": 920
+ },
+ {
+ "epoch": 2.15,
+ "learning_rate": 0.0002,
+ "loss": 0.5756,
+ "step": 930
+ },
+ {
+ "epoch": 2.17,
+ "learning_rate": 0.0002,
+ "loss": 0.5048,
+ "step": 940
+ },
+ {
+ "epoch": 2.2,
+ "learning_rate": 0.0002,
+ "loss": 0.5303,
+ "step": 950
+ },
+ {
+ "epoch": 2.22,
+ "learning_rate": 0.0002,
+ "loss": 0.5121,
+ "step": 960
+ },
+ {
+ "epoch": 2.24,
+ "learning_rate": 0.0002,
+ "loss": 0.574,
+ "step": 970
+ },
+ {
+ "epoch": 2.27,
+ "learning_rate": 0.0002,
+ "loss": 0.5207,
+ "step": 980
+ },
+ {
+ "epoch": 2.29,
+ "learning_rate": 0.0002,
+ "loss": 0.5412,
+ "step": 990
+ },
+ {
+ "epoch": 2.31,
+ "learning_rate": 0.0002,
+ "loss": 0.5303,
+ "step": 1000
+ },
+ {
+ "epoch": 2.31,
+ "eval_loss": 0.7593775391578674,
+ "eval_runtime": 246.7377,
+ "eval_samples_per_second": 4.053,
+ "eval_steps_per_second": 2.026,
+ "step": 1000
+ },
+ {
+ "epoch": 2.31,
+ "mmlu_eval_accuracy": 0.4569260811160823,
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+ "mmlu_eval_accuracy_anatomy": 0.5,
+ "mmlu_eval_accuracy_astronomy": 0.375,
+ "mmlu_eval_accuracy_business_ethics": 0.5454545454545454,
+ "mmlu_eval_accuracy_clinical_knowledge": 0.3793103448275862,
+ "mmlu_eval_accuracy_college_biology": 0.5,
+ "mmlu_eval_accuracy_college_chemistry": 0.25,
+ "mmlu_eval_accuracy_college_computer_science": 0.5454545454545454,
+ "mmlu_eval_accuracy_college_mathematics": 0.18181818181818182,
+ "mmlu_eval_accuracy_college_medicine": 0.3181818181818182,
+ "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+ "mmlu_eval_accuracy_computer_security": 0.2727272727272727,
+ "mmlu_eval_accuracy_conceptual_physics": 0.38461538461538464,
+ "mmlu_eval_accuracy_econometrics": 0.16666666666666666,
+ "mmlu_eval_accuracy_electrical_engineering": 0.3125,
+ "mmlu_eval_accuracy_elementary_mathematics": 0.34146341463414637,
+ "mmlu_eval_accuracy_formal_logic": 0.2857142857142857,
+ "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": 0.6666666666666666,
+ "mmlu_eval_accuracy_high_school_geography": 0.7272727272727273,
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.37209302325581395,
+ "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.46153846153846156,
+ "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
+ "mmlu_eval_accuracy_high_school_psychology": 0.7,
+ "mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
+ "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
+ "mmlu_eval_accuracy_high_school_world_history": 0.5384615384615384,
+ "mmlu_eval_accuracy_human_aging": 0.6956521739130435,
+ "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+ "mmlu_eval_accuracy_international_law": 0.7692307692307693,
+ "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+ "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+ "mmlu_eval_accuracy_management": 0.5454545454545454,
+ "mmlu_eval_accuracy_marketing": 0.76,
+ "mmlu_eval_accuracy_medical_genetics": 0.6363636363636364,
+ "mmlu_eval_accuracy_miscellaneous": 0.6744186046511628,
+ "mmlu_eval_accuracy_moral_disputes": 0.4473684210526316,
+ "mmlu_eval_accuracy_moral_scenarios": 0.24,
+ "mmlu_eval_accuracy_nutrition": 0.5454545454545454,
+ "mmlu_eval_accuracy_philosophy": 0.4411764705882353,
+ "mmlu_eval_accuracy_prehistory": 0.45714285714285713,
+ "mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
+ "mmlu_eval_accuracy_professional_law": 0.3588235294117647,
+ "mmlu_eval_accuracy_professional_medicine": 0.4838709677419355,
+ "mmlu_eval_accuracy_professional_psychology": 0.43478260869565216,
+ "mmlu_eval_accuracy_public_relations": 0.5,
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
+ "mmlu_eval_accuracy_sociology": 0.5909090909090909,
+ "mmlu_eval_accuracy_us_foreign_policy": 0.5454545454545454,
+ "mmlu_eval_accuracy_virology": 0.4444444444444444,
+ "mmlu_eval_accuracy_world_religions": 0.7368421052631579,
+ "mmlu_loss": 1.3353144975267253,
+ "step": 1000
+ },
+ {
+ "epoch": 2.34,
+ "learning_rate": 0.0002,
+ "loss": 0.5821,
+ "step": 1010
+ },
+ {
+ "epoch": 2.36,
+ "learning_rate": 0.0002,
+ "loss": 0.5049,
+ "step": 1020
+ },
+ {
+ "epoch": 2.38,
+ "learning_rate": 0.0002,
+ "loss": 0.5422,
+ "step": 1030
+ },
+ {
+ "epoch": 2.41,
+ "learning_rate": 0.0002,
+ "loss": 0.5449,
+ "step": 1040
+ },
+ {
+ "epoch": 2.43,
+ "learning_rate": 0.0002,
+ "loss": 0.5118,
+ "step": 1050
+ },
+ {
+ "epoch": 2.45,
+ "learning_rate": 0.0002,
+ "loss": 0.5837,
+ "step": 1060
+ },
+ {
+ "epoch": 2.48,
+ "learning_rate": 0.0002,
+ "loss": 0.5563,
+ "step": 1070
+ },
+ {
+ "epoch": 2.5,
+ "learning_rate": 0.0002,
+ "loss": 0.5417,
+ "step": 1080
+ },
+ {
+ "epoch": 2.52,
+ "learning_rate": 0.0002,
+ "loss": 0.5476,
+ "step": 1090
+ },
+ {
+ "epoch": 2.54,
+ "learning_rate": 0.0002,
+ "loss": 0.5308,
+ "step": 1100
+ },
+ {
+ "epoch": 2.57,
+ "learning_rate": 0.0002,
+ "loss": 0.5445,
+ "step": 1110
+ },
+ {
+ "epoch": 2.59,
+ "learning_rate": 0.0002,
+ "loss": 0.4944,
+ "step": 1120
+ },
+ {
+ "epoch": 2.61,
+ "learning_rate": 0.0002,
+ "loss": 0.5082,
+ "step": 1130
+ },
+ {
+ "epoch": 2.64,
+ "learning_rate": 0.0002,
+ "loss": 0.5107,
+ "step": 1140
+ },
+ {
+ "epoch": 2.66,
+ "learning_rate": 0.0002,
+ "loss": 0.4959,
+ "step": 1150
+ },
+ {
+ "epoch": 2.68,
+ "learning_rate": 0.0002,
+ "loss": 0.5311,
+ "step": 1160
+ },
+ {
+ "epoch": 2.71,
+ "learning_rate": 0.0002,
+ "loss": 0.5388,
+ "step": 1170
+ },
+ {
+ "epoch": 2.73,
+ "learning_rate": 0.0002,
+ "loss": 0.4745,
+ "step": 1180
+ },
+ {
+ "epoch": 2.75,
+ "learning_rate": 0.0002,
+ "loss": 0.5232,
+ "step": 1190
+ },
+ {
+ "epoch": 2.78,
+ "learning_rate": 0.0002,
+ "loss": 0.5388,
+ "step": 1200
+ },
+ {
+ "epoch": 2.78,
+ "eval_loss": 0.7669270634651184,
+ "eval_runtime": 246.0098,
+ "eval_samples_per_second": 4.065,
+ "eval_steps_per_second": 2.032,
+ "step": 1200
+ },
+ {
+ "epoch": 2.78,
+ "mmlu_eval_accuracy": 0.4734820857068989,
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+ "mmlu_eval_accuracy_anatomy": 0.5714285714285714,
+ "mmlu_eval_accuracy_astronomy": 0.4375,
+ "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": 0.36363636363636365,
+ "mmlu_eval_accuracy_college_mathematics": 0.2727272727272727,
+ "mmlu_eval_accuracy_college_medicine": 0.36363636363636365,
+ "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+ "mmlu_eval_accuracy_computer_security": 0.6363636363636364,
+ "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
+ "mmlu_eval_accuracy_econometrics": 0.25,
+ "mmlu_eval_accuracy_electrical_engineering": 0.375,
+ "mmlu_eval_accuracy_elementary_mathematics": 0.3170731707317073,
+ "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": 0.4090909090909091,
+ "mmlu_eval_accuracy_high_school_computer_science": 0.6666666666666666,
+ "mmlu_eval_accuracy_high_school_european_history": 0.6111111111111112,
+ "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.5714285714285714,
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.3488372093023256,
+ "mmlu_eval_accuracy_high_school_mathematics": 0.20689655172413793,
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
+ "mmlu_eval_accuracy_high_school_physics": 0.35294117647058826,
+ "mmlu_eval_accuracy_high_school_psychology": 0.7333333333333333,
+ "mmlu_eval_accuracy_high_school_statistics": 0.30434782608695654,
+ "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
+ "mmlu_eval_accuracy_high_school_world_history": 0.5769230769230769,
+ "mmlu_eval_accuracy_human_aging": 0.7391304347826086,
+ "mmlu_eval_accuracy_human_sexuality": 0.3333333333333333,
+ "mmlu_eval_accuracy_international_law": 0.7692307692307693,
+ "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+ "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+ "mmlu_eval_accuracy_management": 0.5454545454545454,
+ "mmlu_eval_accuracy_marketing": 0.8,
+ "mmlu_eval_accuracy_medical_genetics": 0.6363636363636364,
+ "mmlu_eval_accuracy_miscellaneous": 0.6395348837209303,
+ "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
+ "mmlu_eval_accuracy_moral_scenarios": 0.24,
+ "mmlu_eval_accuracy_nutrition": 0.5454545454545454,
+ "mmlu_eval_accuracy_philosophy": 0.4411764705882353,
+ "mmlu_eval_accuracy_prehistory": 0.5142857142857142,
+ "mmlu_eval_accuracy_professional_accounting": 0.3548387096774194,
+ "mmlu_eval_accuracy_professional_law": 0.3176470588235294,
+ "mmlu_eval_accuracy_professional_medicine": 0.4838709677419355,
+ "mmlu_eval_accuracy_professional_psychology": 0.4057971014492754,
+ "mmlu_eval_accuracy_public_relations": 0.4166666666666667,
+ "mmlu_eval_accuracy_security_studies": 0.5185185185185185,
+ "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": 0.7368421052631579,
+ "mmlu_loss": 1.082002373310122,
+ "step": 1200
+ },
+ {
+ "epoch": 2.8,
+ "learning_rate": 0.0002,
+ "loss": 0.5243,
+ "step": 1210
+ },
+ {
+ "epoch": 2.82,
+ "learning_rate": 0.0002,
+ "loss": 0.5168,
+ "step": 1220
+ },
+ {
+ "epoch": 2.85,
+ "learning_rate": 0.0002,
+ "loss": 0.4966,
+ "step": 1230
+ },
+ {
+ "epoch": 2.87,
+ "learning_rate": 0.0002,
+ "loss": 0.5869,
+ "step": 1240
+ },
+ {
+ "epoch": 2.89,
+ "learning_rate": 0.0002,
+ "loss": 0.5336,
+ "step": 1250
+ },
+ {
+ "epoch": 2.91,
+ "learning_rate": 0.0002,
+ "loss": 0.563,
+ "step": 1260
+ },
+ {
+ "epoch": 2.94,
+ "learning_rate": 0.0002,
+ "loss": 0.5002,
+ "step": 1270
+ },
+ {
+ "epoch": 2.96,
+ "learning_rate": 0.0002,
+ "loss": 0.607,
+ "step": 1280
+ },
+ {
+ "epoch": 2.98,
+ "learning_rate": 0.0002,
+ "loss": 0.5458,
+ "step": 1290
+ },
+ {
+ "epoch": 3.01,
+ "learning_rate": 0.0002,
+ "loss": 0.5263,
+ "step": 1300
+ },
+ {
+ "epoch": 3.03,
+ "learning_rate": 0.0002,
+ "loss": 0.4074,
+ "step": 1310
+ },
+ {
+ "epoch": 3.05,
+ "learning_rate": 0.0002,
+ "loss": 0.3737,
+ "step": 1320
+ },
+ {
+ "epoch": 3.08,
+ "learning_rate": 0.0002,
+ "loss": 0.3938,
+ "step": 1330
+ },
+ {
+ "epoch": 3.1,
+ "learning_rate": 0.0002,
+ "loss": 0.3802,
+ "step": 1340
+ },
+ {
+ "epoch": 3.12,
+ "learning_rate": 0.0002,
+ "loss": 0.4216,
+ "step": 1350
+ },
+ {
+ "epoch": 3.15,
+ "learning_rate": 0.0002,
+ "loss": 0.4045,
+ "step": 1360
+ },
+ {
+ "epoch": 3.17,
+ "learning_rate": 0.0002,
+ "loss": 0.4024,
+ "step": 1370
+ },
+ {
+ "epoch": 3.19,
+ "learning_rate": 0.0002,
+ "loss": 0.3943,
+ "step": 1380
+ },
+ {
+ "epoch": 3.22,
+ "learning_rate": 0.0002,
+ "loss": 0.3923,
+ "step": 1390
+ },
+ {
+ "epoch": 3.24,
+ "learning_rate": 0.0002,
+ "loss": 0.4267,
+ "step": 1400
+ },
+ {
+ "epoch": 3.24,
+ "eval_loss": 0.8232849836349487,
+ "eval_runtime": 245.4693,
+ "eval_samples_per_second": 4.074,
+ "eval_steps_per_second": 2.037,
+ "step": 1400
+ },
+ {
+ "epoch": 3.24,
+ "mmlu_eval_accuracy": 0.4552997822968728,
+ "mmlu_eval_accuracy_abstract_algebra": 0.36363636363636365,
+ "mmlu_eval_accuracy_anatomy": 0.5,
+ "mmlu_eval_accuracy_astronomy": 0.3125,
+ "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": 0.25,
+ "mmlu_eval_accuracy_college_computer_science": 0.36363636363636365,
+ "mmlu_eval_accuracy_college_mathematics": 0.36363636363636365,
+ "mmlu_eval_accuracy_college_medicine": 0.4090909090909091,
+ "mmlu_eval_accuracy_college_physics": 0.45454545454545453,
+ "mmlu_eval_accuracy_computer_security": 0.36363636363636365,
+ "mmlu_eval_accuracy_conceptual_physics": 0.4230769230769231,
+ "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": 0.21428571428571427,
+ "mmlu_eval_accuracy_global_facts": 0.4,
+ "mmlu_eval_accuracy_high_school_biology": 0.46875,
+ "mmlu_eval_accuracy_high_school_chemistry": 0.3181818181818182,
+ "mmlu_eval_accuracy_high_school_computer_science": 0.4444444444444444,
+ "mmlu_eval_accuracy_high_school_european_history": 0.5555555555555556,
+ "mmlu_eval_accuracy_high_school_geography": 0.7727272727272727,
+ "mmlu_eval_accuracy_high_school_government_and_politics": 0.5238095238095238,
+ "mmlu_eval_accuracy_high_school_macroeconomics": 0.3488372093023256,
+ "mmlu_eval_accuracy_high_school_mathematics": 0.2413793103448276,
+ "mmlu_eval_accuracy_high_school_microeconomics": 0.5,
+ "mmlu_eval_accuracy_high_school_physics": 0.29411764705882354,
+ "mmlu_eval_accuracy_high_school_psychology": 0.7166666666666667,
+ "mmlu_eval_accuracy_high_school_statistics": 0.2608695652173913,
+ "mmlu_eval_accuracy_high_school_us_history": 0.6363636363636364,
+ "mmlu_eval_accuracy_high_school_world_history": 0.6153846153846154,
+ "mmlu_eval_accuracy_human_aging": 0.6521739130434783,
+ "mmlu_eval_accuracy_human_sexuality": 0.4166666666666667,
+ "mmlu_eval_accuracy_international_law": 0.7692307692307693,
+ "mmlu_eval_accuracy_jurisprudence": 0.2727272727272727,
+ "mmlu_eval_accuracy_logical_fallacies": 0.6111111111111112,
+ "mmlu_eval_accuracy_machine_learning": 0.18181818181818182,
+ "mmlu_eval_accuracy_management": 0.6363636363636364,
+ "mmlu_eval_accuracy_marketing": 0.76,
+ "mmlu_eval_accuracy_medical_genetics": 0.6363636363636364,
+ "mmlu_eval_accuracy_miscellaneous": 0.6627906976744186,
+ "mmlu_eval_accuracy_moral_disputes": 0.47368421052631576,
+ "mmlu_eval_accuracy_moral_scenarios": 0.21,
+ "mmlu_eval_accuracy_nutrition": 0.5454545454545454,
+ "mmlu_eval_accuracy_philosophy": 0.47058823529411764,
+ "mmlu_eval_accuracy_prehistory": 0.4857142857142857,
+ "mmlu_eval_accuracy_professional_accounting": 0.2903225806451613,
+ "mmlu_eval_accuracy_professional_law": 0.29411764705882354,
+ "mmlu_eval_accuracy_professional_medicine": 0.41935483870967744,
+ "mmlu_eval_accuracy_professional_psychology": 0.34782608695652173,
+ "mmlu_eval_accuracy_public_relations": 0.4166666666666667,
+ "mmlu_eval_accuracy_security_studies": 0.5555555555555556,
+ "mmlu_eval_accuracy_sociology": 0.5454545454545454,
+ "mmlu_eval_accuracy_us_foreign_policy": 0.45454545454545453,
+ "mmlu_eval_accuracy_virology": 0.5555555555555556,
+ "mmlu_eval_accuracy_world_religions": 0.7894736842105263,
+ "mmlu_loss": 1.2433716685416338,
+ "step": 1400
+ }
+ ],
+ "max_steps": 5000,
+ "num_train_epochs": 12,
+ "total_flos": 3.994598287814492e+17,
+ "trial_name": null,
+ "trial_params": null
+}

checkpoint-1400/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87c64fb3887e7b9f2e5dae3db2d5bf0c30becc3e3774f7c348e4bd2c620de810
+size 6011

checkpoint-800/adapter_model/adapter_model/README.md

@@ -15,6 +15,17 @@ The following `bitsandbytes` quantization config was used during training:
 - bnb_4bit_use_double_quant: True
 - bnb_4bit_compute_dtype: bfloat16
 
+The following `bitsandbytes` quantization config was used during training:
+- load_in_8bit: False
+- load_in_4bit: True
+- llm_int8_threshold: 6.0
+- llm_int8_skip_modules: None
+- llm_int8_enable_fp32_cpu_offload: False
+- llm_int8_has_fp16_weight: False
+- bnb_4bit_quant_type: nf4
+- bnb_4bit_use_double_quant: True
+- bnb_4bit_compute_dtype: bfloat16
+
 The following `bitsandbytes` quantization config was used during training:
 - load_in_8bit: False
 - load_in_4bit: True
@@ -27,6 +38,7 @@ 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-800/adapter_model/adapter_model/adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:239a990af6deccad780a9c0e8d9b390e6ff81c4c34b3ffcaee21b2683d4a698b
+oid sha256:e344083300bd30eb9777f0a0328191d02ed3d81426dd12bf6e9ae3e704352b5c
 size 319977229