Add files using upload-large-folder tool

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_c-r2-a2/adapter_config.json

@@ -0,0 +1,30 @@
+{
+  "alpha": 4,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "init_weight": "kaiming",
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "metric_tracking": false,
+  "modules_to_save": null,
+  "peft_type": "ABLATION",
+  "r": 2,
+  "revision": null,
+  "seed": 42,
+  "share_weights": false,
+  "target_modules": [
+    "up_proj",
+    "gate_proj",
+    "v_proj",
+    "k_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj"
+  ],
+  "task_type": null,
+  "track_n": 100,
+  "variant": "B"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_c-r2-a2/eval_results.json

@@ -0,0 +1,4 @@
+{
+    "task": "arc_c",
+    "results": 0.38139931740614336
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_c-r2-a2/training_configuration.json

@@ -0,0 +1,38 @@
+{
+  "model_id": "TinyLlama/TinyLlama_v1.1",
+  "dataset": {
+    "name": "ARC_C",
+    "dataset_id": "allenai/ai2_arc",
+    "preprocess_id": "arc_train_deepeval"
+  },
+  "peft_config": {
+    "method": "abl_B",
+    "rank": 2,
+    "alpha": 4,
+    "dropout": 0.0,
+    "bias": "none",
+    "target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "down_proj",
+      "up_proj"
+    ],
+    "trainable_parameter_count": 1182720
+  },
+  "training_config": {
+    "max_dataset_length": null,
+    "batch_size": 64,
+    "per_device_batch_size": 32,
+    "gradient_accumulation_steps": 2,
+    "learning_rate": 0.0003,
+    "num_epochs": 4,
+    "warmup_ratio": 0.1
+  },
+  "model_name": "TinyLlama_v1.1-abl_B-arc_c-r2-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_c-r2-a2",
+  "seed": 42,
+  "timestamp": "2025-08-31T13:31:46.211850"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_c-r2-a2/training_logs.json

@@ -0,0 +1,625 @@
+[
+  {
+    "step": 1,
+    "epoch": 0.05714285714285714,
+    "cpu_mem": 1.506254848,
+    "gpu_mem": 4.425348608,
+    "loss": 4.4614,
+    "grad_norm": 256.6861267089844,
+    "learning_rate": 4.285714285714285e-05
+  },
+  {
+    "step": 2,
+    "epoch": 0.11428571428571428,
+    "cpu_mem": 1.506451456,
+    "gpu_mem": 4.43480064,
+    "loss": 4.6994,
+    "grad_norm": 261.9680480957031,
+    "learning_rate": 8.57142857142857e-05
+  },
+  {
+    "step": 3,
+    "epoch": 0.17142857142857143,
+    "cpu_mem": 1.506451456,
+    "gpu_mem": 4.43483136,
+    "loss": 2.2137,
+    "grad_norm": 135.3577423095703,
+    "learning_rate": 0.00012857142857142855
+  },
+  {
+    "step": 4,
+    "epoch": 0.22857142857142856,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434797568,
+    "loss": 1.5723,
+    "grad_norm": 16.541139602661133,
+    "learning_rate": 0.0001714285714285714
+  },
+  {
+    "step": 5,
+    "epoch": 0.2857142857142857,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.43478528,
+    "loss": 1.4149,
+    "grad_norm": 9.431467056274414,
+    "learning_rate": 0.00021428571428571427
+  },
+  {
+    "step": 6,
+    "epoch": 0.34285714285714286,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434848256,
+    "loss": 1.4493,
+    "grad_norm": 18.17936134338379,
+    "learning_rate": 0.0002571428571428571
+  },
+  {
+    "step": 7,
+    "epoch": 0.4,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.4348544,
+    "loss": 1.5623,
+    "grad_norm": 17.265527725219727,
+    "learning_rate": 0.0003
+  },
+  {
+    "step": 8,
+    "epoch": 0.45714285714285713,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434812928,
+    "loss": 1.3595,
+    "grad_norm": 3.7705695629119873,
+    "learning_rate": 0.00029980111348272456
+  },
+  {
+    "step": 9,
+    "epoch": 0.5142857142857142,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.43480832,
+    "loss": 1.3274,
+    "grad_norm": 7.432311058044434,
+    "learning_rate": 0.00029920498134218835
+  },
+  {
+    "step": 10,
+    "epoch": 0.5714285714285714,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434797568,
+    "loss": 1.6472,
+    "grad_norm": 23.43598747253418,
+    "learning_rate": 0.0002982131844136615
+  },
+  {
+    "step": 11,
+    "epoch": 0.6285714285714286,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.43480832,
+    "loss": 1.5246,
+    "grad_norm": 11.669015884399414,
+    "learning_rate": 0.0002968283527643036
+  },
+  {
+    "step": 12,
+    "epoch": 0.6857142857142857,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434832896,
+    "loss": 1.4175,
+    "grad_norm": 7.646242618560791,
+    "learning_rate": 0.000295054158718698
+  },
+  {
+    "step": 13,
+    "epoch": 0.7428571428571429,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434832896,
+    "loss": 1.3104,
+    "grad_norm": 16.06755828857422,
+    "learning_rate": 0.00029289530712050735
+  },
+  {
+    "step": 14,
+    "epoch": 0.8,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434780672,
+    "loss": 1.5926,
+    "grad_norm": 24.16870880126953,
+    "learning_rate": 0.000290357522856074
+  },
+  {
+    "step": 15,
+    "epoch": 0.8571428571428571,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434855936,
+    "loss": 1.5066,
+    "grad_norm": 12.364516258239746,
+    "learning_rate": 0.0002874475356730507
+  },
+  {
+    "step": 16,
+    "epoch": 0.9142857142857143,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434849792,
+    "loss": 1.4177,
+    "grad_norm": 11.54814624786377,
+    "learning_rate": 0.0002841730623343193
+  },
+  {
+    "step": 17,
+    "epoch": 0.9714285714285714,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.4348544,
+    "loss": 1.3599,
+    "grad_norm": 5.233391284942627,
+    "learning_rate": 0.00028054278615452326
+  },
+  {
+    "step": 18,
+    "epoch": 1.0285714285714285,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439542272,
+    "loss": 2.0643,
+    "grad_norm": 10.010595321655273,
+    "learning_rate": 0.0002765663339734778
+  },
+  {
+    "step": 19,
+    "epoch": 1.0857142857142856,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439540736,
+    "loss": 1.4071,
+    "grad_norm": 7.176828861236572,
+    "learning_rate": 0.00027225425062752165
+  },
+  {
+    "step": 20,
+    "epoch": 1.1428571428571428,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.43951616,
+    "loss": 1.3075,
+    "grad_norm": 5.2276411056518555,
+    "learning_rate": 0.0002676179709865066
+  },
+  {
+    "step": 21,
+    "epoch": 1.2,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.43952384,
+    "loss": 1.4644,
+    "grad_norm": 11.008788108825684,
+    "learning_rate": 0.0002626697896305779
+  },
+  {
+    "step": 22,
+    "epoch": 1.2571428571428571,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439553024,
+    "loss": 1.4464,
+    "grad_norm": 10.353914260864258,
+    "learning_rate": 0.000257422828247159
+  },
+  {
+    "step": 23,
+    "epoch": 1.3142857142857143,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439582208,
+    "loss": 1.3721,
+    "grad_norm": 5.877532482147217,
+    "learning_rate": 0.00025189100083459397
+  },
+  {
+    "step": 24,
+    "epoch": 1.3714285714285714,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439525376,
+    "loss": 1.3644,
+    "grad_norm": 4.342014312744141,
+    "learning_rate": 0.0002460889768047263
+  },
+  {
+    "step": 25,
+    "epoch": 1.4285714285714286,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439594496,
+    "loss": 1.3209,
+    "grad_norm": 3.078639507293701,
+    "learning_rate": 0.00024003214208225522
+  },
+  {
+    "step": 26,
+    "epoch": 1.4857142857142858,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439551488,
+    "loss": 1.3689,
+    "grad_norm": 3.996753454208374,
+    "learning_rate": 0.00023373655830402968
+  },
+  {
+    "step": 27,
+    "epoch": 1.5428571428571427,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439510016,
+    "loss": 1.415,
+    "grad_norm": 4.786011219024658,
+    "learning_rate": 0.00022721892022647462
+  },
+  {
+    "step": 28,
+    "epoch": 1.6,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439556096,
+    "loss": 1.5722,
+    "grad_norm": 8.570415496826172,
+    "learning_rate": 0.000220496511454098
+  },
+  {
+    "step": 29,
+    "epoch": 1.657142857142857,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439551488,
+    "loss": 1.3872,
+    "grad_norm": 2.8165605068206787,
+    "learning_rate": 0.0002135871586064791
+  },
+  {
+    "step": 30,
+    "epoch": 1.7142857142857144,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439540736,
+    "loss": 1.3614,
+    "grad_norm": 3.3478944301605225,
+    "learning_rate": 0.00020650918404527775
+  },
+  {
+    "step": 31,
+    "epoch": 1.7714285714285714,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439571456,
+    "loss": 1.3404,
+    "grad_norm": 2.628307819366455,
+    "learning_rate": 0.00019928135728662522
+  },
+  {
+    "step": 32,
+    "epoch": 1.8285714285714287,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439580672,
+    "loss": 1.3803,
+    "grad_norm": 4.98179292678833,
+    "learning_rate": 0.00019192284522774142
+  },
+  {
+    "step": 33,
+    "epoch": 1.8857142857142857,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439560704,
+    "loss": 1.4267,
+    "grad_norm": 6.194660663604736,
+    "learning_rate": 0.00018445316131976934
+  },
+  {
+    "step": 34,
+    "epoch": 1.9428571428571428,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.4395392,
+    "loss": 1.4209,
+    "grad_norm": 5.258482456207275,
+    "learning_rate": 0.00017689211382161034
+  },
+  {
+    "step": 35,
+    "epoch": 2.0,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439427072,
+    "loss": 2.1003,
+    "grad_norm": 6.661643981933594,
+    "learning_rate": 0.00016925975327198266
+  },
+  {
+    "step": 36,
+    "epoch": 2.057142857142857,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434826752,
+    "loss": 1.399,
+    "grad_norm": 5.453505039215088,
+    "learning_rate": 0.00016157631931899697
+  },
+  {
+    "step": 37,
+    "epoch": 2.1142857142857143,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434835968,
+    "loss": 1.3992,
+    "grad_norm": 4.897593975067139,
+    "learning_rate": 0.0001538621870482483
+  },
+  {
+    "step": 38,
+    "epoch": 2.1714285714285713,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434806784,
+    "loss": 1.3695,
+    "grad_norm": 4.509614944458008,
+    "learning_rate": 0.00014613781295175172
+  },
+  {
+    "step": 39,
+    "epoch": 2.2285714285714286,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434825216,
+    "loss": 1.3894,
+    "grad_norm": 4.451006889343262,
+    "learning_rate": 0.00013842368068100303
+  },
+  {
+    "step": 40,
+    "epoch": 2.2857142857142856,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434802176,
+    "loss": 1.3774,
+    "grad_norm": 2.4717953205108643,
+    "learning_rate": 0.00013074024672801731
+  },
+  {
+    "step": 41,
+    "epoch": 2.342857142857143,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434803712,
+    "loss": 1.3856,
+    "grad_norm": 2.598550796508789,
+    "learning_rate": 0.00012310788617838966
+  },
+  {
+    "step": 42,
+    "epoch": 2.4,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434832896,
+    "loss": 1.3837,
+    "grad_norm": 5.809165954589844,
+    "learning_rate": 0.00011554683868023067
+  },
+  {
+    "step": 43,
+    "epoch": 2.4571428571428573,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434848256,
+    "loss": 1.3518,
+    "grad_norm": 4.0668511390686035,
+    "learning_rate": 0.00010807715477225858
+  },
+  {
+    "step": 44,
+    "epoch": 2.5142857142857142,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434866688,
+    "loss": 1.3572,
+    "grad_norm": 3.543848991394043,
+    "learning_rate": 0.00010071864271337478
+  },
+  {
+    "step": 45,
+    "epoch": 2.571428571428571,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434820608,
+    "loss": 1.3099,
+    "grad_norm": 2.2563862800598145,
+    "learning_rate": 9.34908159547222e-05
+  },
+  {
+    "step": 46,
+    "epoch": 2.6285714285714286,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434814464,
+    "loss": 1.2907,
+    "grad_norm": 3.035221815109253,
+    "learning_rate": 8.641284139352091e-05
+  },
+  {
+    "step": 47,
+    "epoch": 2.685714285714286,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.43480832,
+    "loss": 1.2951,
+    "grad_norm": 2.453887462615967,
+    "learning_rate": 7.950348854590204e-05
+  },
+  {
+    "step": 48,
+    "epoch": 2.742857142857143,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434812928,
+    "loss": 1.2507,
+    "grad_norm": 2.245753526687622,
+    "learning_rate": 7.278107977352543e-05
+  },
+  {
+    "step": 49,
+    "epoch": 2.8,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434803712,
+    "loss": 1.3112,
+    "grad_norm": 2.7252495288848877,
+    "learning_rate": 6.626344169597031e-05
+  },
+  {
+    "step": 50,
+    "epoch": 2.857142857142857,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.43478528,
+    "loss": 1.3279,
+    "grad_norm": 2.834775447845459,
+    "learning_rate": 5.996785791774478e-05
+  },
+  {
+    "step": 51,
+    "epoch": 2.914285714285714,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434809856,
+    "loss": 1.3215,
+    "grad_norm": 3.576658010482788,
+    "learning_rate": 5.391102319527373e-05
+  },
+  {
+    "step": 52,
+    "epoch": 2.9714285714285715,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.434837504,
+    "loss": 1.351,
+    "grad_norm": 2.464062452316284,
+    "learning_rate": 4.8108999165406026e-05
+  },
+  {
+    "step": 53,
+    "epoch": 3.0285714285714285,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439537664,
+    "loss": 1.9408,
+    "grad_norm": 4.601680755615234,
+    "learning_rate": 4.257717175284103e-05
+  },
+  {
+    "step": 54,
+    "epoch": 3.085714285714286,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439506944,
+    "loss": 1.3573,
+    "grad_norm": 3.305359363555908,
+    "learning_rate": 3.733021036942205e-05
+  },
+  {
+    "step": 55,
+    "epoch": 3.142857142857143,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439540736,
+    "loss": 1.3641,
+    "grad_norm": 4.420423984527588,
+    "learning_rate": 3.238202901349345e-05
+  },
+  {
+    "step": 56,
+    "epoch": 3.2,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439614464,
+    "loss": 1.3082,
+    "grad_norm": 2.968301296234131,
+    "learning_rate": 2.774574937247831e-05
+  },
+  {
+    "step": 57,
+    "epoch": 3.257142857142857,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439557632,
+    "loss": 1.287,
+    "grad_norm": 2.3714208602905273,
+    "learning_rate": 2.3433666026522153e-05
+  },
+  {
+    "step": 58,
+    "epoch": 3.314285714285714,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439551488,
+    "loss": 1.2157,
+    "grad_norm": 3.4141745567321777,
+    "learning_rate": 1.945721384547671e-05
+  },
+  {
+    "step": 59,
+    "epoch": 3.3714285714285714,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439602176,
+    "loss": 1.2699,
+    "grad_norm": 3.1746935844421387,
+    "learning_rate": 1.5826937665680693e-05
+  },
+  {
+    "step": 60,
+    "epoch": 3.4285714285714284,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439528448,
+    "loss": 1.3339,
+    "grad_norm": 4.8189496994018555,
+    "learning_rate": 1.2552464326949302e-05
+  },
+  {
+    "step": 61,
+    "epoch": 3.4857142857142858,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439542272,
+    "loss": 1.341,
+    "grad_norm": 4.947413921356201,
+    "learning_rate": 9.64247714392597e-06
+  },
+  {
+    "step": 62,
+    "epoch": 3.5428571428571427,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439543808,
+    "loss": 1.2974,
+    "grad_norm": 3.260141611099243,
+    "learning_rate": 7.104692879492624e-06
+  },
+  {
+    "step": 63,
+    "epoch": 3.6,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439533056,
+    "loss": 1.3065,
+    "grad_norm": 3.7296063899993896,
+    "learning_rate": 4.945841281301943e-06
+  },
+  {
+    "step": 64,
+    "epoch": 3.657142857142857,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439549952,
+    "loss": 1.2641,
+    "grad_norm": 4.162058353424072,
+    "learning_rate": 3.1716472356963286e-06
+  },
+  {
+    "step": 65,
+    "epoch": 3.7142857142857144,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439571456,
+    "loss": 1.2935,
+    "grad_norm": 3.655007839202881,
+    "learning_rate": 1.7868155863384415e-06
+  },
+  {
+    "step": 66,
+    "epoch": 3.7714285714285714,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.43956224,
+    "loss": 1.2682,
+    "grad_norm": 3.4037258625030518,
+    "learning_rate": 7.950186578116413e-07
+  },
+  {
+    "step": 67,
+    "epoch": 3.8285714285714287,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.439588352,
+    "loss": 1.2646,
+    "grad_norm": 3.657987117767334,
+    "learning_rate": 1.988865172754206e-07
+  },
+  {
+    "step": 68,
+    "epoch": 3.8857142857142857,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.4395392,
+    "loss": 1.2839,
+    "grad_norm": 4.297943115234375,
+    "learning_rate": 0.0
+  },
+  {
+    "step": 68,
+    "epoch": 3.8857142857142857,
+    "cpu_mem": 1.506648064,
+    "gpu_mem": 4.4395392,
+    "train_runtime": 380.012,
+    "train_samples_per_second": 11.779,
+    "train_steps_per_second": 0.179,
+    "total_flos": 0.0,
+    "train_loss": 1.5091714701231789
+  }
+]

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_c-r32-a2/adapter_config.json

@@ -0,0 +1,30 @@
+{
+  "alpha": 64,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "init_weight": "kaiming",
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "metric_tracking": false,
+  "modules_to_save": null,
+  "peft_type": "ABLATION",
+  "r": 32,
+  "revision": null,
+  "seed": 42,
+  "share_weights": false,
+  "target_modules": [
+    "up_proj",
+    "gate_proj",
+    "v_proj",
+    "k_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj"
+  ],
+  "task_type": null,
+  "track_n": 100,
+  "variant": "B"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_c-r32-a2/eval_results.json

@@ -0,0 +1,4 @@
+{
+    "task": "arc_c",
+    "results": 0.3293515358361775
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_c-r32-a2/training_configuration.json

@@ -0,0 +1,38 @@
+{
+  "model_id": "TinyLlama/TinyLlama_v1.1",
+  "dataset": {
+    "name": "ARC_C",
+    "dataset_id": "allenai/ai2_arc",
+    "preprocess_id": "arc_train_deepeval"
+  },
+  "peft_config": {
+    "method": "abl_B",
+    "rank": 32,
+    "alpha": 64,
+    "dropout": 0.0,
+    "bias": "none",
+    "target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "down_proj",
+      "up_proj"
+    ],
+    "trainable_parameter_count": 13009920
+  },
+  "training_config": {
+    "max_dataset_length": null,
+    "batch_size": 64,
+    "per_device_batch_size": 32,
+    "gradient_accumulation_steps": 2,
+    "learning_rate": 0.0003,
+    "num_epochs": 4,
+    "warmup_ratio": 0.1
+  },
+  "model_name": "TinyLlama_v1.1-abl_B-arc_c-r32-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_c-r32-a2",
+  "seed": 42,
+  "timestamp": "2025-09-01T03:38:45.959130"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_c-r32-a2/training_logs.json

@@ -0,0 +1,625 @@
+[
+  {
+    "step": 1,
+    "epoch": 0.05714285714285714,
+    "cpu_mem": 1.49262336,
+    "gpu_mem": 4.519966208,
+    "loss": 4.4614,
+    "grad_norm": 15.242693901062012,
+    "learning_rate": 4.285714285714285e-05
+  },
+  {
+    "step": 2,
+    "epoch": 0.11428571428571428,
+    "cpu_mem": 1.492819968,
+    "gpu_mem": 4.62403584,
+    "loss": 4.6994,
+    "grad_norm": 15.576388359069824,
+    "learning_rate": 8.57142857142857e-05
+  },
+  {
+    "step": 3,
+    "epoch": 0.17142857142857143,
+    "cpu_mem": 1.492819968,
+    "gpu_mem": 4.62406656,
+    "loss": 3.5902,
+    "grad_norm": 14.862706184387207,
+    "learning_rate": 0.00012857142857142855
+  },
+  {
+    "step": 4,
+    "epoch": 0.22857142857142856,
+    "cpu_mem": 1.492819968,
+    "gpu_mem": 4.624032768,
+    "loss": 2.4093,
+    "grad_norm": 8.196996688842773,
+    "learning_rate": 0.0001714285714285714
+  },
+  {
+    "step": 5,
+    "epoch": 0.2857142857142857,
+    "cpu_mem": 1.492819968,
+    "gpu_mem": 4.62402048,
+    "loss": 1.8535,
+    "grad_norm": 4.226119518280029,
+    "learning_rate": 0.00021428571428571427
+  },
+  {
+    "step": 6,
+    "epoch": 0.34285714285714286,
+    "cpu_mem": 1.492819968,
+    "gpu_mem": 4.624083456,
+    "loss": 1.7171,
+    "grad_norm": 1.8783652782440186,
+    "learning_rate": 0.0002571428571428571
+  },
+  {
+    "step": 7,
+    "epoch": 0.4,
+    "cpu_mem": 1.492819968,
+    "gpu_mem": 4.6240896,
+    "loss": 1.5023,
+    "grad_norm": 0.8889955282211304,
+    "learning_rate": 0.0003
+  },
+  {
+    "step": 8,
+    "epoch": 0.45714285714285713,
+    "cpu_mem": 1.492819968,
+    "gpu_mem": 4.624048128,
+    "loss": 1.4388,
+    "grad_norm": 0.607149064540863,
+    "learning_rate": 0.00029980111348272456
+  },
+  {
+    "step": 9,
+    "epoch": 0.5142857142857142,
+    "cpu_mem": 1.492819968,
+    "gpu_mem": 4.62404352,
+    "loss": 1.4488,
+    "grad_norm": 1.5999906063079834,
+    "learning_rate": 0.00029920498134218835
+  },
+  {
+    "step": 10,
+    "epoch": 0.5714285714285714,
+    "cpu_mem": 1.492819968,
+    "gpu_mem": 4.624032768,
+    "loss": 1.4183,
+    "grad_norm": 0.8587966561317444,
+    "learning_rate": 0.0002982131844136615
+  },
+  {
+    "step": 11,
+    "epoch": 0.6285714285714286,
+    "cpu_mem": 1.492819968,
+    "gpu_mem": 4.62404352,
+    "loss": 1.3995,
+    "grad_norm": 0.7488240599632263,
+    "learning_rate": 0.0002968283527643036
+  },
+  {
+    "step": 12,
+    "epoch": 0.6857142857142857,
+    "cpu_mem": 1.492819968,
+    "gpu_mem": 4.624068096,
+    "loss": 1.4221,
+    "grad_norm": 0.9820311665534973,
+    "learning_rate": 0.000295054158718698
+  },
+  {
+    "step": 13,
+    "epoch": 0.7428571428571429,
+    "cpu_mem": 1.492819968,
+    "gpu_mem": 4.624068096,
+    "loss": 1.3384,
+    "grad_norm": 0.888247013092041,
+    "learning_rate": 0.00029289530712050735
+  },
+  {
+    "step": 14,
+    "epoch": 0.8,
+    "cpu_mem": 1.492819968,
+    "gpu_mem": 4.624015872,
+    "loss": 1.5276,
+    "grad_norm": 1.6412670612335205,
+    "learning_rate": 0.000290357522856074
+  },
+  {
+    "step": 15,
+    "epoch": 0.8571428571428571,
+    "cpu_mem": 1.492819968,
+    "gpu_mem": 4.624091136,
+    "loss": 1.6764,
+    "grad_norm": 2.260718822479248,
+    "learning_rate": 0.0002874475356730507
+  },
+  {
+    "step": 16,
+    "epoch": 0.9142857142857143,
+    "cpu_mem": 1.492819968,
+    "gpu_mem": 4.624084992,
+    "loss": 1.4155,
+    "grad_norm": 0.7070791125297546,
+    "learning_rate": 0.0002841730623343193
+  },
+  {
+    "step": 17,
+    "epoch": 0.9714285714285714,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.6240896,
+    "loss": 1.3948,
+    "grad_norm": 0.6293061971664429,
+    "learning_rate": 0.00028054278615452326
+  },
+  {
+    "step": 18,
+    "epoch": 1.0285714285714285,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676086272,
+    "loss": 1.9828,
+    "grad_norm": 0.523653507232666,
+    "learning_rate": 0.0002765663339734778
+  },
+  {
+    "step": 19,
+    "epoch": 1.0857142857142856,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676084736,
+    "loss": 1.3436,
+    "grad_norm": 0.35439494252204895,
+    "learning_rate": 0.00027225425062752165
+  },
+  {
+    "step": 20,
+    "epoch": 1.1428571428571428,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.67606016,
+    "loss": 1.4187,
+    "grad_norm": 0.9338923096656799,
+    "learning_rate": 0.0002676179709865066
+  },
+  {
+    "step": 21,
+    "epoch": 1.2,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.67606784,
+    "loss": 1.3475,
+    "grad_norm": 0.28095149993896484,
+    "learning_rate": 0.0002626697896305779
+  },
+  {
+    "step": 22,
+    "epoch": 1.2571428571428571,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676097024,
+    "loss": 1.3209,
+    "grad_norm": 0.2404586225748062,
+    "learning_rate": 0.000257422828247159
+  },
+  {
+    "step": 23,
+    "epoch": 1.3142857142857143,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676126208,
+    "loss": 1.3192,
+    "grad_norm": 0.388563334941864,
+    "learning_rate": 0.00025189100083459397
+  },
+  {
+    "step": 24,
+    "epoch": 1.3714285714285714,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676069376,
+    "loss": 1.3566,
+    "grad_norm": 0.3809518814086914,
+    "learning_rate": 0.0002460889768047263
+  },
+  {
+    "step": 25,
+    "epoch": 1.4285714285714286,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676138496,
+    "loss": 1.2867,
+    "grad_norm": 0.30288708209991455,
+    "learning_rate": 0.00024003214208225522
+  },
+  {
+    "step": 26,
+    "epoch": 1.4857142857142858,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676095488,
+    "loss": 1.3138,
+    "grad_norm": 0.23125135898590088,
+    "learning_rate": 0.00023373655830402968
+  },
+  {
+    "step": 27,
+    "epoch": 1.5428571428571427,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676054016,
+    "loss": 1.36,
+    "grad_norm": 0.4470292925834656,
+    "learning_rate": 0.00022721892022647462
+  },
+  {
+    "step": 28,
+    "epoch": 1.6,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676100096,
+    "loss": 1.5448,
+    "grad_norm": 1.1614753007888794,
+    "learning_rate": 0.000220496511454098
+  },
+  {
+    "step": 29,
+    "epoch": 1.657142857142857,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676095488,
+    "loss": 1.3743,
+    "grad_norm": 0.4348692297935486,
+    "learning_rate": 0.0002135871586064791
+  },
+  {
+    "step": 30,
+    "epoch": 1.7142857142857144,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676084736,
+    "loss": 1.3902,
+    "grad_norm": 0.5657250285148621,
+    "learning_rate": 0.00020650918404527775
+  },
+  {
+    "step": 31,
+    "epoch": 1.7714285714285714,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676115456,
+    "loss": 1.3738,
+    "grad_norm": 0.4655613303184509,
+    "learning_rate": 0.00019928135728662522
+  },
+  {
+    "step": 32,
+    "epoch": 1.8285714285714287,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676124672,
+    "loss": 1.3621,
+    "grad_norm": 0.28164371848106384,
+    "learning_rate": 0.00019192284522774142
+  },
+  {
+    "step": 33,
+    "epoch": 1.8857142857142857,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676104704,
+    "loss": 1.3797,
+    "grad_norm": 0.26845723390579224,
+    "learning_rate": 0.00018445316131976934
+  },
+  {
+    "step": 34,
+    "epoch": 1.9428571428571428,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.6760832,
+    "loss": 1.3641,
+    "grad_norm": 0.3237358331680298,
+    "learning_rate": 0.00017689211382161034
+  },
+  {
+    "step": 35,
+    "epoch": 2.0,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.675971072,
+    "loss": 2.0906,
+    "grad_norm": 0.5221468210220337,
+    "learning_rate": 0.00016925975327198266
+  },
+  {
+    "step": 36,
+    "epoch": 2.057142857142857,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.624061952,
+    "loss": 1.4109,
+    "grad_norm": 0.6460707187652588,
+    "learning_rate": 0.00016157631931899697
+  },
+  {
+    "step": 37,
+    "epoch": 2.1142857142857143,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.624071168,
+    "loss": 1.3558,
+    "grad_norm": 0.27621448040008545,
+    "learning_rate": 0.0001538621870482483
+  },
+  {
+    "step": 38,
+    "epoch": 2.1714285714285713,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.624041984,
+    "loss": 1.3643,
+    "grad_norm": 0.31860432028770447,
+    "learning_rate": 0.00014613781295175172
+  },
+  {
+    "step": 39,
+    "epoch": 2.2285714285714286,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.624060416,
+    "loss": 1.3291,
+    "grad_norm": 0.14847567677497864,
+    "learning_rate": 0.00013842368068100303
+  },
+  {
+    "step": 40,
+    "epoch": 2.2857142857142856,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.624037376,
+    "loss": 1.3953,
+    "grad_norm": 0.37305229902267456,
+    "learning_rate": 0.00013074024672801731
+  },
+  {
+    "step": 41,
+    "epoch": 2.342857142857143,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.624038912,
+    "loss": 1.3988,
+    "grad_norm": 0.34824004769325256,
+    "learning_rate": 0.00012310788617838966
+  },
+  {
+    "step": 42,
+    "epoch": 2.4,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.624068096,
+    "loss": 1.3218,
+    "grad_norm": 0.4003921449184418,
+    "learning_rate": 0.00011554683868023067
+  },
+  {
+    "step": 43,
+    "epoch": 2.4571428571428573,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.624083456,
+    "loss": 1.3708,
+    "grad_norm": 0.5072891116142273,
+    "learning_rate": 0.00010807715477225858
+  },
+  {
+    "step": 44,
+    "epoch": 2.5142857142857142,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.624101888,
+    "loss": 1.3432,
+    "grad_norm": 0.2869243025779724,
+    "learning_rate": 0.00010071864271337478
+  },
+  {
+    "step": 45,
+    "epoch": 2.571428571428571,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.624055808,
+    "loss": 1.326,
+    "grad_norm": 0.1572774201631546,
+    "learning_rate": 9.34908159547222e-05
+  },
+  {
+    "step": 46,
+    "epoch": 2.6285714285714286,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.624049664,
+    "loss": 1.3203,
+    "grad_norm": 0.38407614827156067,
+    "learning_rate": 8.641284139352091e-05
+  },
+  {
+    "step": 47,
+    "epoch": 2.685714285714286,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.62404352,
+    "loss": 1.3093,
+    "grad_norm": 0.3907293677330017,
+    "learning_rate": 7.950348854590204e-05
+  },
+  {
+    "step": 48,
+    "epoch": 2.742857142857143,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.624048128,
+    "loss": 1.2786,
+    "grad_norm": 0.27290043234825134,
+    "learning_rate": 7.278107977352543e-05
+  },
+  {
+    "step": 49,
+    "epoch": 2.8,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.624038912,
+    "loss": 1.3194,
+    "grad_norm": 0.35774123668670654,
+    "learning_rate": 6.626344169597031e-05
+  },
+  {
+    "step": 50,
+    "epoch": 2.857142857142857,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.62402048,
+    "loss": 1.3196,
+    "grad_norm": 0.16639018058776855,
+    "learning_rate": 5.996785791774478e-05
+  },
+  {
+    "step": 51,
+    "epoch": 2.914285714285714,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.624045056,
+    "loss": 1.3523,
+    "grad_norm": 0.514198899269104,
+    "learning_rate": 5.391102319527373e-05
+  },
+  {
+    "step": 52,
+    "epoch": 2.9714285714285715,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.624072704,
+    "loss": 1.3465,
+    "grad_norm": 0.18968720734119415,
+    "learning_rate": 4.8108999165406026e-05
+  },
+  {
+    "step": 53,
+    "epoch": 3.0285714285714285,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676081664,
+    "loss": 1.9683,
+    "grad_norm": 0.4923784136772156,
+    "learning_rate": 4.257717175284103e-05
+  },
+  {
+    "step": 54,
+    "epoch": 3.085714285714286,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676050944,
+    "loss": 1.3609,
+    "grad_norm": 0.3217475414276123,
+    "learning_rate": 3.733021036942205e-05
+  },
+  {
+    "step": 55,
+    "epoch": 3.142857142857143,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676084736,
+    "loss": 1.3716,
+    "grad_norm": 0.44956937432289124,
+    "learning_rate": 3.238202901349345e-05
+  },
+  {
+    "step": 56,
+    "epoch": 3.2,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676158464,
+    "loss": 1.3291,
+    "grad_norm": 0.2953140139579773,
+    "learning_rate": 2.774574937247831e-05
+  },
+  {
+    "step": 57,
+    "epoch": 3.257142857142857,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676101632,
+    "loss": 1.3283,
+    "grad_norm": 0.19999608397483826,
+    "learning_rate": 2.3433666026522153e-05
+  },
+  {
+    "step": 58,
+    "epoch": 3.314285714285714,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676095488,
+    "loss": 1.2358,
+    "grad_norm": 0.26531586050987244,
+    "learning_rate": 1.945721384547671e-05
+  },
+  {
+    "step": 59,
+    "epoch": 3.3714285714285714,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676146176,
+    "loss": 1.3186,
+    "grad_norm": 0.301336407661438,
+    "learning_rate": 1.5826937665680693e-05
+  },
+  {
+    "step": 60,
+    "epoch": 3.4285714285714284,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676072448,
+    "loss": 1.367,
+    "grad_norm": 0.4051586091518402,
+    "learning_rate": 1.2552464326949302e-05
+  },
+  {
+    "step": 61,
+    "epoch": 3.4857142857142858,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676086272,
+    "loss": 1.3862,
+    "grad_norm": 0.36729010939598083,
+    "learning_rate": 9.64247714392597e-06
+  },
+  {
+    "step": 62,
+    "epoch": 3.5428571428571427,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676087808,
+    "loss": 1.328,
+    "grad_norm": 0.2522547245025635,
+    "learning_rate": 7.104692879492624e-06
+  },
+  {
+    "step": 63,
+    "epoch": 3.6,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676077056,
+    "loss": 1.3359,
+    "grad_norm": 0.25604918599128723,
+    "learning_rate": 4.945841281301943e-06
+  },
+  {
+    "step": 64,
+    "epoch": 3.657142857142857,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676093952,
+    "loss": 1.3089,
+    "grad_norm": 0.3017137348651886,
+    "learning_rate": 3.1716472356963286e-06
+  },
+  {
+    "step": 65,
+    "epoch": 3.7142857142857144,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676115456,
+    "loss": 1.3253,
+    "grad_norm": 0.31271353363990784,
+    "learning_rate": 1.7868155863384415e-06
+  },
+  {
+    "step": 66,
+    "epoch": 3.7714285714285714,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.67610624,
+    "loss": 1.2868,
+    "grad_norm": 0.2766549587249756,
+    "learning_rate": 7.950186578116413e-07
+  },
+  {
+    "step": 67,
+    "epoch": 3.8285714285714287,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.676132352,
+    "loss": 1.3103,
+    "grad_norm": 0.2598274350166321,
+    "learning_rate": 1.988865172754206e-07
+  },
+  {
+    "step": 68,
+    "epoch": 3.8857142857142857,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.6760832,
+    "loss": 1.3174,
+    "grad_norm": 0.3739108741283417,
+    "learning_rate": 0.0
+  },
+  {
+    "step": 68,
+    "epoch": 3.8857142857142857,
+    "cpu_mem": 1.492414464,
+    "gpu_mem": 4.6760832,
+    "train_runtime": 383.1823,
+    "train_samples_per_second": 11.681,
+    "train_steps_per_second": 0.177,
+    "total_flos": 0.0,
+    "train_loss": 1.5497514447745155
+  }
+]

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_c-r8-a2/adapter_config.json

@@ -0,0 +1,30 @@
+{
+  "alpha": 16,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "init_weight": "kaiming",
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "metric_tracking": false,
+  "modules_to_save": null,
+  "peft_type": "ABLATION",
+  "r": 8,
+  "revision": null,
+  "seed": 42,
+  "share_weights": false,
+  "target_modules": [
+    "up_proj",
+    "gate_proj",
+    "v_proj",
+    "k_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj"
+  ],
+  "task_type": null,
+  "track_n": 100,
+  "variant": "B"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_c-r8-a2/eval_results.json

@@ -0,0 +1,4 @@
+{
+    "task": "arc_c",
+    "results": 0.38139931740614336
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_c-r8-a2/training_configuration.json

@@ -0,0 +1,38 @@
+{
+  "model_id": "TinyLlama/TinyLlama_v1.1",
+  "dataset": {
+    "name": "ARC_C",
+    "dataset_id": "allenai/ai2_arc",
+    "preprocess_id": "arc_train_deepeval"
+  },
+  "peft_config": {
+    "method": "abl_B",
+    "rank": 8,
+    "alpha": 16,
+    "dropout": 0.0,
+    "bias": "none",
+    "target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "down_proj",
+      "up_proj"
+    ],
+    "trainable_parameter_count": 3548160
+  },
+  "training_config": {
+    "max_dataset_length": null,
+    "batch_size": 64,
+    "per_device_batch_size": 32,
+    "gradient_accumulation_steps": 2,
+    "learning_rate": 0.0003,
+    "num_epochs": 4,
+    "warmup_ratio": 0.1
+  },
+  "model_name": "TinyLlama_v1.1-abl_B-arc_c-r8-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_c-r8-a2",
+  "seed": 42,
+  "timestamp": "2025-08-31T20:33:13.257777"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_c-r8-a2/training_logs.json

@@ -0,0 +1,625 @@
+[
+  {
+    "step": 1,
+    "epoch": 0.05714285714285714,
+    "cpu_mem": 1.491714048,
+    "gpu_mem": 4.444272128,
+    "loss": 4.4614,
+    "grad_norm": 87.54730224609375,
+    "learning_rate": 4.285714285714285e-05
+  },
+  {
+    "step": 2,
+    "epoch": 0.11428571428571428,
+    "cpu_mem": 1.491714048,
+    "gpu_mem": 4.47264768,
+    "loss": 4.6994,
+    "grad_norm": 88.6588363647461,
+    "learning_rate": 8.57142857142857e-05
+  },
+  {
+    "step": 3,
+    "epoch": 0.17142857142857143,
+    "cpu_mem": 1.491714048,
+    "gpu_mem": 4.4726784,
+    "loss": 2.9296,
+    "grad_norm": 59.65691375732422,
+    "learning_rate": 0.00012857142857142855
+  },
+  {
+    "step": 4,
+    "epoch": 0.22857142857142856,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472644608,
+    "loss": 1.9248,
+    "grad_norm": 44.71550750732422,
+    "learning_rate": 0.0001714285714285714
+  },
+  {
+    "step": 5,
+    "epoch": 0.2857142857142857,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.47263232,
+    "loss": 1.5841,
+    "grad_norm": 6.350735664367676,
+    "learning_rate": 0.00021428571428571427
+  },
+  {
+    "step": 6,
+    "epoch": 0.34285714285714286,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472695296,
+    "loss": 1.4618,
+    "grad_norm": 3.4582574367523193,
+    "learning_rate": 0.0002571428571428571
+  },
+  {
+    "step": 7,
+    "epoch": 0.4,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.47270144,
+    "loss": 1.468,
+    "grad_norm": 4.226593971252441,
+    "learning_rate": 0.0003
+  },
+  {
+    "step": 8,
+    "epoch": 0.45714285714285713,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472659968,
+    "loss": 1.4115,
+    "grad_norm": 4.548079490661621,
+    "learning_rate": 0.00029980111348272456
+  },
+  {
+    "step": 9,
+    "epoch": 0.5142857142857142,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.47265536,
+    "loss": 1.3269,
+    "grad_norm": 3.2502329349517822,
+    "learning_rate": 0.00029920498134218835
+  },
+  {
+    "step": 10,
+    "epoch": 0.5714285714285714,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472644608,
+    "loss": 1.5354,
+    "grad_norm": 5.945767402648926,
+    "learning_rate": 0.0002982131844136615
+  },
+  {
+    "step": 11,
+    "epoch": 0.6285714285714286,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.47265536,
+    "loss": 1.4694,
+    "grad_norm": 4.522561550140381,
+    "learning_rate": 0.0002968283527643036
+  },
+  {
+    "step": 12,
+    "epoch": 0.6857142857142857,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472679936,
+    "loss": 1.3612,
+    "grad_norm": 2.172001361846924,
+    "learning_rate": 0.000295054158718698
+  },
+  {
+    "step": 13,
+    "epoch": 0.7428571428571429,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472679936,
+    "loss": 1.3933,
+    "grad_norm": 4.974757194519043,
+    "learning_rate": 0.00029289530712050735
+  },
+  {
+    "step": 14,
+    "epoch": 0.8,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472627712,
+    "loss": 1.6926,
+    "grad_norm": 8.53721809387207,
+    "learning_rate": 0.000290357522856074
+  },
+  {
+    "step": 15,
+    "epoch": 0.8571428571428571,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472702976,
+    "loss": 1.682,
+    "grad_norm": 7.365475654602051,
+    "learning_rate": 0.0002874475356730507
+  },
+  {
+    "step": 16,
+    "epoch": 0.9142857142857143,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472696832,
+    "loss": 1.3926,
+    "grad_norm": 1.4433025121688843,
+    "learning_rate": 0.0002841730623343193
+  },
+  {
+    "step": 17,
+    "epoch": 0.9714285714285714,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.47270144,
+    "loss": 1.3892,
+    "grad_norm": 2.147620677947998,
+    "learning_rate": 0.00028054278615452326
+  },
+  {
+    "step": 18,
+    "epoch": 1.0285714285714285,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486851072,
+    "loss": 2.0418,
+    "grad_norm": 2.8316280841827393,
+    "learning_rate": 0.0002765663339734778
+  },
+  {
+    "step": 19,
+    "epoch": 1.0857142857142856,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486849536,
+    "loss": 1.3974,
+    "grad_norm": 2.231830358505249,
+    "learning_rate": 0.00027225425062752165
+  },
+  {
+    "step": 20,
+    "epoch": 1.1428571428571428,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.48682496,
+    "loss": 1.4072,
+    "grad_norm": 3.0868797302246094,
+    "learning_rate": 0.0002676179709865066
+  },
+  {
+    "step": 21,
+    "epoch": 1.2,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.48683264,
+    "loss": 1.362,
+    "grad_norm": 1.2280853986740112,
+    "learning_rate": 0.0002626697896305779
+  },
+  {
+    "step": 22,
+    "epoch": 1.2571428571428571,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486861824,
+    "loss": 1.3332,
+    "grad_norm": 1.2861239910125732,
+    "learning_rate": 0.000257422828247159
+  },
+  {
+    "step": 23,
+    "epoch": 1.3142857142857143,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486891008,
+    "loss": 1.2994,
+    "grad_norm": 0.9643853306770325,
+    "learning_rate": 0.00025189100083459397
+  },
+  {
+    "step": 24,
+    "epoch": 1.3714285714285714,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486834176,
+    "loss": 1.3951,
+    "grad_norm": 2.0752029418945312,
+    "learning_rate": 0.0002460889768047263
+  },
+  {
+    "step": 25,
+    "epoch": 1.4285714285714286,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486903296,
+    "loss": 1.3398,
+    "grad_norm": 1.7827640771865845,
+    "learning_rate": 0.00024003214208225522
+  },
+  {
+    "step": 26,
+    "epoch": 1.4857142857142858,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486860288,
+    "loss": 1.3324,
+    "grad_norm": 1.2238781452178955,
+    "learning_rate": 0.00023373655830402968
+  },
+  {
+    "step": 27,
+    "epoch": 1.5428571428571427,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486818816,
+    "loss": 1.3679,
+    "grad_norm": 1.499398112297058,
+    "learning_rate": 0.00022721892022647462
+  },
+  {
+    "step": 28,
+    "epoch": 1.6,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486864896,
+    "loss": 1.4935,
+    "grad_norm": 3.0713043212890625,
+    "learning_rate": 0.000220496511454098
+  },
+  {
+    "step": 29,
+    "epoch": 1.657142857142857,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486860288,
+    "loss": 1.3437,
+    "grad_norm": 0.9689168930053711,
+    "learning_rate": 0.0002135871586064791
+  },
+  {
+    "step": 30,
+    "epoch": 1.7142857142857144,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486849536,
+    "loss": 1.3452,
+    "grad_norm": 1.1040581464767456,
+    "learning_rate": 0.00020650918404527775
+  },
+  {
+    "step": 31,
+    "epoch": 1.7714285714285714,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486880256,
+    "loss": 1.3348,
+    "grad_norm": 1.1760023832321167,
+    "learning_rate": 0.00019928135728662522
+  },
+  {
+    "step": 32,
+    "epoch": 1.8285714285714287,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486889472,
+    "loss": 1.3825,
+    "grad_norm": 1.7875124216079712,
+    "learning_rate": 0.00019192284522774142
+  },
+  {
+    "step": 33,
+    "epoch": 1.8857142857142857,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486869504,
+    "loss": 1.4018,
+    "grad_norm": 2.0181446075439453,
+    "learning_rate": 0.00018445316131976934
+  },
+  {
+    "step": 34,
+    "epoch": 1.9428571428571428,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486848,
+    "loss": 1.362,
+    "grad_norm": 1.2658624649047852,
+    "learning_rate": 0.00017689211382161034
+  },
+  {
+    "step": 35,
+    "epoch": 2.0,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486735872,
+    "loss": 2.084,
+    "grad_norm": 2.159017562866211,
+    "learning_rate": 0.00016925975327198266
+  },
+  {
+    "step": 36,
+    "epoch": 2.057142857142857,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472673792,
+    "loss": 1.3323,
+    "grad_norm": 1.2686223983764648,
+    "learning_rate": 0.00016157631931899697
+  },
+  {
+    "step": 37,
+    "epoch": 2.1142857142857143,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472683008,
+    "loss": 1.4262,
+    "grad_norm": 2.404953956604004,
+    "learning_rate": 0.0001538621870482483
+  },
+  {
+    "step": 38,
+    "epoch": 2.1714285714285713,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472653824,
+    "loss": 1.3707,
+    "grad_norm": 1.484898567199707,
+    "learning_rate": 0.00014613781295175172
+  },
+  {
+    "step": 39,
+    "epoch": 2.2285714285714286,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472672256,
+    "loss": 1.4011,
+    "grad_norm": 1.994706153869629,
+    "learning_rate": 0.00013842368068100303
+  },
+  {
+    "step": 40,
+    "epoch": 2.2857142857142856,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472649216,
+    "loss": 1.3828,
+    "grad_norm": 1.137582540512085,
+    "learning_rate": 0.00013074024672801731
+  },
+  {
+    "step": 41,
+    "epoch": 2.342857142857143,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472650752,
+    "loss": 1.3832,
+    "grad_norm": 1.2223880290985107,
+    "learning_rate": 0.00012310788617838966
+  },
+  {
+    "step": 42,
+    "epoch": 2.4,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472679936,
+    "loss": 1.3495,
+    "grad_norm": 1.9196335077285767,
+    "learning_rate": 0.00011554683868023067
+  },
+  {
+    "step": 43,
+    "epoch": 2.4571428571428573,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472695296,
+    "loss": 1.3356,
+    "grad_norm": 1.193792700767517,
+    "learning_rate": 0.00010807715477225858
+  },
+  {
+    "step": 44,
+    "epoch": 2.5142857142857142,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472713728,
+    "loss": 1.3222,
+    "grad_norm": 0.7033736705780029,
+    "learning_rate": 0.00010071864271337478
+  },
+  {
+    "step": 45,
+    "epoch": 2.571428571428571,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472667648,
+    "loss": 1.2936,
+    "grad_norm": 0.6882470846176147,
+    "learning_rate": 9.34908159547222e-05
+  },
+  {
+    "step": 46,
+    "epoch": 2.6285714285714286,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472661504,
+    "loss": 1.2888,
+    "grad_norm": 1.0970045328140259,
+    "learning_rate": 8.641284139352091e-05
+  },
+  {
+    "step": 47,
+    "epoch": 2.685714285714286,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.47265536,
+    "loss": 1.2839,
+    "grad_norm": 0.9996795058250427,
+    "learning_rate": 7.950348854590204e-05
+  },
+  {
+    "step": 48,
+    "epoch": 2.742857142857143,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472659968,
+    "loss": 1.2405,
+    "grad_norm": 0.8805912137031555,
+    "learning_rate": 7.278107977352543e-05
+  },
+  {
+    "step": 49,
+    "epoch": 2.8,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472650752,
+    "loss": 1.2877,
+    "grad_norm": 1.0169543027877808,
+    "learning_rate": 6.626344169597031e-05
+  },
+  {
+    "step": 50,
+    "epoch": 2.857142857142857,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.47263232,
+    "loss": 1.3195,
+    "grad_norm": 1.2063992023468018,
+    "learning_rate": 5.996785791774478e-05
+  },
+  {
+    "step": 51,
+    "epoch": 2.914285714285714,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472656896,
+    "loss": 1.2716,
+    "grad_norm": 1.298764944076538,
+    "learning_rate": 5.391102319527373e-05
+  },
+  {
+    "step": 52,
+    "epoch": 2.9714285714285715,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.472684544,
+    "loss": 1.3397,
+    "grad_norm": 1.3159348964691162,
+    "learning_rate": 4.8108999165406026e-05
+  },
+  {
+    "step": 53,
+    "epoch": 3.0285714285714285,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486846464,
+    "loss": 1.9005,
+    "grad_norm": 1.5513451099395752,
+    "learning_rate": 4.257717175284103e-05
+  },
+  {
+    "step": 54,
+    "epoch": 3.085714285714286,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486815744,
+    "loss": 1.3357,
+    "grad_norm": 1.5055204629898071,
+    "learning_rate": 3.733021036942205e-05
+  },
+  {
+    "step": 55,
+    "epoch": 3.142857142857143,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486849536,
+    "loss": 1.338,
+    "grad_norm": 1.9263288974761963,
+    "learning_rate": 3.238202901349345e-05
+  },
+  {
+    "step": 56,
+    "epoch": 3.2,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486923264,
+    "loss": 1.2948,
+    "grad_norm": 1.05913245677948,
+    "learning_rate": 2.774574937247831e-05
+  },
+  {
+    "step": 57,
+    "epoch": 3.257142857142857,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486866432,
+    "loss": 1.2813,
+    "grad_norm": 1.0687165260314941,
+    "learning_rate": 2.3433666026522153e-05
+  },
+  {
+    "step": 58,
+    "epoch": 3.314285714285714,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486860288,
+    "loss": 1.1917,
+    "grad_norm": 1.3125996589660645,
+    "learning_rate": 1.945721384547671e-05
+  },
+  {
+    "step": 59,
+    "epoch": 3.3714285714285714,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486910976,
+    "loss": 1.2695,
+    "grad_norm": 1.1107628345489502,
+    "learning_rate": 1.5826937665680693e-05
+  },
+  {
+    "step": 60,
+    "epoch": 3.4285714285714284,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486837248,
+    "loss": 1.3273,
+    "grad_norm": 1.5512784719467163,
+    "learning_rate": 1.2552464326949302e-05
+  },
+  {
+    "step": 61,
+    "epoch": 3.4857142857142858,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486851072,
+    "loss": 1.3364,
+    "grad_norm": 1.1313272714614868,
+    "learning_rate": 9.64247714392597e-06
+  },
+  {
+    "step": 62,
+    "epoch": 3.5428571428571427,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486852608,
+    "loss": 1.3142,
+    "grad_norm": 1.285913109779358,
+    "learning_rate": 7.104692879492624e-06
+  },
+  {
+    "step": 63,
+    "epoch": 3.6,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486841856,
+    "loss": 1.3043,
+    "grad_norm": 1.1523516178131104,
+    "learning_rate": 4.945841281301943e-06
+  },
+  {
+    "step": 64,
+    "epoch": 3.657142857142857,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486858752,
+    "loss": 1.2581,
+    "grad_norm": 1.2959764003753662,
+    "learning_rate": 3.1716472356963286e-06
+  },
+  {
+    "step": 65,
+    "epoch": 3.7142857142857144,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486880256,
+    "loss": 1.307,
+    "grad_norm": 1.4972169399261475,
+    "learning_rate": 1.7868155863384415e-06
+  },
+  {
+    "step": 66,
+    "epoch": 3.7714285714285714,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.48687104,
+    "loss": 1.2616,
+    "grad_norm": 1.236932635307312,
+    "learning_rate": 7.950186578116413e-07
+  },
+  {
+    "step": 67,
+    "epoch": 3.8285714285714287,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486897152,
+    "loss": 1.2551,
+    "grad_norm": 1.1077874898910522,
+    "learning_rate": 1.988865172754206e-07
+  },
+  {
+    "step": 68,
+    "epoch": 3.8857142857142857,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486848,
+    "loss": 1.2744,
+    "grad_norm": 1.5151118040084839,
+    "learning_rate": 0.0
+  },
+  {
+    "step": 68,
+    "epoch": 3.8857142857142857,
+    "cpu_mem": 1.491910656,
+    "gpu_mem": 4.486848,
+    "train_runtime": 380.8275,
+    "train_samples_per_second": 11.753,
+    "train_steps_per_second": 0.179,
+    "total_flos": 0.0,
+    "train_loss": 1.5159650935846216
+  }
+]

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_e-r2-a2/adapter_config.json

@@ -0,0 +1,30 @@
+{
+  "alpha": 4,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "init_weight": "kaiming",
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "metric_tracking": false,
+  "modules_to_save": null,
+  "peft_type": "ABLATION",
+  "r": 2,
+  "revision": null,
+  "seed": 42,
+  "share_weights": false,
+  "target_modules": [
+    "up_proj",
+    "gate_proj",
+    "v_proj",
+    "k_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj"
+  ],
+  "task_type": null,
+  "track_n": 100,
+  "variant": "B"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_e-r2-a2/eval_results.json

@@ -0,0 +1,4 @@
+{
+    "task": "arc_e",
+    "results": 0.48695286195286197
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_e-r2-a2/training_configuration.json

@@ -0,0 +1,38 @@
+{
+  "model_id": "TinyLlama/TinyLlama_v1.1",
+  "dataset": {
+    "name": "ARC_E",
+    "dataset_id": "allenai/ai2_arc",
+    "preprocess_id": "arc_train_deepeval"
+  },
+  "peft_config": {
+    "method": "abl_B",
+    "rank": 2,
+    "alpha": 4,
+    "dropout": 0.0,
+    "bias": "none",
+    "target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "down_proj",
+      "up_proj"
+    ],
+    "trainable_parameter_count": 1182720
+  },
+  "training_config": {
+    "max_dataset_length": null,
+    "batch_size": 64,
+    "per_device_batch_size": 32,
+    "gradient_accumulation_steps": 2,
+    "learning_rate": 0.0003,
+    "num_epochs": 4,
+    "warmup_ratio": 0.1
+  },
+  "model_name": "TinyLlama_v1.1-abl_B-arc_e-r2-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_e-r2-a2",
+  "seed": 42,
+  "timestamp": "2025-08-31T12:54:20.156634"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_e-r2-a2/training_logs.json

@@ -0,0 +1,1273 @@
+[
+  {
+    "step": 1,
+    "epoch": 0.028169014084507043,
+    "cpu_mem": 1.50663168,
+    "gpu_mem": 4.425293312,
+    "loss": 4.6319,
+    "grad_norm": 259.6687927246094,
+    "learning_rate": 2.1428571428571425e-05
+  },
+  {
+    "step": 2,
+    "epoch": 0.056338028169014086,
+    "cpu_mem": 1.506828288,
+    "gpu_mem": 4.434816,
+    "loss": 4.4578,
+    "grad_norm": 265.4711608886719,
+    "learning_rate": 4.285714285714285e-05
+  },
+  {
+    "step": 3,
+    "epoch": 0.08450704225352113,
+    "cpu_mem": 1.507024896,
+    "gpu_mem": 4.434794496,
+    "loss": 3.1953,
+    "grad_norm": 169.68458557128906,
+    "learning_rate": 6.428571428571427e-05
+  },
+  {
+    "step": 4,
+    "epoch": 0.11267605633802817,
+    "cpu_mem": 1.507221504,
+    "gpu_mem": 4.434772992,
+    "loss": 2.375,
+    "grad_norm": 182.9171600341797,
+    "learning_rate": 8.57142857142857e-05
+  },
+  {
+    "step": 5,
+    "epoch": 0.14084507042253522,
+    "cpu_mem": 1.507418112,
+    "gpu_mem": 4.434814464,
+    "loss": 1.6471,
+    "grad_norm": 24.682857513427734,
+    "learning_rate": 0.00010714285714285714
+  },
+  {
+    "step": 6,
+    "epoch": 0.16901408450704225,
+    "cpu_mem": 1.507418112,
+    "gpu_mem": 4.434789888,
+    "loss": 1.5066,
+    "grad_norm": 23.628555297851562,
+    "learning_rate": 0.00012857142857142855
+  },
+  {
+    "step": 7,
+    "epoch": 0.19718309859154928,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434812928,
+    "loss": 1.4403,
+    "grad_norm": 15.431472778320312,
+    "learning_rate": 0.00015
+  },
+  {
+    "step": 8,
+    "epoch": 0.22535211267605634,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434771456,
+    "loss": 1.3832,
+    "grad_norm": 12.077364921569824,
+    "learning_rate": 0.0001714285714285714
+  },
+  {
+    "step": 9,
+    "epoch": 0.2535211267605634,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434772992,
+    "loss": 1.37,
+    "grad_norm": 13.536406517028809,
+    "learning_rate": 0.00019285714285714286
+  },
+  {
+    "step": 10,
+    "epoch": 0.28169014084507044,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434768384,
+    "loss": 1.6613,
+    "grad_norm": 55.036895751953125,
+    "learning_rate": 0.00021428571428571427
+  },
+  {
+    "step": 11,
+    "epoch": 0.30985915492957744,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43484672,
+    "loss": 1.4147,
+    "grad_norm": 16.728759765625,
+    "learning_rate": 0.00023571428571428569
+  },
+  {
+    "step": 12,
+    "epoch": 0.3380281690140845,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434820608,
+    "loss": 1.3499,
+    "grad_norm": 8.593183517456055,
+    "learning_rate": 0.0002571428571428571
+  },
+  {
+    "step": 13,
+    "epoch": 0.36619718309859156,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434771456,
+    "loss": 1.4139,
+    "grad_norm": 10.364399909973145,
+    "learning_rate": 0.00027857142857142854
+  },
+  {
+    "step": 14,
+    "epoch": 0.39436619718309857,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43479296,
+    "loss": 1.3907,
+    "grad_norm": 8.241915702819824,
+    "learning_rate": 0.0003
+  },
+  {
+    "step": 15,
+    "epoch": 0.4225352112676056,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43476992,
+    "loss": 1.3206,
+    "grad_norm": 3.446953296661377,
+    "learning_rate": 0.0002999533773001224
+  },
+  {
+    "step": 16,
+    "epoch": 0.4507042253521127,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434774528,
+    "loss": 1.4047,
+    "grad_norm": 7.349297523498535,
+    "learning_rate": 0.0002998135381828383
+  },
+  {
+    "step": 17,
+    "epoch": 0.4788732394366197,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434811392,
+    "loss": 1.3294,
+    "grad_norm": 4.483480453491211,
+    "learning_rate": 0.00029958056957717696
+  },
+  {
+    "step": 18,
+    "epoch": 0.5070422535211268,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434822144,
+    "loss": 1.339,
+    "grad_norm": 5.043764114379883,
+    "learning_rate": 0.0002992546163048102
+  },
+  {
+    "step": 19,
+    "epoch": 0.5352112676056338,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434765312,
+    "loss": 1.3036,
+    "grad_norm": 5.584592819213867,
+    "learning_rate": 0.0002988358809900258
+  },
+  {
+    "step": 20,
+    "epoch": 0.5633802816901409,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434835968,
+    "loss": 1.3402,
+    "grad_norm": 2.576645851135254,
+    "learning_rate": 0.0002983246239337692
+  },
+  {
+    "step": 21,
+    "epoch": 0.5915492957746479,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434834432,
+    "loss": 1.3322,
+    "grad_norm": 4.193614959716797,
+    "learning_rate": 0.0002977211629518312
+  },
+  {
+    "step": 22,
+    "epoch": 0.6197183098591549,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434791424,
+    "loss": 1.3737,
+    "grad_norm": 4.681839942932129,
+    "learning_rate": 0.00029702587317728153
+  },
+  {
+    "step": 23,
+    "epoch": 0.647887323943662,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43480832,
+    "loss": 1.33,
+    "grad_norm": 3.810232400894165,
+    "learning_rate": 0.0002962391868272735
+  },
+  {
+    "step": 24,
+    "epoch": 0.676056338028169,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434765312,
+    "loss": 1.3134,
+    "grad_norm": 3.2136316299438477,
+    "learning_rate": 0.00029536159293436166
+  },
+  {
+    "step": 25,
+    "epoch": 0.704225352112676,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434794496,
+    "loss": 1.512,
+    "grad_norm": 13.608572006225586,
+    "learning_rate": 0.00029439363704250176
+  },
+  {
+    "step": 26,
+    "epoch": 0.7323943661971831,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434774528,
+    "loss": 1.6166,
+    "grad_norm": 14.765206336975098,
+    "learning_rate": 0.00029333592086792107
+  },
+  {
+    "step": 27,
+    "epoch": 0.7605633802816901,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43480064,
+    "loss": 1.3597,
+    "grad_norm": 4.8024702072143555,
+    "learning_rate": 0.0002921891019250697
+  },
+  {
+    "step": 28,
+    "epoch": 0.7887323943661971,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43480064,
+    "loss": 1.3922,
+    "grad_norm": 3.2379608154296875,
+    "learning_rate": 0.0002909538931178862
+  },
+  {
+    "step": 29,
+    "epoch": 0.8169014084507042,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434779136,
+    "loss": 1.3246,
+    "grad_norm": 4.987905025482178,
+    "learning_rate": 0.00028963106229663063
+  },
+  {
+    "step": 30,
+    "epoch": 0.8450704225352113,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43476992,
+    "loss": 1.3724,
+    "grad_norm": 4.277651786804199,
+    "learning_rate": 0.00028822143178056114
+  },
+  {
+    "step": 31,
+    "epoch": 0.8732394366197183,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434788352,
+    "loss": 1.3908,
+    "grad_norm": 3.794233560562134,
+    "learning_rate": 0.00028672587784675096
+  },
+  {
+    "step": 32,
+    "epoch": 0.9014084507042254,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434811392,
+    "loss": 1.35,
+    "grad_norm": 3.4455714225769043,
+    "learning_rate": 0.0002851453301853628
+  },
+  {
+    "step": 33,
+    "epoch": 0.9295774647887324,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43480832,
+    "loss": 1.3438,
+    "grad_norm": 2.1978235244750977,
+    "learning_rate": 0.00028348077132172027
+  },
+  {
+    "step": 34,
+    "epoch": 0.9577464788732394,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434811392,
+    "loss": 1.3851,
+    "grad_norm": 3.6687681674957275,
+    "learning_rate": 0.0002817332360055343
+  },
+  {
+    "step": 35,
+    "epoch": 0.9859154929577465,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43479296,
+    "loss": 1.3028,
+    "grad_norm": 2.3399291038513184,
+    "learning_rate": 0.0002799038105676658
+  },
+  {
+    "step": 36,
+    "epoch": 1.0140845070422535,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439514624,
+    "loss": 1.9693,
+    "grad_norm": 6.734819412231445,
+    "learning_rate": 0.0002779936322448233
+  },
+  {
+    "step": 37,
+    "epoch": 1.0422535211267605,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439519232,
+    "loss": 1.3661,
+    "grad_norm": 3.9805550575256348,
+    "learning_rate": 0.0002760038884726157
+  },
+  {
+    "step": 38,
+    "epoch": 1.0704225352112675,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439497728,
+    "loss": 1.2265,
+    "grad_norm": 3.169705390930176,
+    "learning_rate": 0.00027393581614739923
+  },
+  {
+    "step": 39,
+    "epoch": 1.0985915492957747,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439486976,
+    "loss": 1.4149,
+    "grad_norm": 7.221917629241943,
+    "learning_rate": 0.0002717907008573785
+  },
+  {
+    "step": 40,
+    "epoch": 1.1267605633802817,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439549952,
+    "loss": 1.4342,
+    "grad_norm": 6.082188129425049,
+    "learning_rate": 0.0002695698760834384
+  },
+  {
+    "step": 41,
+    "epoch": 1.1549295774647887,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439510016,
+    "loss": 1.3285,
+    "grad_norm": 3.6209311485290527,
+    "learning_rate": 0.00026727472237020447
+  },
+  {
+    "step": 42,
+    "epoch": 1.1830985915492958,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439553024,
+    "loss": 1.3293,
+    "grad_norm": 2.048436164855957,
+    "learning_rate": 0.00026490666646784665
+  },
+  {
+    "step": 43,
+    "epoch": 1.2112676056338028,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439502336,
+    "loss": 1.3793,
+    "grad_norm": 2.7470545768737793,
+    "learning_rate": 0.0002624671804451601
+  },
+  {
+    "step": 44,
+    "epoch": 1.2394366197183098,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439566848,
+    "loss": 1.3528,
+    "grad_norm": 2.686230421066284,
+    "learning_rate": 0.0002599577807744739
+  },
+  {
+    "step": 45,
+    "epoch": 1.267605633802817,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439534592,
+    "loss": 1.3625,
+    "grad_norm": 2.582090139389038,
+    "learning_rate": 0.0002573800273889577
+  },
+  {
+    "step": 46,
+    "epoch": 1.295774647887324,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.4395392,
+    "loss": 1.3958,
+    "grad_norm": 3.0206172466278076,
+    "learning_rate": 0.0002547355227129109
+  },
+  {
+    "step": 47,
+    "epoch": 1.323943661971831,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43948544,
+    "loss": 1.3788,
+    "grad_norm": 4.6008501052856445,
+    "learning_rate": 0.00025202591066563786
+  },
+  {
+    "step": 48,
+    "epoch": 1.352112676056338,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439499264,
+    "loss": 1.3214,
+    "grad_norm": 2.9762120246887207,
+    "learning_rate": 0.0002492528756395289
+  },
+  {
+    "step": 49,
+    "epoch": 1.380281690140845,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439488512,
+    "loss": 1.306,
+    "grad_norm": 2.677886724472046,
+    "learning_rate": 0.0002464181414529809
+  },
+  {
+    "step": 50,
+    "epoch": 1.408450704225352,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439502336,
+    "loss": 1.3691,
+    "grad_norm": 4.243171691894531,
+    "learning_rate": 0.00024352347027881003
+  },
+  {
+    "step": 51,
+    "epoch": 1.436619718309859,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43955456,
+    "loss": 1.4131,
+    "grad_norm": 5.571732521057129,
+    "learning_rate": 0.0002405706615488216
+  },
+  {
+    "step": 52,
+    "epoch": 1.4647887323943662,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439502336,
+    "loss": 1.375,
+    "grad_norm": 4.6382060050964355,
+    "learning_rate": 0.00023756155083521846
+  },
+  {
+    "step": 53,
+    "epoch": 1.4929577464788732,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439571456,
+    "loss": 1.3311,
+    "grad_norm": 3.3071107864379883,
+    "learning_rate": 0.00023449800870954326
+  },
+  {
+    "step": 54,
+    "epoch": 1.5211267605633803,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.4395392,
+    "loss": 1.3042,
+    "grad_norm": 3.131740093231201,
+    "learning_rate": 0.0002313819395798639
+  },
+  {
+    "step": 55,
+    "epoch": 1.5492957746478875,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439548416,
+    "loss": 1.3975,
+    "grad_norm": 4.488919258117676,
+    "learning_rate": 0.0002282152805069247
+  },
+  {
+    "step": 56,
+    "epoch": 1.5774647887323945,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43952384,
+    "loss": 1.3192,
+    "grad_norm": 3.3439676761627197,
+    "learning_rate": 0.000225
+  },
+  {
+    "step": 57,
+    "epoch": 1.6056338028169015,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439557632,
+    "loss": 1.299,
+    "grad_norm": 1.8695595264434814,
+    "learning_rate": 0.00022173809679319772
+  },
+  {
+    "step": 58,
+    "epoch": 1.6338028169014085,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.4395392,
+    "loss": 1.3204,
+    "grad_norm": 2.3153836727142334,
+    "learning_rate": 0.00021843159860297442
+  },
+  {
+    "step": 59,
+    "epoch": 1.6619718309859155,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439525376,
+    "loss": 1.3336,
+    "grad_norm": 2.244190216064453,
+    "learning_rate": 0.00021508256086763368
+  },
+  {
+    "step": 60,
+    "epoch": 1.6901408450704225,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439563776,
+    "loss": 1.3321,
+    "grad_norm": 3.2910544872283936,
+    "learning_rate": 0.00021169306546959174
+  },
+  {
+    "step": 61,
+    "epoch": 1.7183098591549295,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439496192,
+    "loss": 1.3344,
+    "grad_norm": 3.4403202533721924,
+    "learning_rate": 0.0002082652194412042
+  },
+  {
+    "step": 62,
+    "epoch": 1.7464788732394365,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439543808,
+    "loss": 1.3526,
+    "grad_norm": 1.968464732170105,
+    "learning_rate": 0.00020480115365495926
+  },
+  {
+    "step": 63,
+    "epoch": 1.7746478873239435,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43949312,
+    "loss": 1.3002,
+    "grad_norm": 2.35099196434021,
+    "learning_rate": 0.00020130302149885031
+  },
+  {
+    "step": 64,
+    "epoch": 1.8028169014084507,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439542272,
+    "loss": 1.3641,
+    "grad_norm": 3.8307111263275146,
+    "learning_rate": 0.00019777299753775265
+  },
+  {
+    "step": 65,
+    "epoch": 1.8309859154929577,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439540736,
+    "loss": 1.337,
+    "grad_norm": 2.769908905029297,
+    "learning_rate": 0.00019421327616163563
+  },
+  {
+    "step": 66,
+    "epoch": 1.8591549295774648,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439559168,
+    "loss": 1.3166,
+    "grad_norm": 3.008100748062134,
+    "learning_rate": 0.00019062607022145078
+  },
+  {
+    "step": 67,
+    "epoch": 1.887323943661972,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.4395008,
+    "loss": 1.3218,
+    "grad_norm": 2.7646050453186035,
+    "learning_rate": 0.00018701360965354402
+  },
+  {
+    "step": 68,
+    "epoch": 1.915492957746479,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439513088,
+    "loss": 1.343,
+    "grad_norm": 2.6678524017333984,
+    "learning_rate": 0.00018337814009344714
+  },
+  {
+    "step": 69,
+    "epoch": 1.943661971830986,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439537664,
+    "loss": 1.3152,
+    "grad_norm": 3.518205404281616,
+    "learning_rate": 0.0001797219214799096
+  },
+  {
+    "step": 70,
+    "epoch": 1.971830985915493,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439514624,
+    "loss": 1.2606,
+    "grad_norm": 1.7116825580596924,
+    "learning_rate": 0.00017604722665003956
+  },
+  {
+    "step": 71,
+    "epoch": 2.0,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439350272,
+    "loss": 2.0035,
+    "grad_norm": 7.806818962097168,
+    "learning_rate": 0.00017235633992642615
+  },
+  {
+    "step": 72,
+    "epoch": 2.028169014084507,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434803712,
+    "loss": 1.3873,
+    "grad_norm": 4.118950843811035,
+    "learning_rate": 0.00016865155569712278
+  },
+  {
+    "step": 73,
+    "epoch": 2.056338028169014,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434766848,
+    "loss": 1.2968,
+    "grad_norm": 2.314663887023926,
+    "learning_rate": 0.0001649351769893725
+  },
+  {
+    "step": 74,
+    "epoch": 2.084507042253521,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434826752,
+    "loss": 1.2931,
+    "grad_norm": 2.5138306617736816,
+    "learning_rate": 0.00016120951403796364
+  },
+  {
+    "step": 75,
+    "epoch": 2.112676056338028,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434794496,
+    "loss": 1.2959,
+    "grad_norm": 2.421915054321289,
+    "learning_rate": 0.00015747688284910457
+  },
+  {
+    "step": 76,
+    "epoch": 2.140845070422535,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434805248,
+    "loss": 1.2741,
+    "grad_norm": 1.9817582368850708,
+    "learning_rate": 0.00015373960376071093
+  },
+  {
+    "step": 77,
+    "epoch": 2.169014084507042,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434842112,
+    "loss": 1.3254,
+    "grad_norm": 2.0456702709198,
+    "learning_rate": 0.00015
+  },
+  {
+    "step": 78,
+    "epoch": 2.1971830985915495,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434826752,
+    "loss": 1.3843,
+    "grad_norm": 2.987546443939209,
+    "learning_rate": 0.00014626039623928907
+  },
+  {
+    "step": 79,
+    "epoch": 2.2253521126760565,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.4347776,
+    "loss": 1.2465,
+    "grad_norm": 2.78094744682312,
+    "learning_rate": 0.0001425231171508954
+  },
+  {
+    "step": 80,
+    "epoch": 2.2535211267605635,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434822144,
+    "loss": 1.3265,
+    "grad_norm": 2.5948455333709717,
+    "learning_rate": 0.00013879048596203636
+  },
+  {
+    "step": 81,
+    "epoch": 2.2816901408450705,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43480832,
+    "loss": 1.3044,
+    "grad_norm": 3.1513984203338623,
+    "learning_rate": 0.0001350648230106275
+  },
+  {
+    "step": 82,
+    "epoch": 2.3098591549295775,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434776064,
+    "loss": 1.2917,
+    "grad_norm": 3.952747344970703,
+    "learning_rate": 0.00013134844430287725
+  },
+  {
+    "step": 83,
+    "epoch": 2.3380281690140845,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434826752,
+    "loss": 1.295,
+    "grad_norm": 4.772861480712891,
+    "learning_rate": 0.0001276436600735738
+  },
+  {
+    "step": 84,
+    "epoch": 2.3661971830985915,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434765312,
+    "loss": 1.35,
+    "grad_norm": 5.37269401550293,
+    "learning_rate": 0.00012395277334996044
+  },
+  {
+    "step": 85,
+    "epoch": 2.3943661971830985,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434811392,
+    "loss": 1.3514,
+    "grad_norm": 4.7276458740234375,
+    "learning_rate": 0.00012027807852009038
+  },
+  {
+    "step": 86,
+    "epoch": 2.4225352112676055,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434765312,
+    "loss": 1.3044,
+    "grad_norm": 4.575231075286865,
+    "learning_rate": 0.00011662185990655284
+  },
+  {
+    "step": 87,
+    "epoch": 2.4507042253521125,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434796032,
+    "loss": 1.3396,
+    "grad_norm": 5.390990257263184,
+    "learning_rate": 0.00011298639034645593
+  },
+  {
+    "step": 88,
+    "epoch": 2.4788732394366195,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434771456,
+    "loss": 1.273,
+    "grad_norm": 7.428049564361572,
+    "learning_rate": 0.00010937392977854923
+  },
+  {
+    "step": 89,
+    "epoch": 2.507042253521127,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434825216,
+    "loss": 1.2715,
+    "grad_norm": 5.177944660186768,
+    "learning_rate": 0.00010578672383836435
+  },
+  {
+    "step": 90,
+    "epoch": 2.535211267605634,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434806784,
+    "loss": 1.2945,
+    "grad_norm": 4.6876220703125,
+    "learning_rate": 0.00010222700246224735
+  },
+  {
+    "step": 91,
+    "epoch": 2.563380281690141,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434756096,
+    "loss": 1.362,
+    "grad_norm": 6.015515327453613,
+    "learning_rate": 9.869697850114969e-05
+  },
+  {
+    "step": 92,
+    "epoch": 2.591549295774648,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434780672,
+    "loss": 1.2821,
+    "grad_norm": 4.289063453674316,
+    "learning_rate": 9.519884634504074e-05
+  },
+  {
+    "step": 93,
+    "epoch": 2.619718309859155,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434783744,
+    "loss": 1.2382,
+    "grad_norm": 4.802374362945557,
+    "learning_rate": 9.17347805587958e-05
+  },
+  {
+    "step": 94,
+    "epoch": 2.647887323943662,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434776064,
+    "loss": 1.224,
+    "grad_norm": 4.2559494972229,
+    "learning_rate": 8.830693453040829e-05
+  },
+  {
+    "step": 95,
+    "epoch": 2.676056338028169,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434814464,
+    "loss": 1.2576,
+    "grad_norm": 4.885746955871582,
+    "learning_rate": 8.491743913236628e-05
+  },
+  {
+    "step": 96,
+    "epoch": 2.704225352112676,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43482368,
+    "loss": 1.1952,
+    "grad_norm": 5.2525129318237305,
+    "learning_rate": 8.156840139702554e-05
+  },
+  {
+    "step": 97,
+    "epoch": 2.732394366197183,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434766848,
+    "loss": 1.313,
+    "grad_norm": 6.298973083496094,
+    "learning_rate": 7.82619032068023e-05
+  },
+  {
+    "step": 98,
+    "epoch": 2.76056338028169,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434766848,
+    "loss": 1.2644,
+    "grad_norm": 5.16410493850708,
+    "learning_rate": 7.500000000000002e-05
+  },
+  {
+    "step": 99,
+    "epoch": 2.788732394366197,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434763776,
+    "loss": 1.2443,
+    "grad_norm": 5.660140514373779,
+    "learning_rate": 7.17847194930753e-05
+  },
+  {
+    "step": 100,
+    "epoch": 2.816901408450704,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43476224,
+    "loss": 1.1987,
+    "grad_norm": 6.567827224731445,
+    "learning_rate": 6.86180604201361e-05
+  },
+  {
+    "step": 101,
+    "epoch": 2.845070422535211,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434805248,
+    "loss": 1.1556,
+    "grad_norm": 4.935121059417725,
+    "learning_rate": 6.550199129045668e-05
+  },
+  {
+    "step": 102,
+    "epoch": 2.873239436619718,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434743808,
+    "loss": 1.2275,
+    "grad_norm": 6.281479358673096,
+    "learning_rate": 6.243844916478155e-05
+  },
+  {
+    "step": 103,
+    "epoch": 2.9014084507042255,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43479296,
+    "loss": 1.2214,
+    "grad_norm": 5.953158855438232,
+    "learning_rate": 5.9429338451178355e-05
+  },
+  {
+    "step": 104,
+    "epoch": 2.9295774647887325,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434855936,
+    "loss": 1.2748,
+    "grad_norm": 9.715459823608398,
+    "learning_rate": 5.6476529721189974e-05
+  },
+  {
+    "step": 105,
+    "epoch": 2.9577464788732395,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43480832,
+    "loss": 1.2011,
+    "grad_norm": 5.6507368087768555,
+    "learning_rate": 5.358185854701909e-05
+  },
+  {
+    "step": 106,
+    "epoch": 2.9859154929577465,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.434789888,
+    "loss": 1.2578,
+    "grad_norm": 6.601775646209717,
+    "learning_rate": 5.074712436047112e-05
+  },
+  {
+    "step": 107,
+    "epoch": 3.0140845070422535,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439537664,
+    "loss": 1.7458,
+    "grad_norm": 18.245891571044922,
+    "learning_rate": 4.7974089334362057e-05
+  },
+  {
+    "step": 108,
+    "epoch": 3.0422535211267605,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439519232,
+    "loss": 1.2056,
+    "grad_norm": 6.946835994720459,
+    "learning_rate": 4.526447728708908e-05
+  },
+  {
+    "step": 109,
+    "epoch": 3.0704225352112675,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43950848,
+    "loss": 1.1129,
+    "grad_norm": 6.792351245880127,
+    "learning_rate": 4.261997261104223e-05
+  },
+  {
+    "step": 110,
+    "epoch": 3.0985915492957745,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43956224,
+    "loss": 1.1966,
+    "grad_norm": 10.979827880859375,
+    "learning_rate": 4.004221922552608e-05
+  },
+  {
+    "step": 111,
+    "epoch": 3.1267605633802815,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439522304,
+    "loss": 1.1583,
+    "grad_norm": 8.557021141052246,
+    "learning_rate": 3.753281955483985e-05
+  },
+  {
+    "step": 112,
+    "epoch": 3.1549295774647885,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439540736,
+    "loss": 1.1496,
+    "grad_norm": 7.465101718902588,
+    "learning_rate": 3.509333353215331e-05
+  },
+  {
+    "step": 113,
+    "epoch": 3.183098591549296,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439603712,
+    "loss": 1.1222,
+    "grad_norm": 7.832935333251953,
+    "learning_rate": 3.2725277629795526e-05
+  },
+  {
+    "step": 114,
+    "epoch": 3.211267605633803,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43953152,
+    "loss": 1.1831,
+    "grad_norm": 14.592466354370117,
+    "learning_rate": 3.0430123916561672e-05
+  },
+  {
+    "step": 115,
+    "epoch": 3.23943661971831,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439525376,
+    "loss": 1.2385,
+    "grad_norm": 8.725632667541504,
+    "learning_rate": 2.8209299142621522e-05
+  },
+  {
+    "step": 116,
+    "epoch": 3.267605633802817,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439540736,
+    "loss": 1.0783,
+    "grad_norm": 9.574886322021484,
+    "learning_rate": 2.6064183852600797e-05
+  },
+  {
+    "step": 117,
+    "epoch": 3.295774647887324,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439556096,
+    "loss": 1.126,
+    "grad_norm": 9.345331192016602,
+    "learning_rate": 2.3996111527384288e-05
+  },
+  {
+    "step": 118,
+    "epoch": 3.323943661971831,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43954688,
+    "loss": 1.1187,
+    "grad_norm": 10.025165557861328,
+    "learning_rate": 2.2006367755176655e-05
+  },
+  {
+    "step": 119,
+    "epoch": 3.352112676056338,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439537664,
+    "loss": 1.108,
+    "grad_norm": 8.818310737609863,
+    "learning_rate": 2.009618943233419e-05
+  },
+  {
+    "step": 120,
+    "epoch": 3.380281690140845,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439556096,
+    "loss": 1.1482,
+    "grad_norm": 9.437005043029785,
+    "learning_rate": 1.82667639944657e-05
+  },
+  {
+    "step": 121,
+    "epoch": 3.408450704225352,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43955456,
+    "loss": 1.0596,
+    "grad_norm": 9.051286697387695,
+    "learning_rate": 1.6519228678279718e-05
+  },
+  {
+    "step": 122,
+    "epoch": 3.436619718309859,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439511552,
+    "loss": 0.9866,
+    "grad_norm": 8.84517765045166,
+    "learning_rate": 1.4854669814637143e-05
+  },
+  {
+    "step": 123,
+    "epoch": 3.464788732394366,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439543808,
+    "loss": 1.1725,
+    "grad_norm": 13.230541229248047,
+    "learning_rate": 1.3274122153249028e-05
+  },
+  {
+    "step": 124,
+    "epoch": 3.492957746478873,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439497728,
+    "loss": 1.1315,
+    "grad_norm": 9.880842208862305,
+    "learning_rate": 1.1778568219438839e-05
+  },
+  {
+    "step": 125,
+    "epoch": 3.52112676056338,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439542272,
+    "loss": 1.0279,
+    "grad_norm": 12.509747505187988,
+    "learning_rate": 1.036893770336938e-05
+  },
+  {
+    "step": 126,
+    "epoch": 3.5492957746478875,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43949312,
+    "loss": 1.0526,
+    "grad_norm": 14.200373649597168,
+    "learning_rate": 9.046106882113751e-06
+  },
+  {
+    "step": 127,
+    "epoch": 3.5774647887323945,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439505408,
+    "loss": 1.0684,
+    "grad_norm": 12.265142440795898,
+    "learning_rate": 7.810898074930243e-06
+  },
+  {
+    "step": 128,
+    "epoch": 3.6056338028169015,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439529984,
+    "loss": 1.0435,
+    "grad_norm": 9.855721473693848,
+    "learning_rate": 6.664079132078881e-06
+  },
+  {
+    "step": 129,
+    "epoch": 3.6338028169014085,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439491584,
+    "loss": 1.0585,
+    "grad_norm": 9.79635238647461,
+    "learning_rate": 5.606362957498195e-06
+  },
+  {
+    "step": 130,
+    "epoch": 3.6619718309859155,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439494656,
+    "loss": 1.1007,
+    "grad_norm": 12.417722702026367,
+    "learning_rate": 4.638407065638322e-06
+  },
+  {
+    "step": 131,
+    "epoch": 3.6901408450704225,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439506944,
+    "loss": 1.0559,
+    "grad_norm": 9.638754844665527,
+    "learning_rate": 3.760813172726457e-06
+  },
+  {
+    "step": 132,
+    "epoch": 3.7183098591549295,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439471616,
+    "loss": 1.1001,
+    "grad_norm": 12.243162155151367,
+    "learning_rate": 2.9741268227184255e-06
+  },
+  {
+    "step": 133,
+    "epoch": 3.7464788732394365,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439513088,
+    "loss": 1.0142,
+    "grad_norm": 8.843420028686523,
+    "learning_rate": 2.2788370481687965e-06
+  },
+  {
+    "step": 134,
+    "epoch": 3.7746478873239435,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439528448,
+    "loss": 1.1127,
+    "grad_norm": 17.717222213745117,
+    "learning_rate": 1.6753760662307215e-06
+  },
+  {
+    "step": 135,
+    "epoch": 3.802816901408451,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.43949312,
+    "loss": 1.1962,
+    "grad_norm": 12.415128707885742,
+    "learning_rate": 1.1641190099741904e-06
+  },
+  {
+    "step": 136,
+    "epoch": 3.830985915492958,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.4395008,
+    "loss": 1.1189,
+    "grad_norm": 13.41920280456543,
+    "learning_rate": 7.453836951897885e-07
+  },
+  {
+    "step": 137,
+    "epoch": 3.859154929577465,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439522304,
+    "loss": 1.0789,
+    "grad_norm": 12.066742897033691,
+    "learning_rate": 4.194304228229806e-07
+  },
+  {
+    "step": 138,
+    "epoch": 3.887323943661972,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439533056,
+    "loss": 1.1993,
+    "grad_norm": 11.856720924377441,
+    "learning_rate": 1.8646181716164831e-07
+  },
+  {
+    "step": 139,
+    "epoch": 3.915492957746479,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439525376,
+    "loss": 1.1628,
+    "grad_norm": 14.922451972961426,
+    "learning_rate": 4.662269987756317e-08
+  },
+  {
+    "step": 140,
+    "epoch": 3.943661971830986,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439559168,
+    "loss": 1.0973,
+    "grad_norm": 9.704615592956543,
+    "learning_rate": 0.0
+  },
+  {
+    "step": 140,
+    "epoch": 3.943661971830986,
+    "cpu_mem": 1.50761472,
+    "gpu_mem": 4.439559168,
+    "train_runtime": 683.2424,
+    "train_samples_per_second": 13.178,
+    "train_steps_per_second": 0.205,
+    "total_flos": 0.0,
+    "train_loss": 1.3638307809829713
+  }
+]

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_e-r32-a2/adapter_config.json

@@ -0,0 +1,30 @@
+{
+  "alpha": 64,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "init_weight": "kaiming",
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "metric_tracking": false,
+  "modules_to_save": null,
+  "peft_type": "ABLATION",
+  "r": 32,
+  "revision": null,
+  "seed": 42,
+  "share_weights": false,
+  "target_modules": [
+    "up_proj",
+    "gate_proj",
+    "v_proj",
+    "k_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj"
+  ],
+  "task_type": null,
+  "track_n": 100,
+  "variant": "B"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_e-r32-a2/eval_results.json

@@ -0,0 +1,4 @@
+{
+    "task": "arc_e",
+    "results": 0.32196969696969696
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_e-r32-a2/training_configuration.json

@@ -0,0 +1,38 @@
+{
+  "model_id": "TinyLlama/TinyLlama_v1.1",
+  "dataset": {
+    "name": "ARC_E",
+    "dataset_id": "allenai/ai2_arc",
+    "preprocess_id": "arc_train_deepeval"
+  },
+  "peft_config": {
+    "method": "abl_B",
+    "rank": 32,
+    "alpha": 64,
+    "dropout": 0.0,
+    "bias": "none",
+    "target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "down_proj",
+      "up_proj"
+    ],
+    "trainable_parameter_count": 13009920
+  },
+  "training_config": {
+    "max_dataset_length": null,
+    "batch_size": 64,
+    "per_device_batch_size": 32,
+    "gradient_accumulation_steps": 2,
+    "learning_rate": 0.0003,
+    "num_epochs": 4,
+    "warmup_ratio": 0.1
+  },
+  "model_name": "TinyLlama_v1.1-abl_B-arc_e-r32-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_e-r32-a2",
+  "seed": 42,
+  "timestamp": "2025-09-01T02:59:40.533088"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_e-r32-a2/training_logs.json

@@ -0,0 +1,1273 @@
+[
+  {
+    "step": 1,
+    "epoch": 0.028169014084507043,
+    "cpu_mem": 1.506336768,
+    "gpu_mem": 4.519910912,
+    "loss": 4.6319,
+    "grad_norm": 15.45500373840332,
+    "learning_rate": 2.1428571428571425e-05
+  },
+  {
+    "step": 2,
+    "epoch": 0.056338028169014086,
+    "cpu_mem": 1.506533376,
+    "gpu_mem": 4.6240512,
+    "loss": 4.4578,
+    "grad_norm": 15.729421615600586,
+    "learning_rate": 4.285714285714285e-05
+  },
+  {
+    "step": 3,
+    "epoch": 0.08450704225352113,
+    "cpu_mem": 1.506729984,
+    "gpu_mem": 4.624029696,
+    "loss": 4.2421,
+    "grad_norm": 15.735064506530762,
+    "learning_rate": 6.428571428571427e-05
+  },
+  {
+    "step": 4,
+    "epoch": 0.11267605633802817,
+    "cpu_mem": 1.506926592,
+    "gpu_mem": 4.624008192,
+    "loss": 3.7264,
+    "grad_norm": 12.702402114868164,
+    "learning_rate": 8.57142857142857e-05
+  },
+  {
+    "step": 5,
+    "epoch": 0.14084507042253522,
+    "cpu_mem": 1.506926592,
+    "gpu_mem": 4.624049664,
+    "loss": 2.8628,
+    "grad_norm": 16.878061294555664,
+    "learning_rate": 0.00010714285714285714
+  },
+  {
+    "step": 6,
+    "epoch": 0.16901408450704225,
+    "cpu_mem": 1.5071232,
+    "gpu_mem": 4.624025088,
+    "loss": 2.1096,
+    "grad_norm": 6.79240083694458,
+    "learning_rate": 0.00012857142857142855
+  },
+  {
+    "step": 7,
+    "epoch": 0.19718309859154928,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624048128,
+    "loss": 1.9211,
+    "grad_norm": 2.7695512771606445,
+    "learning_rate": 0.00015
+  },
+  {
+    "step": 8,
+    "epoch": 0.22535211267605634,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624006656,
+    "loss": 1.6771,
+    "grad_norm": 1.5878878831863403,
+    "learning_rate": 0.0001714285714285714
+  },
+  {
+    "step": 9,
+    "epoch": 0.2535211267605634,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624008192,
+    "loss": 1.552,
+    "grad_norm": 1.507412075996399,
+    "learning_rate": 0.00019285714285714286
+  },
+  {
+    "step": 10,
+    "epoch": 0.28169014084507044,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624003584,
+    "loss": 1.5163,
+    "grad_norm": 0.9610570073127747,
+    "learning_rate": 0.00021428571428571427
+  },
+  {
+    "step": 11,
+    "epoch": 0.30985915492957744,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.62408192,
+    "loss": 1.4057,
+    "grad_norm": 0.9633939862251282,
+    "learning_rate": 0.00023571428571428569
+  },
+  {
+    "step": 12,
+    "epoch": 0.3380281690140845,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624055808,
+    "loss": 1.3628,
+    "grad_norm": 0.8077499270439148,
+    "learning_rate": 0.0002571428571428571
+  },
+  {
+    "step": 13,
+    "epoch": 0.36619718309859156,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624006656,
+    "loss": 1.3515,
+    "grad_norm": 0.8354933261871338,
+    "learning_rate": 0.00027857142857142854
+  },
+  {
+    "step": 14,
+    "epoch": 0.39436619718309857,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.62402816,
+    "loss": 1.3241,
+    "grad_norm": 0.6401271820068359,
+    "learning_rate": 0.0003
+  },
+  {
+    "step": 15,
+    "epoch": 0.4225352112676056,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.62400512,
+    "loss": 1.3536,
+    "grad_norm": 0.915156364440918,
+    "learning_rate": 0.0002999533773001224
+  },
+  {
+    "step": 16,
+    "epoch": 0.4507042253521127,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624009728,
+    "loss": 1.386,
+    "grad_norm": 1.0561822652816772,
+    "learning_rate": 0.0002998135381828383
+  },
+  {
+    "step": 17,
+    "epoch": 0.4788732394366197,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624046592,
+    "loss": 1.3383,
+    "grad_norm": 0.8924583196640015,
+    "learning_rate": 0.00029958056957717696
+  },
+  {
+    "step": 18,
+    "epoch": 0.5070422535211268,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624057344,
+    "loss": 1.3418,
+    "grad_norm": 1.1358929872512817,
+    "learning_rate": 0.0002992546163048102
+  },
+  {
+    "step": 19,
+    "epoch": 0.5352112676056338,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624000512,
+    "loss": 1.3633,
+    "grad_norm": 1.5113840103149414,
+    "learning_rate": 0.0002988358809900258
+  },
+  {
+    "step": 20,
+    "epoch": 0.5633802816901409,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624071168,
+    "loss": 1.3698,
+    "grad_norm": 0.9107031226158142,
+    "learning_rate": 0.0002983246239337692
+  },
+  {
+    "step": 21,
+    "epoch": 0.5915492957746479,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624069632,
+    "loss": 1.3323,
+    "grad_norm": 0.9094489216804504,
+    "learning_rate": 0.0002977211629518312
+  },
+  {
+    "step": 22,
+    "epoch": 0.6197183098591549,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624026624,
+    "loss": 1.3659,
+    "grad_norm": 1.0449373722076416,
+    "learning_rate": 0.00029702587317728153
+  },
+  {
+    "step": 23,
+    "epoch": 0.647887323943662,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.62404352,
+    "loss": 1.3191,
+    "grad_norm": 0.6375622153282166,
+    "learning_rate": 0.0002962391868272735
+  },
+  {
+    "step": 24,
+    "epoch": 0.676056338028169,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624000512,
+    "loss": 1.3682,
+    "grad_norm": 0.6815070509910583,
+    "learning_rate": 0.00029536159293436166
+  },
+  {
+    "step": 25,
+    "epoch": 0.704225352112676,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624029696,
+    "loss": 1.3646,
+    "grad_norm": 0.43889474868774414,
+    "learning_rate": 0.00029439363704250176
+  },
+  {
+    "step": 26,
+    "epoch": 0.7323943661971831,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624009728,
+    "loss": 1.4326,
+    "grad_norm": 0.6083115935325623,
+    "learning_rate": 0.00029333592086792107
+  },
+  {
+    "step": 27,
+    "epoch": 0.7605633802816901,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.62403584,
+    "loss": 1.3713,
+    "grad_norm": 0.5347280502319336,
+    "learning_rate": 0.0002921891019250697
+  },
+  {
+    "step": 28,
+    "epoch": 0.7887323943661971,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.62403584,
+    "loss": 1.342,
+    "grad_norm": 0.31040897965431213,
+    "learning_rate": 0.0002909538931178862
+  },
+  {
+    "step": 29,
+    "epoch": 0.8169014084507042,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624014336,
+    "loss": 1.2572,
+    "grad_norm": 0.41217169165611267,
+    "learning_rate": 0.00028963106229663063
+  },
+  {
+    "step": 30,
+    "epoch": 0.8450704225352113,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.62400512,
+    "loss": 1.3938,
+    "grad_norm": 0.6330100893974304,
+    "learning_rate": 0.00028822143178056114
+  },
+  {
+    "step": 31,
+    "epoch": 0.8732394366197183,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624023552,
+    "loss": 1.3735,
+    "grad_norm": 0.5508550405502319,
+    "learning_rate": 0.00028672587784675096
+  },
+  {
+    "step": 32,
+    "epoch": 0.9014084507042254,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624046592,
+    "loss": 1.3095,
+    "grad_norm": 0.24852244555950165,
+    "learning_rate": 0.0002851453301853628
+  },
+  {
+    "step": 33,
+    "epoch": 0.9295774647887324,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.62404352,
+    "loss": 1.3292,
+    "grad_norm": 0.25524747371673584,
+    "learning_rate": 0.00028348077132172027
+  },
+  {
+    "step": 34,
+    "epoch": 0.9577464788732394,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624046592,
+    "loss": 1.4214,
+    "grad_norm": 0.6475468873977661,
+    "learning_rate": 0.0002817332360055343
+  },
+  {
+    "step": 35,
+    "epoch": 0.9859154929577465,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.62402816,
+    "loss": 1.3162,
+    "grad_norm": 0.3794228434562683,
+    "learning_rate": 0.0002799038105676658
+  },
+  {
+    "step": 36,
+    "epoch": 1.0140845070422535,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676058624,
+    "loss": 2.0002,
+    "grad_norm": 0.7675129175186157,
+    "learning_rate": 0.0002779936322448233
+  },
+  {
+    "step": 37,
+    "epoch": 1.0422535211267605,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676063232,
+    "loss": 1.3351,
+    "grad_norm": 0.1956872045993805,
+    "learning_rate": 0.0002760038884726157
+  },
+  {
+    "step": 38,
+    "epoch": 1.0704225352112675,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676041728,
+    "loss": 1.245,
+    "grad_norm": 0.3838285803794861,
+    "learning_rate": 0.00027393581614739923
+  },
+  {
+    "step": 39,
+    "epoch": 1.0985915492957747,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676030976,
+    "loss": 1.3188,
+    "grad_norm": 0.49647337198257446,
+    "learning_rate": 0.0002717907008573785
+  },
+  {
+    "step": 40,
+    "epoch": 1.1267605633802817,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676093952,
+    "loss": 1.3697,
+    "grad_norm": 0.4313946068286896,
+    "learning_rate": 0.0002695698760834384
+  },
+  {
+    "step": 41,
+    "epoch": 1.1549295774647887,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676054016,
+    "loss": 1.3231,
+    "grad_norm": 0.2102261781692505,
+    "learning_rate": 0.00026727472237020447
+  },
+  {
+    "step": 42,
+    "epoch": 1.1830985915492958,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676097024,
+    "loss": 1.3535,
+    "grad_norm": 0.3038178086280823,
+    "learning_rate": 0.00026490666646784665
+  },
+  {
+    "step": 43,
+    "epoch": 1.2112676056338028,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676046336,
+    "loss": 1.3822,
+    "grad_norm": 0.22646814584732056,
+    "learning_rate": 0.0002624671804451601
+  },
+  {
+    "step": 44,
+    "epoch": 1.2394366197183098,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676110848,
+    "loss": 1.3555,
+    "grad_norm": 0.2905891537666321,
+    "learning_rate": 0.0002599577807744739
+  },
+  {
+    "step": 45,
+    "epoch": 1.267605633802817,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676078592,
+    "loss": 1.3617,
+    "grad_norm": 0.22614973783493042,
+    "learning_rate": 0.0002573800273889577
+  },
+  {
+    "step": 46,
+    "epoch": 1.295774647887324,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.6760832,
+    "loss": 1.3523,
+    "grad_norm": 0.19625195860862732,
+    "learning_rate": 0.0002547355227129109
+  },
+  {
+    "step": 47,
+    "epoch": 1.323943661971831,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.67602944,
+    "loss": 1.3234,
+    "grad_norm": 0.5174495577812195,
+    "learning_rate": 0.00025202591066563786
+  },
+  {
+    "step": 48,
+    "epoch": 1.352112676056338,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676043264,
+    "loss": 1.3128,
+    "grad_norm": 0.23632147908210754,
+    "learning_rate": 0.0002492528756395289
+  },
+  {
+    "step": 49,
+    "epoch": 1.380281690140845,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676032512,
+    "loss": 1.3276,
+    "grad_norm": 0.3750535249710083,
+    "learning_rate": 0.0002464181414529809
+  },
+  {
+    "step": 50,
+    "epoch": 1.408450704225352,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676046336,
+    "loss": 1.3333,
+    "grad_norm": 0.326159805059433,
+    "learning_rate": 0.00024352347027881003
+  },
+  {
+    "step": 51,
+    "epoch": 1.436619718309859,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.67609856,
+    "loss": 1.3605,
+    "grad_norm": 0.43731603026390076,
+    "learning_rate": 0.0002405706615488216
+  },
+  {
+    "step": 52,
+    "epoch": 1.4647887323943662,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676046336,
+    "loss": 1.3531,
+    "grad_norm": 0.3836096525192261,
+    "learning_rate": 0.00023756155083521846
+  },
+  {
+    "step": 53,
+    "epoch": 1.4929577464788732,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676115456,
+    "loss": 1.3079,
+    "grad_norm": 0.3928913474082947,
+    "learning_rate": 0.00023449800870954326
+  },
+  {
+    "step": 54,
+    "epoch": 1.5211267605633803,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.6760832,
+    "loss": 1.3158,
+    "grad_norm": 0.4037013351917267,
+    "learning_rate": 0.0002313819395798639
+  },
+  {
+    "step": 55,
+    "epoch": 1.5492957746478875,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676092416,
+    "loss": 1.3535,
+    "grad_norm": 0.22461530566215515,
+    "learning_rate": 0.0002282152805069247
+  },
+  {
+    "step": 56,
+    "epoch": 1.5774647887323945,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.67606784,
+    "loss": 1.3064,
+    "grad_norm": 0.1745154857635498,
+    "learning_rate": 0.000225
+  },
+  {
+    "step": 57,
+    "epoch": 1.6056338028169015,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676101632,
+    "loss": 1.311,
+    "grad_norm": 0.1602332592010498,
+    "learning_rate": 0.00022173809679319772
+  },
+  {
+    "step": 58,
+    "epoch": 1.6338028169014085,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.6760832,
+    "loss": 1.3149,
+    "grad_norm": 0.2614760994911194,
+    "learning_rate": 0.00021843159860297442
+  },
+  {
+    "step": 59,
+    "epoch": 1.6619718309859155,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676069376,
+    "loss": 1.3111,
+    "grad_norm": 0.15861870348453522,
+    "learning_rate": 0.00021508256086763368
+  },
+  {
+    "step": 60,
+    "epoch": 1.6901408450704225,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676107776,
+    "loss": 1.3024,
+    "grad_norm": 0.4162784814834595,
+    "learning_rate": 0.00021169306546959174
+  },
+  {
+    "step": 61,
+    "epoch": 1.7183098591549295,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676040192,
+    "loss": 1.3036,
+    "grad_norm": 0.376035213470459,
+    "learning_rate": 0.0002082652194412042
+  },
+  {
+    "step": 62,
+    "epoch": 1.7464788732394365,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676087808,
+    "loss": 1.3625,
+    "grad_norm": 0.26167038083076477,
+    "learning_rate": 0.00020480115365495926
+  },
+  {
+    "step": 63,
+    "epoch": 1.7746478873239435,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.67603712,
+    "loss": 1.3322,
+    "grad_norm": 0.36649367213249207,
+    "learning_rate": 0.00020130302149885031
+  },
+  {
+    "step": 64,
+    "epoch": 1.8028169014084507,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676086272,
+    "loss": 1.3437,
+    "grad_norm": 0.45341187715530396,
+    "learning_rate": 0.00019777299753775265
+  },
+  {
+    "step": 65,
+    "epoch": 1.8309859154929577,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676084736,
+    "loss": 1.3348,
+    "grad_norm": 0.34720247983932495,
+    "learning_rate": 0.00019421327616163563
+  },
+  {
+    "step": 66,
+    "epoch": 1.8591549295774648,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676103168,
+    "loss": 1.2949,
+    "grad_norm": 0.29440757632255554,
+    "learning_rate": 0.00019062607022145078
+  },
+  {
+    "step": 67,
+    "epoch": 1.887323943661972,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.6760448,
+    "loss": 1.3086,
+    "grad_norm": 0.22650545835494995,
+    "learning_rate": 0.00018701360965354402
+  },
+  {
+    "step": 68,
+    "epoch": 1.915492957746479,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676057088,
+    "loss": 1.3501,
+    "grad_norm": 0.18094563484191895,
+    "learning_rate": 0.00018337814009344714
+  },
+  {
+    "step": 69,
+    "epoch": 1.943661971830986,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676081664,
+    "loss": 1.3019,
+    "grad_norm": 0.4657556116580963,
+    "learning_rate": 0.0001797219214799096
+  },
+  {
+    "step": 70,
+    "epoch": 1.971830985915493,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676058624,
+    "loss": 1.2815,
+    "grad_norm": 0.32910600304603577,
+    "learning_rate": 0.00017604722665003956
+  },
+  {
+    "step": 71,
+    "epoch": 2.0,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.675894272,
+    "loss": 1.9533,
+    "grad_norm": 0.9542618989944458,
+    "learning_rate": 0.00017235633992642615
+  },
+  {
+    "step": 72,
+    "epoch": 2.028169014084507,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624038912,
+    "loss": 1.3393,
+    "grad_norm": 0.26049795746803284,
+    "learning_rate": 0.00016865155569712278
+  },
+  {
+    "step": 73,
+    "epoch": 2.056338028169014,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624002048,
+    "loss": 1.3217,
+    "grad_norm": 0.27294543385505676,
+    "learning_rate": 0.0001649351769893725
+  },
+  {
+    "step": 74,
+    "epoch": 2.084507042253521,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624061952,
+    "loss": 1.3088,
+    "grad_norm": 0.2529256045818329,
+    "learning_rate": 0.00016120951403796364
+  },
+  {
+    "step": 75,
+    "epoch": 2.112676056338028,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624029696,
+    "loss": 1.3425,
+    "grad_norm": 0.39007511734962463,
+    "learning_rate": 0.00015747688284910457
+  },
+  {
+    "step": 76,
+    "epoch": 2.140845070422535,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624040448,
+    "loss": 1.2894,
+    "grad_norm": 0.2318618893623352,
+    "learning_rate": 0.00015373960376071093
+  },
+  {
+    "step": 77,
+    "epoch": 2.169014084507042,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624077312,
+    "loss": 1.3408,
+    "grad_norm": 0.17781618237495422,
+    "learning_rate": 0.00015
+  },
+  {
+    "step": 78,
+    "epoch": 2.1971830985915495,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624061952,
+    "loss": 1.3582,
+    "grad_norm": 0.3542206287384033,
+    "learning_rate": 0.00014626039623928907
+  },
+  {
+    "step": 79,
+    "epoch": 2.2253521126760565,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.6240128,
+    "loss": 1.2668,
+    "grad_norm": 0.3510591387748718,
+    "learning_rate": 0.0001425231171508954
+  },
+  {
+    "step": 80,
+    "epoch": 2.2535211267605635,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624057344,
+    "loss": 1.3527,
+    "grad_norm": 0.4764325022697449,
+    "learning_rate": 0.00013879048596203636
+  },
+  {
+    "step": 81,
+    "epoch": 2.2816901408450705,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.62404352,
+    "loss": 1.3285,
+    "grad_norm": 0.29078051447868347,
+    "learning_rate": 0.0001350648230106275
+  },
+  {
+    "step": 82,
+    "epoch": 2.3098591549295775,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624011264,
+    "loss": 1.3261,
+    "grad_norm": 0.329904705286026,
+    "learning_rate": 0.00013134844430287725
+  },
+  {
+    "step": 83,
+    "epoch": 2.3380281690140845,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624061952,
+    "loss": 1.3449,
+    "grad_norm": 0.671593427658081,
+    "learning_rate": 0.0001276436600735738
+  },
+  {
+    "step": 84,
+    "epoch": 2.3661971830985915,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624000512,
+    "loss": 1.3563,
+    "grad_norm": 0.3557969629764557,
+    "learning_rate": 0.00012395277334996044
+  },
+  {
+    "step": 85,
+    "epoch": 2.3943661971830985,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624046592,
+    "loss": 1.3714,
+    "grad_norm": 0.5320531129837036,
+    "learning_rate": 0.00012027807852009038
+  },
+  {
+    "step": 86,
+    "epoch": 2.4225352112676055,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624000512,
+    "loss": 1.337,
+    "grad_norm": 0.38929489254951477,
+    "learning_rate": 0.00011662185990655284
+  },
+  {
+    "step": 87,
+    "epoch": 2.4507042253521125,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624031232,
+    "loss": 1.3616,
+    "grad_norm": 0.3265232741832733,
+    "learning_rate": 0.00011298639034645593
+  },
+  {
+    "step": 88,
+    "epoch": 2.4788732394366195,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624006656,
+    "loss": 1.3341,
+    "grad_norm": 0.5538458824157715,
+    "learning_rate": 0.00010937392977854923
+  },
+  {
+    "step": 89,
+    "epoch": 2.507042253521127,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624060416,
+    "loss": 1.2784,
+    "grad_norm": 0.30906733870506287,
+    "learning_rate": 0.00010578672383836435
+  },
+  {
+    "step": 90,
+    "epoch": 2.535211267605634,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624041984,
+    "loss": 1.3251,
+    "grad_norm": 0.30996736884117126,
+    "learning_rate": 0.00010222700246224735
+  },
+  {
+    "step": 91,
+    "epoch": 2.563380281690141,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.623991296,
+    "loss": 1.4019,
+    "grad_norm": 0.8204323053359985,
+    "learning_rate": 9.869697850114969e-05
+  },
+  {
+    "step": 92,
+    "epoch": 2.591549295774648,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624015872,
+    "loss": 1.3102,
+    "grad_norm": 0.2713940143585205,
+    "learning_rate": 9.519884634504074e-05
+  },
+  {
+    "step": 93,
+    "epoch": 2.619718309859155,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624018944,
+    "loss": 1.2545,
+    "grad_norm": 0.43170469999313354,
+    "learning_rate": 9.17347805587958e-05
+  },
+  {
+    "step": 94,
+    "epoch": 2.647887323943662,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624011264,
+    "loss": 1.279,
+    "grad_norm": 0.260023832321167,
+    "learning_rate": 8.830693453040829e-05
+  },
+  {
+    "step": 95,
+    "epoch": 2.676056338028169,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624049664,
+    "loss": 1.3035,
+    "grad_norm": 0.2027701735496521,
+    "learning_rate": 8.491743913236628e-05
+  },
+  {
+    "step": 96,
+    "epoch": 2.704225352112676,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.62405888,
+    "loss": 1.2398,
+    "grad_norm": 0.5005522966384888,
+    "learning_rate": 8.156840139702554e-05
+  },
+  {
+    "step": 97,
+    "epoch": 2.732394366197183,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624002048,
+    "loss": 1.3283,
+    "grad_norm": 0.4746371805667877,
+    "learning_rate": 7.82619032068023e-05
+  },
+  {
+    "step": 98,
+    "epoch": 2.76056338028169,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624002048,
+    "loss": 1.3273,
+    "grad_norm": 0.1946246474981308,
+    "learning_rate": 7.500000000000002e-05
+  },
+  {
+    "step": 99,
+    "epoch": 2.788732394366197,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.623998976,
+    "loss": 1.2802,
+    "grad_norm": 0.2766704559326172,
+    "learning_rate": 7.17847194930753e-05
+  },
+  {
+    "step": 100,
+    "epoch": 2.816901408450704,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.62399744,
+    "loss": 1.2638,
+    "grad_norm": 0.29240837693214417,
+    "learning_rate": 6.86180604201361e-05
+  },
+  {
+    "step": 101,
+    "epoch": 2.845070422535211,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624040448,
+    "loss": 1.2422,
+    "grad_norm": 0.3155769407749176,
+    "learning_rate": 6.550199129045668e-05
+  },
+  {
+    "step": 102,
+    "epoch": 2.873239436619718,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.623979008,
+    "loss": 1.2926,
+    "grad_norm": 0.27102938294410706,
+    "learning_rate": 6.243844916478155e-05
+  },
+  {
+    "step": 103,
+    "epoch": 2.9014084507042255,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.62402816,
+    "loss": 1.2868,
+    "grad_norm": 0.2362990826368332,
+    "learning_rate": 5.9429338451178355e-05
+  },
+  {
+    "step": 104,
+    "epoch": 2.9295774647887325,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624091136,
+    "loss": 1.323,
+    "grad_norm": 0.4770653247833252,
+    "learning_rate": 5.6476529721189974e-05
+  },
+  {
+    "step": 105,
+    "epoch": 2.9577464788732395,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.62404352,
+    "loss": 1.2533,
+    "grad_norm": 0.3264782726764679,
+    "learning_rate": 5.358185854701909e-05
+  },
+  {
+    "step": 106,
+    "epoch": 2.9859154929577465,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.624025088,
+    "loss": 1.3119,
+    "grad_norm": 0.20415638387203217,
+    "learning_rate": 5.074712436047112e-05
+  },
+  {
+    "step": 107,
+    "epoch": 3.0140845070422535,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676081664,
+    "loss": 1.8464,
+    "grad_norm": 0.7195807695388794,
+    "learning_rate": 4.7974089334362057e-05
+  },
+  {
+    "step": 108,
+    "epoch": 3.0422535211267605,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676063232,
+    "loss": 1.2953,
+    "grad_norm": 0.45035746693611145,
+    "learning_rate": 4.526447728708908e-05
+  },
+  {
+    "step": 109,
+    "epoch": 3.0704225352112675,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.67605248,
+    "loss": 1.2954,
+    "grad_norm": 0.2725832462310791,
+    "learning_rate": 4.261997261104223e-05
+  },
+  {
+    "step": 110,
+    "epoch": 3.0985915492957745,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.67610624,
+    "loss": 1.3324,
+    "grad_norm": 0.6822441816329956,
+    "learning_rate": 4.004221922552608e-05
+  },
+  {
+    "step": 111,
+    "epoch": 3.1267605633802815,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676066304,
+    "loss": 1.3058,
+    "grad_norm": 0.3432196378707886,
+    "learning_rate": 3.753281955483985e-05
+  },
+  {
+    "step": 112,
+    "epoch": 3.1549295774647885,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676084736,
+    "loss": 1.2886,
+    "grad_norm": 0.21611112356185913,
+    "learning_rate": 3.509333353215331e-05
+  },
+  {
+    "step": 113,
+    "epoch": 3.183098591549296,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676147712,
+    "loss": 1.2902,
+    "grad_norm": 0.24663004279136658,
+    "learning_rate": 3.2725277629795526e-05
+  },
+  {
+    "step": 114,
+    "epoch": 3.211267605633803,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.67607552,
+    "loss": 1.3149,
+    "grad_norm": 0.3060291111469269,
+    "learning_rate": 3.0430123916561672e-05
+  },
+  {
+    "step": 115,
+    "epoch": 3.23943661971831,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676069376,
+    "loss": 1.3368,
+    "grad_norm": 0.4497460722923279,
+    "learning_rate": 2.8209299142621522e-05
+  },
+  {
+    "step": 116,
+    "epoch": 3.267605633802817,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676084736,
+    "loss": 1.3084,
+    "grad_norm": 0.27042055130004883,
+    "learning_rate": 2.6064183852600797e-05
+  },
+  {
+    "step": 117,
+    "epoch": 3.295774647887324,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676100096,
+    "loss": 1.2573,
+    "grad_norm": 0.2981705367565155,
+    "learning_rate": 2.3996111527384288e-05
+  },
+  {
+    "step": 118,
+    "epoch": 3.323943661971831,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.67609088,
+    "loss": 1.2871,
+    "grad_norm": 0.305347740650177,
+    "learning_rate": 2.2006367755176655e-05
+  },
+  {
+    "step": 119,
+    "epoch": 3.352112676056338,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676081664,
+    "loss": 1.3115,
+    "grad_norm": 0.23850424587726593,
+    "learning_rate": 2.009618943233419e-05
+  },
+  {
+    "step": 120,
+    "epoch": 3.380281690140845,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676100096,
+    "loss": 1.2868,
+    "grad_norm": 0.288796603679657,
+    "learning_rate": 1.82667639944657e-05
+  },
+  {
+    "step": 121,
+    "epoch": 3.408450704225352,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.67609856,
+    "loss": 1.2793,
+    "grad_norm": 0.31622397899627686,
+    "learning_rate": 1.6519228678279718e-05
+  },
+  {
+    "step": 122,
+    "epoch": 3.436619718309859,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676055552,
+    "loss": 1.279,
+    "grad_norm": 0.3226993680000305,
+    "learning_rate": 1.4854669814637143e-05
+  },
+  {
+    "step": 123,
+    "epoch": 3.464788732394366,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676087808,
+    "loss": 1.2784,
+    "grad_norm": 0.3610694706439972,
+    "learning_rate": 1.3274122153249028e-05
+  },
+  {
+    "step": 124,
+    "epoch": 3.492957746478873,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676041728,
+    "loss": 1.3062,
+    "grad_norm": 0.27508386969566345,
+    "learning_rate": 1.1778568219438839e-05
+  },
+  {
+    "step": 125,
+    "epoch": 3.52112676056338,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676086272,
+    "loss": 1.2735,
+    "grad_norm": 0.5249335169792175,
+    "learning_rate": 1.036893770336938e-05
+  },
+  {
+    "step": 126,
+    "epoch": 3.5492957746478875,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.67603712,
+    "loss": 1.267,
+    "grad_norm": 0.382548063993454,
+    "learning_rate": 9.046106882113751e-06
+  },
+  {
+    "step": 127,
+    "epoch": 3.5774647887323945,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676049408,
+    "loss": 1.2886,
+    "grad_norm": 0.2308499962091446,
+    "learning_rate": 7.810898074930243e-06
+  },
+  {
+    "step": 128,
+    "epoch": 3.6056338028169015,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676073984,
+    "loss": 1.2877,
+    "grad_norm": 0.2855219841003418,
+    "learning_rate": 6.664079132078881e-06
+  },
+  {
+    "step": 129,
+    "epoch": 3.6338028169014085,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676035584,
+    "loss": 1.2868,
+    "grad_norm": 0.21983076632022858,
+    "learning_rate": 5.606362957498195e-06
+  },
+  {
+    "step": 130,
+    "epoch": 3.6619718309859155,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676038656,
+    "loss": 1.3086,
+    "grad_norm": 0.26207077503204346,
+    "learning_rate": 4.638407065638322e-06
+  },
+  {
+    "step": 131,
+    "epoch": 3.6901408450704225,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676050944,
+    "loss": 1.2698,
+    "grad_norm": 0.31314849853515625,
+    "learning_rate": 3.760813172726457e-06
+  },
+  {
+    "step": 132,
+    "epoch": 3.7183098591549295,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676015616,
+    "loss": 1.2776,
+    "grad_norm": 0.25790277123451233,
+    "learning_rate": 2.9741268227184255e-06
+  },
+  {
+    "step": 133,
+    "epoch": 3.7464788732394365,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676057088,
+    "loss": 1.2492,
+    "grad_norm": 0.2964359223842621,
+    "learning_rate": 2.2788370481687965e-06
+  },
+  {
+    "step": 134,
+    "epoch": 3.7746478873239435,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676072448,
+    "loss": 1.251,
+    "grad_norm": 0.3126276433467865,
+    "learning_rate": 1.6753760662307215e-06
+  },
+  {
+    "step": 135,
+    "epoch": 3.802816901408451,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.67603712,
+    "loss": 1.3024,
+    "grad_norm": 0.4269898235797882,
+    "learning_rate": 1.1641190099741904e-06
+  },
+  {
+    "step": 136,
+    "epoch": 3.830985915492958,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.6760448,
+    "loss": 1.3213,
+    "grad_norm": 0.368845134973526,
+    "learning_rate": 7.453836951897885e-07
+  },
+  {
+    "step": 137,
+    "epoch": 3.859154929577465,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676066304,
+    "loss": 1.2884,
+    "grad_norm": 0.4196622669696808,
+    "learning_rate": 4.194304228229806e-07
+  },
+  {
+    "step": 138,
+    "epoch": 3.887323943661972,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676077056,
+    "loss": 1.3246,
+    "grad_norm": 0.25930461287498474,
+    "learning_rate": 1.8646181716164831e-07
+  },
+  {
+    "step": 139,
+    "epoch": 3.915492957746479,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676069376,
+    "loss": 1.351,
+    "grad_norm": 0.3037481904029846,
+    "learning_rate": 4.662269987756317e-08
+  },
+  {
+    "step": 140,
+    "epoch": 3.943661971830986,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676103168,
+    "loss": 1.312,
+    "grad_norm": 0.31067949533462524,
+    "learning_rate": 0.0
+  },
+  {
+    "step": 140,
+    "epoch": 3.943661971830986,
+    "cpu_mem": 1.507319808,
+    "gpu_mem": 4.676103168,
+    "train_runtime": 689.3565,
+    "train_samples_per_second": 13.061,
+    "train_steps_per_second": 0.203,
+    "total_flos": 0.0,
+    "train_loss": 1.444502578462873
+  }
+]

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_e-r8-a2/adapter_config.json

@@ -0,0 +1,30 @@
+{
+  "alpha": 16,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "init_weight": "kaiming",
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "metric_tracking": false,
+  "modules_to_save": null,
+  "peft_type": "ABLATION",
+  "r": 8,
+  "revision": null,
+  "seed": 42,
+  "share_weights": false,
+  "target_modules": [
+    "up_proj",
+    "gate_proj",
+    "v_proj",
+    "k_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj"
+  ],
+  "task_type": null,
+  "track_n": 100,
+  "variant": "B"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_e-r8-a2/eval_results.json

@@ -0,0 +1,4 @@
+{
+    "task": "arc_e",
+    "results": 0.4718013468013468
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_e-r8-a2/training_configuration.json

@@ -0,0 +1,38 @@
+{
+  "model_id": "TinyLlama/TinyLlama_v1.1",
+  "dataset": {
+    "name": "ARC_E",
+    "dataset_id": "allenai/ai2_arc",
+    "preprocess_id": "arc_train_deepeval"
+  },
+  "peft_config": {
+    "method": "abl_B",
+    "rank": 8,
+    "alpha": 16,
+    "dropout": 0.0,
+    "bias": "none",
+    "target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "down_proj",
+      "up_proj"
+    ],
+    "trainable_parameter_count": 3548160
+  },
+  "training_config": {
+    "max_dataset_length": null,
+    "batch_size": 64,
+    "per_device_batch_size": 32,
+    "gradient_accumulation_steps": 2,
+    "learning_rate": 0.0003,
+    "num_epochs": 4,
+    "warmup_ratio": 0.1
+  },
+  "model_name": "TinyLlama_v1.1-abl_B-arc_e-r8-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_e-r8-a2",
+  "seed": 42,
+  "timestamp": "2025-08-31T19:54:43.238015"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-arc_e-r8-a2/training_logs.json

@@ -0,0 +1,1273 @@
+[
+  {
+    "step": 1,
+    "epoch": 0.028169014084507043,
+    "cpu_mem": 1.497096192,
+    "gpu_mem": 4.444216832,
+    "loss": 4.6319,
+    "grad_norm": 90.68135070800781,
+    "learning_rate": 2.1428571428571425e-05
+  },
+  {
+    "step": 2,
+    "epoch": 0.056338028169014086,
+    "cpu_mem": 1.500438528,
+    "gpu_mem": 4.47266304,
+    "loss": 4.4578,
+    "grad_norm": 88.92479705810547,
+    "learning_rate": 4.285714285714285e-05
+  },
+  {
+    "step": 3,
+    "epoch": 0.08450704225352113,
+    "cpu_mem": 1.500635136,
+    "gpu_mem": 4.472641536,
+    "loss": 3.8569,
+    "grad_norm": 78.79303741455078,
+    "learning_rate": 6.428571428571427e-05
+  },
+  {
+    "step": 4,
+    "epoch": 0.11267605633802817,
+    "cpu_mem": 1.500831744,
+    "gpu_mem": 4.472620032,
+    "loss": 3.0224,
+    "grad_norm": 53.755924224853516,
+    "learning_rate": 8.57142857142857e-05
+  },
+  {
+    "step": 5,
+    "epoch": 0.14084507042253522,
+    "cpu_mem": 1.500831744,
+    "gpu_mem": 4.472661504,
+    "loss": 2.0876,
+    "grad_norm": 23.539710998535156,
+    "learning_rate": 0.00010714285714285714
+  },
+  {
+    "step": 6,
+    "epoch": 0.16901408450704225,
+    "cpu_mem": 1.500831744,
+    "gpu_mem": 4.472636928,
+    "loss": 1.6383,
+    "grad_norm": 9.254476547241211,
+    "learning_rate": 0.00012857142857142855
+  },
+  {
+    "step": 7,
+    "epoch": 0.19718309859154928,
+    "cpu_mem": 1.500831744,
+    "gpu_mem": 4.472659968,
+    "loss": 1.5543,
+    "grad_norm": 5.9162468910217285,
+    "learning_rate": 0.00015
+  },
+  {
+    "step": 8,
+    "epoch": 0.22535211267605634,
+    "cpu_mem": 1.500831744,
+    "gpu_mem": 4.472618496,
+    "loss": 1.4366,
+    "grad_norm": 4.133505344390869,
+    "learning_rate": 0.0001714285714285714
+  },
+  {
+    "step": 9,
+    "epoch": 0.2535211267605634,
+    "cpu_mem": 1.500831744,
+    "gpu_mem": 4.472620032,
+    "loss": 1.4241,
+    "grad_norm": 8.360941886901855,
+    "learning_rate": 0.00019285714285714286
+  },
+  {
+    "step": 10,
+    "epoch": 0.28169014084507044,
+    "cpu_mem": 1.501028352,
+    "gpu_mem": 4.472615424,
+    "loss": 1.533,
+    "grad_norm": 9.691158294677734,
+    "learning_rate": 0.00021428571428571427
+  },
+  {
+    "step": 11,
+    "epoch": 0.30985915492957744,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.47269376,
+    "loss": 1.4012,
+    "grad_norm": 4.719605922698975,
+    "learning_rate": 0.00023571428571428569
+  },
+  {
+    "step": 12,
+    "epoch": 0.3380281690140845,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472667648,
+    "loss": 1.3237,
+    "grad_norm": 3.155473470687866,
+    "learning_rate": 0.0002571428571428571
+  },
+  {
+    "step": 13,
+    "epoch": 0.36619718309859156,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472618496,
+    "loss": 1.4359,
+    "grad_norm": 5.259221076965332,
+    "learning_rate": 0.00027857142857142854
+  },
+  {
+    "step": 14,
+    "epoch": 0.39436619718309857,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.47264,
+    "loss": 1.3195,
+    "grad_norm": 1.7507895231246948,
+    "learning_rate": 0.0003
+  },
+  {
+    "step": 15,
+    "epoch": 0.4225352112676056,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.47261696,
+    "loss": 1.406,
+    "grad_norm": 4.849987030029297,
+    "learning_rate": 0.0002999533773001224
+  },
+  {
+    "step": 16,
+    "epoch": 0.4507042253521127,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472621568,
+    "loss": 1.4401,
+    "grad_norm": 4.631597518920898,
+    "learning_rate": 0.0002998135381828383
+  },
+  {
+    "step": 17,
+    "epoch": 0.4788732394366197,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472658432,
+    "loss": 1.3536,
+    "grad_norm": 2.7577764987945557,
+    "learning_rate": 0.00029958056957717696
+  },
+  {
+    "step": 18,
+    "epoch": 0.5070422535211268,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472669184,
+    "loss": 1.3395,
+    "grad_norm": 2.595319986343384,
+    "learning_rate": 0.0002992546163048102
+  },
+  {
+    "step": 19,
+    "epoch": 0.5352112676056338,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472612352,
+    "loss": 1.2984,
+    "grad_norm": 2.2393436431884766,
+    "learning_rate": 0.0002988358809900258
+  },
+  {
+    "step": 20,
+    "epoch": 0.5633802816901409,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472683008,
+    "loss": 1.3524,
+    "grad_norm": 1.4507402181625366,
+    "learning_rate": 0.0002983246239337692
+  },
+  {
+    "step": 21,
+    "epoch": 0.5915492957746479,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472681472,
+    "loss": 1.3479,
+    "grad_norm": 2.475274085998535,
+    "learning_rate": 0.0002977211629518312
+  },
+  {
+    "step": 22,
+    "epoch": 0.6197183098591549,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472638464,
+    "loss": 1.3795,
+    "grad_norm": 2.8394975662231445,
+    "learning_rate": 0.00029702587317728153
+  },
+  {
+    "step": 23,
+    "epoch": 0.647887323943662,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.47265536,
+    "loss": 1.3218,
+    "grad_norm": 1.7564839124679565,
+    "learning_rate": 0.0002962391868272735
+  },
+  {
+    "step": 24,
+    "epoch": 0.676056338028169,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472612352,
+    "loss": 1.3391,
+    "grad_norm": 1.4141991138458252,
+    "learning_rate": 0.00029536159293436166
+  },
+  {
+    "step": 25,
+    "epoch": 0.704225352112676,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472641536,
+    "loss": 1.3794,
+    "grad_norm": 1.3704174757003784,
+    "learning_rate": 0.00029439363704250176
+  },
+  {
+    "step": 26,
+    "epoch": 0.7323943661971831,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472621568,
+    "loss": 1.4543,
+    "grad_norm": 1.8014308214187622,
+    "learning_rate": 0.00029333592086792107
+  },
+  {
+    "step": 27,
+    "epoch": 0.7605633802816901,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.47264768,
+    "loss": 1.3353,
+    "grad_norm": 1.1347827911376953,
+    "learning_rate": 0.0002921891019250697
+  },
+  {
+    "step": 28,
+    "epoch": 0.7887323943661971,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.47264768,
+    "loss": 1.386,
+    "grad_norm": 1.4631065130233765,
+    "learning_rate": 0.0002909538931178862
+  },
+  {
+    "step": 29,
+    "epoch": 0.8169014084507042,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472626176,
+    "loss": 1.2431,
+    "grad_norm": 1.1491742134094238,
+    "learning_rate": 0.00028963106229663063
+  },
+  {
+    "step": 30,
+    "epoch": 0.8450704225352113,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.47261696,
+    "loss": 1.3533,
+    "grad_norm": 1.2432610988616943,
+    "learning_rate": 0.00028822143178056114
+  },
+  {
+    "step": 31,
+    "epoch": 0.8732394366197183,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472635392,
+    "loss": 1.3643,
+    "grad_norm": 1.3368273973464966,
+    "learning_rate": 0.00028672587784675096
+  },
+  {
+    "step": 32,
+    "epoch": 0.9014084507042254,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472658432,
+    "loss": 1.3383,
+    "grad_norm": 1.5168676376342773,
+    "learning_rate": 0.0002851453301853628
+  },
+  {
+    "step": 33,
+    "epoch": 0.9295774647887324,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.47265536,
+    "loss": 1.356,
+    "grad_norm": 0.9831933975219727,
+    "learning_rate": 0.00028348077132172027
+  },
+  {
+    "step": 34,
+    "epoch": 0.9577464788732394,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472658432,
+    "loss": 1.4134,
+    "grad_norm": 1.6918983459472656,
+    "learning_rate": 0.0002817332360055343
+  },
+  {
+    "step": 35,
+    "epoch": 0.9859154929577465,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.47264,
+    "loss": 1.328,
+    "grad_norm": 1.3503168821334839,
+    "learning_rate": 0.0002799038105676658
+  },
+  {
+    "step": 36,
+    "epoch": 1.0140845070422535,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486823424,
+    "loss": 1.9758,
+    "grad_norm": 2.2334275245666504,
+    "learning_rate": 0.0002779936322448233
+  },
+  {
+    "step": 37,
+    "epoch": 1.0422535211267605,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486828032,
+    "loss": 1.3284,
+    "grad_norm": 0.5176252722740173,
+    "learning_rate": 0.0002760038884726157
+  },
+  {
+    "step": 38,
+    "epoch": 1.0704225352112675,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486806528,
+    "loss": 1.23,
+    "grad_norm": 0.9931122660636902,
+    "learning_rate": 0.00027393581614739923
+  },
+  {
+    "step": 39,
+    "epoch": 1.0985915492957747,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486795776,
+    "loss": 1.3856,
+    "grad_norm": 2.389493227005005,
+    "learning_rate": 0.0002717907008573785
+  },
+  {
+    "step": 40,
+    "epoch": 1.1267605633802817,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486858752,
+    "loss": 1.4521,
+    "grad_norm": 2.60158109664917,
+    "learning_rate": 0.0002695698760834384
+  },
+  {
+    "step": 41,
+    "epoch": 1.1549295774647887,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486818816,
+    "loss": 1.3356,
+    "grad_norm": 1.25742769241333,
+    "learning_rate": 0.00026727472237020447
+  },
+  {
+    "step": 42,
+    "epoch": 1.1830985915492958,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486861824,
+    "loss": 1.3599,
+    "grad_norm": 1.3035224676132202,
+    "learning_rate": 0.00026490666646784665
+  },
+  {
+    "step": 43,
+    "epoch": 1.2112676056338028,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486811136,
+    "loss": 1.4067,
+    "grad_norm": 1.275063157081604,
+    "learning_rate": 0.0002624671804451601
+  },
+  {
+    "step": 44,
+    "epoch": 1.2394366197183098,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486875648,
+    "loss": 1.3442,
+    "grad_norm": 1.034511685371399,
+    "learning_rate": 0.0002599577807744739
+  },
+  {
+    "step": 45,
+    "epoch": 1.267605633802817,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486843392,
+    "loss": 1.3551,
+    "grad_norm": 0.8410855531692505,
+    "learning_rate": 0.0002573800273889577
+  },
+  {
+    "step": 46,
+    "epoch": 1.295774647887324,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486848,
+    "loss": 1.3701,
+    "grad_norm": 0.9161557555198669,
+    "learning_rate": 0.0002547355227129109
+  },
+  {
+    "step": 47,
+    "epoch": 1.323943661971831,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.48679424,
+    "loss": 1.3437,
+    "grad_norm": 1.512771725654602,
+    "learning_rate": 0.00025202591066563786
+  },
+  {
+    "step": 48,
+    "epoch": 1.352112676056338,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486808064,
+    "loss": 1.346,
+    "grad_norm": 1.1189930438995361,
+    "learning_rate": 0.0002492528756395289
+  },
+  {
+    "step": 49,
+    "epoch": 1.380281690140845,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486797312,
+    "loss": 1.3254,
+    "grad_norm": 0.9964473247528076,
+    "learning_rate": 0.0002464181414529809
+  },
+  {
+    "step": 50,
+    "epoch": 1.408450704225352,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486811136,
+    "loss": 1.3244,
+    "grad_norm": 0.8401857018470764,
+    "learning_rate": 0.00024352347027881003
+  },
+  {
+    "step": 51,
+    "epoch": 1.436619718309859,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.48686336,
+    "loss": 1.3491,
+    "grad_norm": 0.9280771017074585,
+    "learning_rate": 0.0002405706615488216
+  },
+  {
+    "step": 52,
+    "epoch": 1.4647887323943662,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486811136,
+    "loss": 1.3614,
+    "grad_norm": 1.0336897373199463,
+    "learning_rate": 0.00023756155083521846
+  },
+  {
+    "step": 53,
+    "epoch": 1.4929577464788732,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486880256,
+    "loss": 1.3085,
+    "grad_norm": 0.844360888004303,
+    "learning_rate": 0.00023449800870954326
+  },
+  {
+    "step": 54,
+    "epoch": 1.5211267605633803,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486848,
+    "loss": 1.2801,
+    "grad_norm": 0.8565107583999634,
+    "learning_rate": 0.0002313819395798639
+  },
+  {
+    "step": 55,
+    "epoch": 1.5492957746478875,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486857216,
+    "loss": 1.3645,
+    "grad_norm": 0.8742291331291199,
+    "learning_rate": 0.0002282152805069247
+  },
+  {
+    "step": 56,
+    "epoch": 1.5774647887323945,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.48683264,
+    "loss": 1.3072,
+    "grad_norm": 0.6819919943809509,
+    "learning_rate": 0.000225
+  },
+  {
+    "step": 57,
+    "epoch": 1.6056338028169015,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486866432,
+    "loss": 1.2991,
+    "grad_norm": 0.5670233964920044,
+    "learning_rate": 0.00022173809679319772
+  },
+  {
+    "step": 58,
+    "epoch": 1.6338028169014085,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486848,
+    "loss": 1.3135,
+    "grad_norm": 0.7747777700424194,
+    "learning_rate": 0.00021843159860297442
+  },
+  {
+    "step": 59,
+    "epoch": 1.6619718309859155,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486834176,
+    "loss": 1.3134,
+    "grad_norm": 0.5816622972488403,
+    "learning_rate": 0.00021508256086763368
+  },
+  {
+    "step": 60,
+    "epoch": 1.6901408450704225,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486872576,
+    "loss": 1.2977,
+    "grad_norm": 1.1774585247039795,
+    "learning_rate": 0.00021169306546959174
+  },
+  {
+    "step": 61,
+    "epoch": 1.7183098591549295,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486804992,
+    "loss": 1.2838,
+    "grad_norm": 0.9861280918121338,
+    "learning_rate": 0.0002082652194412042
+  },
+  {
+    "step": 62,
+    "epoch": 1.7464788732394365,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486852608,
+    "loss": 1.3528,
+    "grad_norm": 1.0061546564102173,
+    "learning_rate": 0.00020480115365495926
+  },
+  {
+    "step": 63,
+    "epoch": 1.7746478873239435,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.48680192,
+    "loss": 1.316,
+    "grad_norm": 1.0553845167160034,
+    "learning_rate": 0.00020130302149885031
+  },
+  {
+    "step": 64,
+    "epoch": 1.8028169014084507,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486851072,
+    "loss": 1.3355,
+    "grad_norm": 1.1395550966262817,
+    "learning_rate": 0.00019777299753775265
+  },
+  {
+    "step": 65,
+    "epoch": 1.8309859154929577,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486849536,
+    "loss": 1.3297,
+    "grad_norm": 1.126117467880249,
+    "learning_rate": 0.00019421327616163563
+  },
+  {
+    "step": 66,
+    "epoch": 1.8591549295774648,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486867968,
+    "loss": 1.2832,
+    "grad_norm": 1.099063754081726,
+    "learning_rate": 0.00019062607022145078
+  },
+  {
+    "step": 67,
+    "epoch": 1.887323943661972,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.4868096,
+    "loss": 1.314,
+    "grad_norm": 1.0150798559188843,
+    "learning_rate": 0.00018701360965354402
+  },
+  {
+    "step": 68,
+    "epoch": 1.915492957746479,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486821888,
+    "loss": 1.3478,
+    "grad_norm": 1.0355052947998047,
+    "learning_rate": 0.00018337814009344714
+  },
+  {
+    "step": 69,
+    "epoch": 1.943661971830986,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486846464,
+    "loss": 1.3066,
+    "grad_norm": 1.6115487813949585,
+    "learning_rate": 0.0001797219214799096
+  },
+  {
+    "step": 70,
+    "epoch": 1.971830985915493,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486823424,
+    "loss": 1.2561,
+    "grad_norm": 0.8908656239509583,
+    "learning_rate": 0.00017604722665003956
+  },
+  {
+    "step": 71,
+    "epoch": 2.0,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486659072,
+    "loss": 1.9607,
+    "grad_norm": 3.0582046508789062,
+    "learning_rate": 0.00017235633992642615
+  },
+  {
+    "step": 72,
+    "epoch": 2.028169014084507,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472650752,
+    "loss": 1.3384,
+    "grad_norm": 1.3399873971939087,
+    "learning_rate": 0.00016865155569712278
+  },
+  {
+    "step": 73,
+    "epoch": 2.056338028169014,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472613888,
+    "loss": 1.2917,
+    "grad_norm": 0.8912840485572815,
+    "learning_rate": 0.0001649351769893725
+  },
+  {
+    "step": 74,
+    "epoch": 2.084507042253521,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472673792,
+    "loss": 1.278,
+    "grad_norm": 0.9684709906578064,
+    "learning_rate": 0.00016120951403796364
+  },
+  {
+    "step": 75,
+    "epoch": 2.112676056338028,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472641536,
+    "loss": 1.3237,
+    "grad_norm": 1.1640229225158691,
+    "learning_rate": 0.00015747688284910457
+  },
+  {
+    "step": 76,
+    "epoch": 2.140845070422535,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472652288,
+    "loss": 1.2621,
+    "grad_norm": 0.9244154095649719,
+    "learning_rate": 0.00015373960376071093
+  },
+  {
+    "step": 77,
+    "epoch": 2.169014084507042,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472689152,
+    "loss": 1.3137,
+    "grad_norm": 1.2289559841156006,
+    "learning_rate": 0.00015
+  },
+  {
+    "step": 78,
+    "epoch": 2.1971830985915495,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472673792,
+    "loss": 1.3245,
+    "grad_norm": 1.2233545780181885,
+    "learning_rate": 0.00014626039623928907
+  },
+  {
+    "step": 79,
+    "epoch": 2.2253521126760565,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.47262464,
+    "loss": 1.2031,
+    "grad_norm": 1.283607006072998,
+    "learning_rate": 0.0001425231171508954
+  },
+  {
+    "step": 80,
+    "epoch": 2.2535211267605635,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472669184,
+    "loss": 1.299,
+    "grad_norm": 1.1909815073013306,
+    "learning_rate": 0.00013879048596203636
+  },
+  {
+    "step": 81,
+    "epoch": 2.2816901408450705,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.47265536,
+    "loss": 1.3448,
+    "grad_norm": 2.1881954669952393,
+    "learning_rate": 0.0001350648230106275
+  },
+  {
+    "step": 82,
+    "epoch": 2.3098591549295775,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472623104,
+    "loss": 1.3249,
+    "grad_norm": 2.604688882827759,
+    "learning_rate": 0.00013134844430287725
+  },
+  {
+    "step": 83,
+    "epoch": 2.3380281690140845,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472673792,
+    "loss": 1.2162,
+    "grad_norm": 2.0677669048309326,
+    "learning_rate": 0.0001276436600735738
+  },
+  {
+    "step": 84,
+    "epoch": 2.3661971830985915,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472612352,
+    "loss": 1.3604,
+    "grad_norm": 2.706294536590576,
+    "learning_rate": 0.00012395277334996044
+  },
+  {
+    "step": 85,
+    "epoch": 2.3943661971830985,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472658432,
+    "loss": 1.2945,
+    "grad_norm": 1.944639801979065,
+    "learning_rate": 0.00012027807852009038
+  },
+  {
+    "step": 86,
+    "epoch": 2.4225352112676055,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472612352,
+    "loss": 1.2707,
+    "grad_norm": 1.4408667087554932,
+    "learning_rate": 0.00011662185990655284
+  },
+  {
+    "step": 87,
+    "epoch": 2.4507042253521125,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472643072,
+    "loss": 1.3298,
+    "grad_norm": 2.4344043731689453,
+    "learning_rate": 0.00011298639034645593
+  },
+  {
+    "step": 88,
+    "epoch": 2.4788732394366195,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472618496,
+    "loss": 1.2685,
+    "grad_norm": 1.9301280975341797,
+    "learning_rate": 0.00010937392977854923
+  },
+  {
+    "step": 89,
+    "epoch": 2.507042253521127,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472672256,
+    "loss": 1.2436,
+    "grad_norm": 1.8113797903060913,
+    "learning_rate": 0.00010578672383836435
+  },
+  {
+    "step": 90,
+    "epoch": 2.535211267605634,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472653824,
+    "loss": 1.2687,
+    "grad_norm": 1.6126199960708618,
+    "learning_rate": 0.00010222700246224735
+  },
+  {
+    "step": 91,
+    "epoch": 2.563380281690141,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472603136,
+    "loss": 1.3103,
+    "grad_norm": 2.3668925762176514,
+    "learning_rate": 9.869697850114969e-05
+  },
+  {
+    "step": 92,
+    "epoch": 2.591549295774648,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472627712,
+    "loss": 1.282,
+    "grad_norm": 1.9559086561203003,
+    "learning_rate": 9.519884634504074e-05
+  },
+  {
+    "step": 93,
+    "epoch": 2.619718309859155,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472630784,
+    "loss": 1.2231,
+    "grad_norm": 1.8475598096847534,
+    "learning_rate": 9.17347805587958e-05
+  },
+  {
+    "step": 94,
+    "epoch": 2.647887323943662,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472623104,
+    "loss": 1.2242,
+    "grad_norm": 2.266403913497925,
+    "learning_rate": 8.830693453040829e-05
+  },
+  {
+    "step": 95,
+    "epoch": 2.676056338028169,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472661504,
+    "loss": 1.2711,
+    "grad_norm": 1.83024263381958,
+    "learning_rate": 8.491743913236628e-05
+  },
+  {
+    "step": 96,
+    "epoch": 2.704225352112676,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.47267072,
+    "loss": 1.202,
+    "grad_norm": 2.0701658725738525,
+    "learning_rate": 8.156840139702554e-05
+  },
+  {
+    "step": 97,
+    "epoch": 2.732394366197183,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472613888,
+    "loss": 1.2861,
+    "grad_norm": 2.78592848777771,
+    "learning_rate": 7.82619032068023e-05
+  },
+  {
+    "step": 98,
+    "epoch": 2.76056338028169,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472613888,
+    "loss": 1.2864,
+    "grad_norm": 1.7303667068481445,
+    "learning_rate": 7.500000000000002e-05
+  },
+  {
+    "step": 99,
+    "epoch": 2.788732394366197,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472610816,
+    "loss": 1.2419,
+    "grad_norm": 1.7865890264511108,
+    "learning_rate": 7.17847194930753e-05
+  },
+  {
+    "step": 100,
+    "epoch": 2.816901408450704,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.47260928,
+    "loss": 1.1842,
+    "grad_norm": 2.1414265632629395,
+    "learning_rate": 6.86180604201361e-05
+  },
+  {
+    "step": 101,
+    "epoch": 2.845070422535211,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472652288,
+    "loss": 1.1673,
+    "grad_norm": 2.0667600631713867,
+    "learning_rate": 6.550199129045668e-05
+  },
+  {
+    "step": 102,
+    "epoch": 2.873239436619718,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472590848,
+    "loss": 1.2343,
+    "grad_norm": 1.725559949874878,
+    "learning_rate": 6.243844916478155e-05
+  },
+  {
+    "step": 103,
+    "epoch": 2.9014084507042255,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.47264,
+    "loss": 1.2293,
+    "grad_norm": 2.002014636993408,
+    "learning_rate": 5.9429338451178355e-05
+  },
+  {
+    "step": 104,
+    "epoch": 2.9295774647887325,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472702976,
+    "loss": 1.3138,
+    "grad_norm": 3.1678977012634277,
+    "learning_rate": 5.6476529721189974e-05
+  },
+  {
+    "step": 105,
+    "epoch": 2.9577464788732395,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.47265536,
+    "loss": 1.167,
+    "grad_norm": 2.2888123989105225,
+    "learning_rate": 5.358185854701909e-05
+  },
+  {
+    "step": 106,
+    "epoch": 2.9859154929577465,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.472636928,
+    "loss": 1.2732,
+    "grad_norm": 2.5358312129974365,
+    "learning_rate": 5.074712436047112e-05
+  },
+  {
+    "step": 107,
+    "epoch": 3.0140845070422535,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486846464,
+    "loss": 1.681,
+    "grad_norm": 4.78402042388916,
+    "learning_rate": 4.7974089334362057e-05
+  },
+  {
+    "step": 108,
+    "epoch": 3.0422535211267605,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486828032,
+    "loss": 1.1895,
+    "grad_norm": 2.0640642642974854,
+    "learning_rate": 4.526447728708908e-05
+  },
+  {
+    "step": 109,
+    "epoch": 3.0704225352112675,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.48681728,
+    "loss": 1.151,
+    "grad_norm": 2.4174320697784424,
+    "learning_rate": 4.261997261104223e-05
+  },
+  {
+    "step": 110,
+    "epoch": 3.0985915492957745,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.48687104,
+    "loss": 1.2597,
+    "grad_norm": 3.956610679626465,
+    "learning_rate": 4.004221922552608e-05
+  },
+  {
+    "step": 111,
+    "epoch": 3.1267605633802815,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486831104,
+    "loss": 1.1957,
+    "grad_norm": 1.9434964656829834,
+    "learning_rate": 3.753281955483985e-05
+  },
+  {
+    "step": 112,
+    "epoch": 3.1549295774647885,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486849536,
+    "loss": 1.1921,
+    "grad_norm": 1.951103925704956,
+    "learning_rate": 3.509333353215331e-05
+  },
+  {
+    "step": 113,
+    "epoch": 3.183098591549296,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486912512,
+    "loss": 1.1653,
+    "grad_norm": 1.9486435651779175,
+    "learning_rate": 3.2725277629795526e-05
+  },
+  {
+    "step": 114,
+    "epoch": 3.211267605633803,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.48684032,
+    "loss": 1.2145,
+    "grad_norm": 2.336803913116455,
+    "learning_rate": 3.0430123916561672e-05
+  },
+  {
+    "step": 115,
+    "epoch": 3.23943661971831,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486834176,
+    "loss": 1.234,
+    "grad_norm": 2.942000389099121,
+    "learning_rate": 2.8209299142621522e-05
+  },
+  {
+    "step": 116,
+    "epoch": 3.267605633802817,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486849536,
+    "loss": 1.1358,
+    "grad_norm": 1.8644171953201294,
+    "learning_rate": 2.6064183852600797e-05
+  },
+  {
+    "step": 117,
+    "epoch": 3.295774647887324,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486864896,
+    "loss": 1.1317,
+    "grad_norm": 2.1343953609466553,
+    "learning_rate": 2.3996111527384288e-05
+  },
+  {
+    "step": 118,
+    "epoch": 3.323943661971831,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.48685568,
+    "loss": 1.1576,
+    "grad_norm": 2.370473861694336,
+    "learning_rate": 2.2006367755176655e-05
+  },
+  {
+    "step": 119,
+    "epoch": 3.352112676056338,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486846464,
+    "loss": 1.1716,
+    "grad_norm": 2.426577091217041,
+    "learning_rate": 2.009618943233419e-05
+  },
+  {
+    "step": 120,
+    "epoch": 3.380281690140845,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486864896,
+    "loss": 1.176,
+    "grad_norm": 2.2006006240844727,
+    "learning_rate": 1.82667639944657e-05
+  },
+  {
+    "step": 121,
+    "epoch": 3.408450704225352,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.48686336,
+    "loss": 1.1531,
+    "grad_norm": 2.4512929916381836,
+    "learning_rate": 1.6519228678279718e-05
+  },
+  {
+    "step": 122,
+    "epoch": 3.436619718309859,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486820352,
+    "loss": 1.0892,
+    "grad_norm": 3.1564323902130127,
+    "learning_rate": 1.4854669814637143e-05
+  },
+  {
+    "step": 123,
+    "epoch": 3.464788732394366,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486852608,
+    "loss": 1.144,
+    "grad_norm": 2.4003117084503174,
+    "learning_rate": 1.3274122153249028e-05
+  },
+  {
+    "step": 124,
+    "epoch": 3.492957746478873,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486806528,
+    "loss": 1.1869,
+    "grad_norm": 2.464019298553467,
+    "learning_rate": 1.1778568219438839e-05
+  },
+  {
+    "step": 125,
+    "epoch": 3.52112676056338,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486851072,
+    "loss": 1.1083,
+    "grad_norm": 3.2846879959106445,
+    "learning_rate": 1.036893770336938e-05
+  },
+  {
+    "step": 126,
+    "epoch": 3.5492957746478875,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.48680192,
+    "loss": 1.0998,
+    "grad_norm": 2.879990577697754,
+    "learning_rate": 9.046106882113751e-06
+  },
+  {
+    "step": 127,
+    "epoch": 3.5774647887323945,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486814208,
+    "loss": 1.1369,
+    "grad_norm": 2.237708568572998,
+    "learning_rate": 7.810898074930243e-06
+  },
+  {
+    "step": 128,
+    "epoch": 3.6056338028169015,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486838784,
+    "loss": 1.1598,
+    "grad_norm": 2.6036179065704346,
+    "learning_rate": 6.664079132078881e-06
+  },
+  {
+    "step": 129,
+    "epoch": 3.6338028169014085,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486800384,
+    "loss": 1.1029,
+    "grad_norm": 2.8772053718566895,
+    "learning_rate": 5.606362957498195e-06
+  },
+  {
+    "step": 130,
+    "epoch": 3.6619718309859155,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486803456,
+    "loss": 1.167,
+    "grad_norm": 2.4127655029296875,
+    "learning_rate": 4.638407065638322e-06
+  },
+  {
+    "step": 131,
+    "epoch": 3.6901408450704225,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486815744,
+    "loss": 1.0721,
+    "grad_norm": 2.208256721496582,
+    "learning_rate": 3.760813172726457e-06
+  },
+  {
+    "step": 132,
+    "epoch": 3.7183098591549295,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486780416,
+    "loss": 1.1263,
+    "grad_norm": 2.983234167098999,
+    "learning_rate": 2.9741268227184255e-06
+  },
+  {
+    "step": 133,
+    "epoch": 3.7464788732394365,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486821888,
+    "loss": 1.1618,
+    "grad_norm": 2.294952630996704,
+    "learning_rate": 2.2788370481687965e-06
+  },
+  {
+    "step": 134,
+    "epoch": 3.7746478873239435,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486837248,
+    "loss": 1.1226,
+    "grad_norm": 2.202094078063965,
+    "learning_rate": 1.6753760662307215e-06
+  },
+  {
+    "step": 135,
+    "epoch": 3.802816901408451,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.48680192,
+    "loss": 1.1488,
+    "grad_norm": 2.804492712020874,
+    "learning_rate": 1.1641190099741904e-06
+  },
+  {
+    "step": 136,
+    "epoch": 3.830985915492958,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.4868096,
+    "loss": 1.173,
+    "grad_norm": 3.213949203491211,
+    "learning_rate": 7.453836951897885e-07
+  },
+  {
+    "step": 137,
+    "epoch": 3.859154929577465,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486831104,
+    "loss": 1.112,
+    "grad_norm": 3.0322439670562744,
+    "learning_rate": 4.194304228229806e-07
+  },
+  {
+    "step": 138,
+    "epoch": 3.887323943661972,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486841856,
+    "loss": 1.1919,
+    "grad_norm": 2.2161591053009033,
+    "learning_rate": 1.8646181716164831e-07
+  },
+  {
+    "step": 139,
+    "epoch": 3.915492957746479,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486834176,
+    "loss": 1.1989,
+    "grad_norm": 2.8926453590393066,
+    "learning_rate": 4.662269987756317e-08
+  },
+  {
+    "step": 140,
+    "epoch": 3.943661971830986,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486867968,
+    "loss": 1.1475,
+    "grad_norm": 2.6210014820098877,
+    "learning_rate": 0.0
+  },
+  {
+    "step": 140,
+    "epoch": 3.943661971830986,
+    "cpu_mem": 1.50122496,
+    "gpu_mem": 4.486867968,
+    "train_runtime": 684.2008,
+    "train_samples_per_second": 13.16,
+    "train_steps_per_second": 0.205,
+    "total_flos": 0.0,
+    "train_loss": 1.3794820129871368
+  }
+]

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-boolq-r2-a2/adapter_config.json

@@ -0,0 +1,30 @@
+{
+  "alpha": 4,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "init_weight": "kaiming",
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "metric_tracking": false,
+  "modules_to_save": null,
+  "peft_type": "ABLATION",
+  "r": 2,
+  "revision": null,
+  "seed": 42,
+  "share_weights": false,
+  "target_modules": [
+    "up_proj",
+    "gate_proj",
+    "v_proj",
+    "k_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj"
+  ],
+  "task_type": null,
+  "track_n": 100,
+  "variant": "B"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-boolq-r2-a2/eval_results.json

@@ -0,0 +1,4 @@
+{
+    "task": "boolq",
+    "results": 0.745565749235474
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-boolq-r2-a2/training_configuration.json

@@ -0,0 +1,38 @@
+{
+  "model_id": "TinyLlama/TinyLlama_v1.1",
+  "dataset": {
+    "name": "BOOLQ",
+    "dataset_id": "google/boolq",
+    "preprocess_id": "boolq_train_deepeval"
+  },
+  "peft_config": {
+    "method": "abl_B",
+    "rank": 2,
+    "alpha": 4,
+    "dropout": 0.0,
+    "bias": "none",
+    "target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "down_proj",
+      "up_proj"
+    ],
+    "trainable_parameter_count": 1182720
+  },
+  "training_config": {
+    "max_dataset_length": null,
+    "batch_size": 64,
+    "per_device_batch_size": 32,
+    "gradient_accumulation_steps": 2,
+    "learning_rate": 0.0003,
+    "num_epochs": 2,
+    "warmup_ratio": 0.1
+  },
+  "model_name": "TinyLlama_v1.1-abl_B-boolq-r2-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-boolq-r2-a2",
+  "seed": 42,
+  "timestamp": "2025-08-31T09:05:54.792561"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-boolq-r2-a2/training_logs.json

@@ -0,0 +1,2659 @@
+[
+  {
+    "step": 1,
+    "epoch": 0.006779661016949152,
+    "cpu_mem": 1.497100288,
+    "gpu_mem": 4.425657344,
+    "loss": 8.869,
+    "grad_norm": 207.28607177734375,
+    "learning_rate": 9.999999999999999e-06
+  },
+  {
+    "step": 2,
+    "epoch": 0.013559322033898305,
+    "cpu_mem": 1.503391744,
+    "gpu_mem": 4.435256832,
+    "loss": 8.9376,
+    "grad_norm": 212.0312957763672,
+    "learning_rate": 1.9999999999999998e-05
+  },
+  {
+    "step": 3,
+    "epoch": 0.020338983050847456,
+    "cpu_mem": 1.504178176,
+    "gpu_mem": 4.435175424,
+    "loss": 8.332,
+    "grad_norm": 210.19793701171875,
+    "learning_rate": 2.9999999999999997e-05
+  },
+  {
+    "step": 4,
+    "epoch": 0.02711864406779661,
+    "cpu_mem": 1.504768,
+    "gpu_mem": 4.435175424,
+    "loss": 7.0801,
+    "grad_norm": 211.4102783203125,
+    "learning_rate": 3.9999999999999996e-05
+  },
+  {
+    "step": 5,
+    "epoch": 0.03389830508474576,
+    "cpu_mem": 1.505161216,
+    "gpu_mem": 4.435110912,
+    "loss": 5.4329,
+    "grad_norm": 188.5390167236328,
+    "learning_rate": 4.9999999999999996e-05
+  },
+  {
+    "step": 6,
+    "epoch": 0.04067796610169491,
+    "cpu_mem": 1.50575104,
+    "gpu_mem": 4.43513088,
+    "loss": 3.7775,
+    "grad_norm": 175.3853759765625,
+    "learning_rate": 5.9999999999999995e-05
+  },
+  {
+    "step": 7,
+    "epoch": 0.04745762711864407,
+    "cpu_mem": 1.506340864,
+    "gpu_mem": 4.435183104,
+    "loss": 2.0568,
+    "grad_norm": 97.72562408447266,
+    "learning_rate": 7e-05
+  },
+  {
+    "step": 8,
+    "epoch": 0.05423728813559322,
+    "cpu_mem": 1.506930688,
+    "gpu_mem": 4.43526912,
+    "loss": 1.3082,
+    "grad_norm": 58.757537841796875,
+    "learning_rate": 7.999999999999999e-05
+  },
+  {
+    "step": 9,
+    "epoch": 0.061016949152542375,
+    "cpu_mem": 1.507323904,
+    "gpu_mem": 4.43517696,
+    "loss": 1.0698,
+    "grad_norm": 30.670557022094727,
+    "learning_rate": 8.999999999999999e-05
+  },
+  {
+    "step": 10,
+    "epoch": 0.06779661016949153,
+    "cpu_mem": 1.50771712,
+    "gpu_mem": 4.43507712,
+    "loss": 0.8257,
+    "grad_norm": 19.861032485961914,
+    "learning_rate": 9.999999999999999e-05
+  },
+  {
+    "step": 11,
+    "epoch": 0.07457627118644068,
+    "cpu_mem": 1.508110336,
+    "gpu_mem": 4.435181568,
+    "loss": 0.7766,
+    "grad_norm": 37.903358459472656,
+    "learning_rate": 0.00010999999999999998
+  },
+  {
+    "step": 12,
+    "epoch": 0.08135593220338982,
+    "cpu_mem": 1.50870016,
+    "gpu_mem": 4.43555328,
+    "loss": 0.7157,
+    "grad_norm": 27.937366485595703,
+    "learning_rate": 0.00011999999999999999
+  },
+  {
+    "step": 13,
+    "epoch": 0.08813559322033898,
+    "cpu_mem": 1.509093376,
+    "gpu_mem": 4.435156992,
+    "loss": 0.8872,
+    "grad_norm": 60.55340576171875,
+    "learning_rate": 0.00013
+  },
+  {
+    "step": 14,
+    "epoch": 0.09491525423728814,
+    "cpu_mem": 1.509486592,
+    "gpu_mem": 4.435133952,
+    "loss": 0.8207,
+    "grad_norm": 46.76423263549805,
+    "learning_rate": 0.00014
+  },
+  {
+    "step": 15,
+    "epoch": 0.1016949152542373,
+    "cpu_mem": 1.509879808,
+    "gpu_mem": 4.435072512,
+    "loss": 0.7102,
+    "grad_norm": 11.082263946533203,
+    "learning_rate": 0.00015
+  },
+  {
+    "step": 16,
+    "epoch": 0.10847457627118644,
+    "cpu_mem": 1.510273024,
+    "gpu_mem": 4.435156992,
+    "loss": 1.083,
+    "grad_norm": 63.073631286621094,
+    "learning_rate": 0.00015999999999999999
+  },
+  {
+    "step": 17,
+    "epoch": 0.1152542372881356,
+    "cpu_mem": 1.51066624,
+    "gpu_mem": 4.435196928,
+    "loss": 0.8847,
+    "grad_norm": 41.545772552490234,
+    "learning_rate": 0.00016999999999999999
+  },
+  {
+    "step": 18,
+    "epoch": 0.12203389830508475,
+    "cpu_mem": 1.510862848,
+    "gpu_mem": 4.435259904,
+    "loss": 0.6895,
+    "grad_norm": 7.644906997680664,
+    "learning_rate": 0.00017999999999999998
+  },
+  {
+    "step": 19,
+    "epoch": 0.1288135593220339,
+    "cpu_mem": 1.511256064,
+    "gpu_mem": 4.435097088,
+    "loss": 1.0504,
+    "grad_norm": 38.996315002441406,
+    "learning_rate": 0.00018999999999999998
+  },
+  {
+    "step": 20,
+    "epoch": 0.13559322033898305,
+    "cpu_mem": 1.51164928,
+    "gpu_mem": 4.435209216,
+    "loss": 0.7336,
+    "grad_norm": 17.267147064208984,
+    "learning_rate": 0.00019999999999999998
+  },
+  {
+    "step": 21,
+    "epoch": 0.1423728813559322,
+    "cpu_mem": 1.511845888,
+    "gpu_mem": 4.435367424,
+    "loss": 0.6753,
+    "grad_norm": 6.746187210083008,
+    "learning_rate": 0.00020999999999999998
+  },
+  {
+    "step": 22,
+    "epoch": 0.14915254237288136,
+    "cpu_mem": 1.512239104,
+    "gpu_mem": 4.435259904,
+    "loss": 0.7545,
+    "grad_norm": 8.339751243591309,
+    "learning_rate": 0.00021999999999999995
+  },
+  {
+    "step": 23,
+    "epoch": 0.15593220338983052,
+    "cpu_mem": 1.512435712,
+    "gpu_mem": 4.435232256,
+    "loss": 0.7935,
+    "grad_norm": 18.06570053100586,
+    "learning_rate": 0.00023
+  },
+  {
+    "step": 24,
+    "epoch": 0.16271186440677965,
+    "cpu_mem": 1.512828928,
+    "gpu_mem": 4.435289088,
+    "loss": 0.6579,
+    "grad_norm": 11.259378433227539,
+    "learning_rate": 0.00023999999999999998
+  },
+  {
+    "step": 25,
+    "epoch": 0.1694915254237288,
+    "cpu_mem": 1.513025536,
+    "gpu_mem": 4.435074048,
+    "loss": 0.8052,
+    "grad_norm": 18.03961181640625,
+    "learning_rate": 0.00025
+  },
+  {
+    "step": 26,
+    "epoch": 0.17627118644067796,
+    "cpu_mem": 1.513418752,
+    "gpu_mem": 4.435129344,
+    "loss": 0.8655,
+    "grad_norm": 23.352096557617188,
+    "learning_rate": 0.00026
+  },
+  {
+    "step": 27,
+    "epoch": 0.18305084745762712,
+    "cpu_mem": 1.51361536,
+    "gpu_mem": 4.435421184,
+    "loss": 0.718,
+    "grad_norm": 33.54894256591797,
+    "learning_rate": 0.00027
+  },
+  {
+    "step": 28,
+    "epoch": 0.18983050847457628,
+    "cpu_mem": 1.513811968,
+    "gpu_mem": 4.43510016,
+    "loss": 0.7011,
+    "grad_norm": 5.752996444702148,
+    "learning_rate": 0.00028
+  },
+  {
+    "step": 29,
+    "epoch": 0.19661016949152543,
+    "cpu_mem": 1.514205184,
+    "gpu_mem": 4.435164672,
+    "loss": 0.7024,
+    "grad_norm": 24.037891387939453,
+    "learning_rate": 0.00029
+  },
+  {
+    "step": 30,
+    "epoch": 0.2033898305084746,
+    "cpu_mem": 1.514401792,
+    "gpu_mem": 4.435243008,
+    "loss": 0.9994,
+    "grad_norm": 46.48765563964844,
+    "learning_rate": 0.0003
+  },
+  {
+    "step": 31,
+    "epoch": 0.21016949152542372,
+    "cpu_mem": 1.5145984,
+    "gpu_mem": 4.4350464,
+    "loss": 0.6251,
+    "grad_norm": 3.328892946243286,
+    "learning_rate": 0.0002999893794250036
+  },
+  {
+    "step": 32,
+    "epoch": 0.21694915254237288,
+    "cpu_mem": 1.514795008,
+    "gpu_mem": 4.435160064,
+    "loss": 0.6636,
+    "grad_norm": 2.738682746887207,
+    "learning_rate": 0.00029995751920396937
+  },
+  {
+    "step": 33,
+    "epoch": 0.22372881355932203,
+    "cpu_mem": 1.514991616,
+    "gpu_mem": 4.435398144,
+    "loss": 0.7977,
+    "grad_norm": 45.13636779785156,
+    "learning_rate": 0.00029990442384854874
+  },
+  {
+    "step": 34,
+    "epoch": 0.2305084745762712,
+    "cpu_mem": 1.515188224,
+    "gpu_mem": 4.43510016,
+    "loss": 0.5935,
+    "grad_norm": 10.75675106048584,
+    "learning_rate": 0.0002998301008774512
+  },
+  {
+    "step": 35,
+    "epoch": 0.23728813559322035,
+    "cpu_mem": 1.51558144,
+    "gpu_mem": 4.435310592,
+    "loss": 1.3179,
+    "grad_norm": 55.147945404052734,
+    "learning_rate": 0.0002997345608153792
+  },
+  {
+    "step": 36,
+    "epoch": 0.2440677966101695,
+    "cpu_mem": 1.51558144,
+    "gpu_mem": 4.43526144,
+    "loss": 0.8981,
+    "grad_norm": 45.676902770996094,
+    "learning_rate": 0.000299617817191538
+  },
+  {
+    "step": 37,
+    "epoch": 0.25084745762711863,
+    "cpu_mem": 1.515778048,
+    "gpu_mem": 4.435072512,
+    "loss": 0.6745,
+    "grad_norm": 13.375166893005371,
+    "learning_rate": 0.0002994798865377198
+  },
+  {
+    "step": 38,
+    "epoch": 0.2576271186440678,
+    "cpu_mem": 1.516171264,
+    "gpu_mem": 4.435319808,
+    "loss": 0.7386,
+    "grad_norm": 9.287933349609375,
+    "learning_rate": 0.0002993207883859627
+  },
+  {
+    "step": 39,
+    "epoch": 0.26440677966101694,
+    "cpu_mem": 1.516367872,
+    "gpu_mem": 4.4356992,
+    "loss": 0.6679,
+    "grad_norm": 2.182807445526123,
+    "learning_rate": 0.0002991405452657846
+  },
+  {
+    "step": 40,
+    "epoch": 0.2711864406779661,
+    "cpu_mem": 1.51656448,
+    "gpu_mem": 4.43526912,
+    "loss": 0.664,
+    "grad_norm": 4.061365127563477,
+    "learning_rate": 0.00029893918270099324
+  },
+  {
+    "step": 41,
+    "epoch": 0.27796610169491526,
+    "cpu_mem": 1.516761088,
+    "gpu_mem": 4.435496448,
+    "loss": 0.6778,
+    "grad_norm": 4.150041580200195,
+    "learning_rate": 0.00029871672920607153
+  },
+  {
+    "step": 42,
+    "epoch": 0.2847457627118644,
+    "cpu_mem": 1.516957696,
+    "gpu_mem": 4.435393536,
+    "loss": 0.6646,
+    "grad_norm": 1.6074423789978027,
+    "learning_rate": 0.0002984732162821399
+  },
+  {
+    "step": 43,
+    "epoch": 0.29152542372881357,
+    "cpu_mem": 1.516957696,
+    "gpu_mem": 4.43521536,
+    "loss": 0.6099,
+    "grad_norm": 3.170870780944824,
+    "learning_rate": 0.0002982086784124952
+  },
+  {
+    "step": 44,
+    "epoch": 0.2983050847457627,
+    "cpu_mem": 1.517154304,
+    "gpu_mem": 4.435358208,
+    "loss": 0.6507,
+    "grad_norm": 9.807236671447754,
+    "learning_rate": 0.00029792315305772796
+  },
+  {
+    "step": 45,
+    "epoch": 0.3050847457627119,
+    "cpu_mem": 1.517154304,
+    "gpu_mem": 4.43513856,
+    "loss": 0.9994,
+    "grad_norm": 25.055511474609375,
+    "learning_rate": 0.0002976166806504174
+  },
+  {
+    "step": 46,
+    "epoch": 0.31186440677966104,
+    "cpu_mem": 1.517350912,
+    "gpu_mem": 4.435381248,
+    "loss": 0.7107,
+    "grad_norm": 7.407582759857178,
+    "learning_rate": 0.00029728930458940595
+  },
+  {
+    "step": 47,
+    "epoch": 0.31864406779661014,
+    "cpu_mem": 1.51754752,
+    "gpu_mem": 4.435104768,
+    "loss": 0.6873,
+    "grad_norm": 13.582527160644531,
+    "learning_rate": 0.00029694107123365385
+  },
+  {
+    "step": 48,
+    "epoch": 0.3254237288135593,
+    "cpu_mem": 1.51754752,
+    "gpu_mem": 4.435181568,
+    "loss": 0.8356,
+    "grad_norm": 24.060340881347656,
+    "learning_rate": 0.00029657202989567393
+  },
+  {
+    "step": 49,
+    "epoch": 0.33220338983050846,
+    "cpu_mem": 1.517744128,
+    "gpu_mem": 4.435198464,
+    "loss": 0.7423,
+    "grad_norm": 3.9435641765594482,
+    "learning_rate": 0.00029618223283454893
+  },
+  {
+    "step": 50,
+    "epoch": 0.3389830508474576,
+    "cpu_mem": 1.517940736,
+    "gpu_mem": 4.435137024,
+    "loss": 0.6187,
+    "grad_norm": 2.644559860229492,
+    "learning_rate": 0.00029577173524853123
+  },
+  {
+    "step": 51,
+    "epoch": 0.34576271186440677,
+    "cpu_mem": 1.517940736,
+    "gpu_mem": 4.435141632,
+    "loss": 0.5859,
+    "grad_norm": 2.845573663711548,
+    "learning_rate": 0.0002953405952672261
+  },
+  {
+    "step": 52,
+    "epoch": 0.3525423728813559,
+    "cpu_mem": 1.518137344,
+    "gpu_mem": 4.435221504,
+    "loss": 0.7151,
+    "grad_norm": 9.970312118530273,
+    "learning_rate": 0.0002948888739433602
+  },
+  {
+    "step": 53,
+    "epoch": 0.3593220338983051,
+    "cpu_mem": 1.518333952,
+    "gpu_mem": 4.435244544,
+    "loss": 0.5678,
+    "grad_norm": 5.6443681716918945,
+    "learning_rate": 0.0002944166352441363
+  },
+  {
+    "step": 54,
+    "epoch": 0.36610169491525424,
+    "cpu_mem": 1.518333952,
+    "gpu_mem": 4.435172352,
+    "loss": 0.7156,
+    "grad_norm": 6.423981189727783,
+    "learning_rate": 0.0002939239460421746
+  },
+  {
+    "step": 55,
+    "epoch": 0.3728813559322034,
+    "cpu_mem": 1.51853056,
+    "gpu_mem": 4.435442688,
+    "loss": 0.6943,
+    "grad_norm": 8.590938568115234,
+    "learning_rate": 0.00029341087610604337
+  },
+  {
+    "step": 56,
+    "epoch": 0.37966101694915255,
+    "cpu_mem": 1.51853056,
+    "gpu_mem": 4.435229184,
+    "loss": 0.6623,
+    "grad_norm": 2.5790529251098633,
+    "learning_rate": 0.00029287749809037904
+  },
+  {
+    "step": 57,
+    "epoch": 0.3864406779661017,
+    "cpu_mem": 1.518727168,
+    "gpu_mem": 4.43522304,
+    "loss": 0.6916,
+    "grad_norm": 7.401466369628906,
+    "learning_rate": 0.0002923238875255979
+  },
+  {
+    "step": 58,
+    "epoch": 0.39322033898305087,
+    "cpu_mem": 1.518923776,
+    "gpu_mem": 4.435118592,
+    "loss": 0.6048,
+    "grad_norm": 2.3391366004943848,
+    "learning_rate": 0.00029175012280720024
+  },
+  {
+    "step": 59,
+    "epoch": 0.4,
+    "cpu_mem": 1.518923776,
+    "gpu_mem": 4.435135488,
+    "loss": 0.808,
+    "grad_norm": 12.574873924255371,
+    "learning_rate": 0.000291156285184669
+  },
+  {
+    "step": 60,
+    "epoch": 0.4067796610169492,
+    "cpu_mem": 1.519120384,
+    "gpu_mem": 4.435229184,
+    "loss": 0.6018,
+    "grad_norm": 4.364191055297852,
+    "learning_rate": 0.00029054245874996426
+  },
+  {
+    "step": 61,
+    "epoch": 0.4135593220338983,
+    "cpu_mem": 1.519120384,
+    "gpu_mem": 4.435239936,
+    "loss": 0.5963,
+    "grad_norm": 2.257843255996704,
+    "learning_rate": 0.0002899087304256151
+  },
+  {
+    "step": 62,
+    "epoch": 0.42033898305084744,
+    "cpu_mem": 1.519316992,
+    "gpu_mem": 4.435227648,
+    "loss": 0.7185,
+    "grad_norm": 5.300965309143066,
+    "learning_rate": 0.0002892551899524109
+  },
+  {
+    "step": 63,
+    "epoch": 0.4271186440677966,
+    "cpu_mem": 1.5195136,
+    "gpu_mem": 4.435219968,
+    "loss": 0.7026,
+    "grad_norm": 18.619279861450195,
+    "learning_rate": 0.000288581929876693
+  },
+  {
+    "step": 64,
+    "epoch": 0.43389830508474575,
+    "cpu_mem": 1.5195136,
+    "gpu_mem": 4.435149312,
+    "loss": 0.7096,
+    "grad_norm": 15.257659912109375,
+    "learning_rate": 0.0002878890455372498
+  },
+  {
+    "step": 65,
+    "epoch": 0.4406779661016949,
+    "cpu_mem": 1.519710208,
+    "gpu_mem": 4.435193856,
+    "loss": 0.6736,
+    "grad_norm": 3.462217330932617,
+    "learning_rate": 0.0002871766350518159
+  },
+  {
+    "step": 66,
+    "epoch": 0.44745762711864406,
+    "cpu_mem": 1.519710208,
+    "gpu_mem": 4.435387392,
+    "loss": 0.6095,
+    "grad_norm": 5.371577262878418,
+    "learning_rate": 0.00028644479930317775
+  },
+  {
+    "step": 67,
+    "epoch": 0.4542372881355932,
+    "cpu_mem": 1.519710208,
+    "gpu_mem": 4.435097088,
+    "loss": 0.6853,
+    "grad_norm": 7.674093246459961,
+    "learning_rate": 0.00028569364192488803
+  },
+  {
+    "step": 68,
+    "epoch": 0.4610169491525424,
+    "cpu_mem": 1.519710208,
+    "gpu_mem": 4.435064832,
+    "loss": 0.7368,
+    "grad_norm": 8.521327018737793,
+    "learning_rate": 0.00028492326928659045
+  },
+  {
+    "step": 69,
+    "epoch": 0.46779661016949153,
+    "cpu_mem": 1.519906816,
+    "gpu_mem": 4.43513088,
+    "loss": 0.6601,
+    "grad_norm": 9.785889625549316,
+    "learning_rate": 0.00028413379047895665
+  },
+  {
+    "step": 70,
+    "epoch": 0.4745762711864407,
+    "cpu_mem": 1.519906816,
+    "gpu_mem": 4.435124736,
+    "loss": 0.6927,
+    "grad_norm": 10.841557502746582,
+    "learning_rate": 0.0002833253172982385
+  },
+  {
+    "step": 71,
+    "epoch": 0.48135593220338985,
+    "cpu_mem": 1.519906816,
+    "gpu_mem": 4.4353536,
+    "loss": 0.6767,
+    "grad_norm": 10.955390930175781,
+    "learning_rate": 0.0002824979642304366
+  },
+  {
+    "step": 72,
+    "epoch": 0.488135593220339,
+    "cpu_mem": 1.519906816,
+    "gpu_mem": 4.43534592,
+    "loss": 0.6142,
+    "grad_norm": 2.390930414199829,
+    "learning_rate": 0.0002816518484350883
+  },
+  {
+    "step": 73,
+    "epoch": 0.49491525423728816,
+    "cpu_mem": 1.520103424,
+    "gpu_mem": 4.435312128,
+    "loss": 0.7588,
+    "grad_norm": 11.52270793914795,
+    "learning_rate": 0.0002807870897286772
+  },
+  {
+    "step": 74,
+    "epoch": 0.5016949152542373,
+    "cpu_mem": 1.520103424,
+    "gpu_mem": 4.435172352,
+    "loss": 0.6045,
+    "grad_norm": 3.1932780742645264,
+    "learning_rate": 0.0002799038105676658
+  },
+  {
+    "step": 75,
+    "epoch": 0.5084745762711864,
+    "cpu_mem": 1.520103424,
+    "gpu_mem": 4.435097088,
+    "loss": 0.589,
+    "grad_norm": 7.821681022644043,
+    "learning_rate": 0.000279002136031155
+  },
+  {
+    "step": 76,
+    "epoch": 0.5152542372881356,
+    "cpu_mem": 1.520103424,
+    "gpu_mem": 4.435037184,
+    "loss": 0.6281,
+    "grad_norm": 4.9963226318359375,
+    "learning_rate": 0.00027808219380317216
+  },
+  {
+    "step": 77,
+    "epoch": 0.5220338983050847,
+    "cpu_mem": 1.520103424,
+    "gpu_mem": 4.435110912,
+    "loss": 0.5632,
+    "grad_norm": 2.8097784519195557,
+    "learning_rate": 0.0002771441141545895
+  },
+  {
+    "step": 78,
+    "epoch": 0.5288135593220339,
+    "cpu_mem": 1.520103424,
+    "gpu_mem": 4.435163136,
+    "loss": 0.9578,
+    "grad_norm": 19.26412582397461,
+    "learning_rate": 0.0002761880299246772
+  },
+  {
+    "step": 79,
+    "epoch": 0.535593220338983,
+    "cpu_mem": 1.520300032,
+    "gpu_mem": 4.435295232,
+    "loss": 0.7552,
+    "grad_norm": 11.019858360290527,
+    "learning_rate": 0.000275214076502292
+  },
+  {
+    "step": 80,
+    "epoch": 0.5423728813559322,
+    "cpu_mem": 1.520300032,
+    "gpu_mem": 4.435186176,
+    "loss": 0.6377,
+    "grad_norm": 6.720715045928955,
+    "learning_rate": 0.0002742223918067056
+  },
+  {
+    "step": 81,
+    "epoch": 0.5491525423728814,
+    "cpu_mem": 1.52049664,
+    "gpu_mem": 4.435066368,
+    "loss": 0.5784,
+    "grad_norm": 3.100093364715576,
+    "learning_rate": 0.00027321311626807374
+  },
+  {
+    "step": 82,
+    "epoch": 0.5559322033898305,
+    "cpu_mem": 1.52049664,
+    "gpu_mem": 4.435135488,
+    "loss": 0.6546,
+    "grad_norm": 7.847732067108154,
+    "learning_rate": 0.0002721863928075503
+  },
+  {
+    "step": 83,
+    "epoch": 0.5627118644067797,
+    "cpu_mem": 1.52049664,
+    "gpu_mem": 4.435235328,
+    "loss": 0.6415,
+    "grad_norm": 7.157805442810059,
+    "learning_rate": 0.000271142366817049
+  },
+  {
+    "step": 84,
+    "epoch": 0.5694915254237288,
+    "cpu_mem": 1.52049664,
+    "gpu_mem": 4.435198464,
+    "loss": 0.6897,
+    "grad_norm": 10.05351734161377,
+    "learning_rate": 0.00027008118613865406
+  },
+  {
+    "step": 85,
+    "epoch": 0.576271186440678,
+    "cpu_mem": 1.52049664,
+    "gpu_mem": 4.43523072,
+    "loss": 0.5937,
+    "grad_norm": 3.456000804901123,
+    "learning_rate": 0.00026900300104368524
+  },
+  {
+    "step": 86,
+    "epoch": 0.5830508474576271,
+    "cpu_mem": 1.52049664,
+    "gpu_mem": 4.435181568,
+    "loss": 0.6938,
+    "grad_norm": 6.894677639007568,
+    "learning_rate": 0.00026790796421141813
+  },
+  {
+    "step": 87,
+    "epoch": 0.5898305084745763,
+    "cpu_mem": 1.52049664,
+    "gpu_mem": 4.435189248,
+    "loss": 0.6673,
+    "grad_norm": 9.324849128723145,
+    "learning_rate": 0.00026679623070746325
+  },
+  {
+    "step": 88,
+    "epoch": 0.5966101694915255,
+    "cpu_mem": 1.520693248,
+    "gpu_mem": 4.435333632,
+    "loss": 0.5659,
+    "grad_norm": 5.584108352661133,
+    "learning_rate": 0.0002656679579618081
+  },
+  {
+    "step": 89,
+    "epoch": 0.6033898305084746,
+    "cpu_mem": 1.520693248,
+    "gpu_mem": 4.43511552,
+    "loss": 0.6689,
+    "grad_norm": 5.593686580657959,
+    "learning_rate": 0.0002645233057465235
+  },
+  {
+    "step": 90,
+    "epoch": 0.6101694915254238,
+    "cpu_mem": 1.520693248,
+    "gpu_mem": 4.43516928,
+    "loss": 0.6263,
+    "grad_norm": 5.61871862411499,
+    "learning_rate": 0.00026336243615313873
+  },
+  {
+    "step": 91,
+    "epoch": 0.6169491525423729,
+    "cpu_mem": 1.520693248,
+    "gpu_mem": 4.435137024,
+    "loss": 0.6892,
+    "grad_norm": 10.633136749267578,
+    "learning_rate": 0.00026218551356968814
+  },
+  {
+    "step": 92,
+    "epoch": 0.6237288135593221,
+    "cpu_mem": 1.520693248,
+    "gpu_mem": 4.435218432,
+    "loss": 0.6887,
+    "grad_norm": 3.8378069400787354,
+    "learning_rate": 0.00026099270465743254
+  },
+  {
+    "step": 93,
+    "epoch": 0.6305084745762712,
+    "cpu_mem": 1.520693248,
+    "gpu_mem": 4.435021824,
+    "loss": 0.7163,
+    "grad_norm": 3.294466018676758,
+    "learning_rate": 0.0002597841783272588
+  },
+  {
+    "step": 94,
+    "epoch": 0.6372881355932203,
+    "cpu_mem": 1.520693248,
+    "gpu_mem": 4.435135488,
+    "loss": 0.5549,
+    "grad_norm": 4.280175685882568,
+    "learning_rate": 0.0002585601057157605
+  },
+  {
+    "step": 95,
+    "epoch": 0.6440677966101694,
+    "cpu_mem": 1.520693248,
+    "gpu_mem": 4.435155456,
+    "loss": 0.6492,
+    "grad_norm": 5.376728057861328,
+    "learning_rate": 0.00025732066016100394
+  },
+  {
+    "step": 96,
+    "epoch": 0.6508474576271186,
+    "cpu_mem": 1.520693248,
+    "gpu_mem": 4.435193856,
+    "loss": 0.6214,
+    "grad_norm": 5.850833415985107,
+    "learning_rate": 0.00025606601717798207
+  },
+  {
+    "step": 97,
+    "epoch": 0.6576271186440678,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435178496,
+    "loss": 0.6908,
+    "grad_norm": 6.41079568862915,
+    "learning_rate": 0.0002547963544337602
+  },
+  {
+    "step": 98,
+    "epoch": 0.6644067796610169,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435090944,
+    "loss": 0.6782,
+    "grad_norm": 3.8216965198516846,
+    "learning_rate": 0.0002535118517223168
+  },
+  {
+    "step": 99,
+    "epoch": 0.6711864406779661,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435040256,
+    "loss": 0.6113,
+    "grad_norm": 2.445533037185669,
+    "learning_rate": 0.00025221269093908365
+  },
+  {
+    "step": 100,
+    "epoch": 0.6779661016949152,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435156992,
+    "loss": 0.6634,
+    "grad_norm": 7.595210552215576,
+    "learning_rate": 0.0002508990560551879
+  },
+  {
+    "step": 101,
+    "epoch": 0.6847457627118644,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435189248,
+    "loss": 0.6559,
+    "grad_norm": 6.5138936042785645,
+    "learning_rate": 0.0002495711330914001
+  },
+  {
+    "step": 102,
+    "epoch": 0.6915254237288135,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.43522304,
+    "loss": 0.635,
+    "grad_norm": 4.007444858551025,
+    "learning_rate": 0.00024822911009179276
+  },
+  {
+    "step": 103,
+    "epoch": 0.6983050847457627,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435273728,
+    "loss": 0.6088,
+    "grad_norm": 2.9485924243927,
+    "learning_rate": 0.0002468731770971113
+  },
+  {
+    "step": 104,
+    "epoch": 0.7050847457627119,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435178496,
+    "loss": 0.6598,
+    "grad_norm": 5.93380880355835,
+    "learning_rate": 0.0002455035261178632
+  },
+  {
+    "step": 105,
+    "epoch": 0.711864406779661,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435279872,
+    "loss": 0.657,
+    "grad_norm": 6.0811567306518555,
+    "learning_rate": 0.0002441203511071278
+  },
+  {
+    "step": 106,
+    "epoch": 0.7186440677966102,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.43523072,
+    "loss": 0.5668,
+    "grad_norm": 3.6183319091796875,
+    "learning_rate": 0.00024272384793309077
+  },
+  {
+    "step": 107,
+    "epoch": 0.7254237288135593,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435118592,
+    "loss": 0.5207,
+    "grad_norm": 3.5147411823272705,
+    "learning_rate": 0.00024131421435130807
+  },
+  {
+    "step": 108,
+    "epoch": 0.7322033898305085,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435302912,
+    "loss": 0.603,
+    "grad_norm": 5.246246814727783,
+    "learning_rate": 0.00023989164997670202
+  },
+  {
+    "step": 109,
+    "epoch": 0.7389830508474576,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435156992,
+    "loss": 0.6161,
+    "grad_norm": 4.675862789154053,
+    "learning_rate": 0.0002384563562552943
+  },
+  {
+    "step": 110,
+    "epoch": 0.7457627118644068,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435160064,
+    "loss": 0.5996,
+    "grad_norm": 3.654783010482788,
+    "learning_rate": 0.0002370085364356797
+  },
+  {
+    "step": 111,
+    "epoch": 0.752542372881356,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435129344,
+    "loss": 0.5225,
+    "grad_norm": 3.8030014038085938,
+    "learning_rate": 0.0002355483955402446
+  },
+  {
+    "step": 112,
+    "epoch": 0.7593220338983051,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435175424,
+    "loss": 0.5475,
+    "grad_norm": 5.957515239715576,
+    "learning_rate": 0.00023407614033613407
+  },
+  {
+    "step": 113,
+    "epoch": 0.7661016949152543,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435166208,
+    "loss": 0.626,
+    "grad_norm": 5.172255039215088,
+    "learning_rate": 0.0002325919793059723
+  },
+  {
+    "step": 114,
+    "epoch": 0.7728813559322034,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435147776,
+    "loss": 0.5508,
+    "grad_norm": 4.3366289138793945,
+    "learning_rate": 0.00023109612261833963
+  },
+  {
+    "step": 115,
+    "epoch": 0.7796610169491526,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.43522304,
+    "loss": 0.5571,
+    "grad_norm": 4.991285800933838,
+    "learning_rate": 0.0002295887820980112
+  },
+  {
+    "step": 116,
+    "epoch": 0.7864406779661017,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435143168,
+    "loss": 0.5788,
+    "grad_norm": 4.1896491050720215,
+    "learning_rate": 0.0002280701711959608
+  },
+  {
+    "step": 117,
+    "epoch": 0.7932203389830509,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435034112,
+    "loss": 0.5795,
+    "grad_norm": 8.061439514160156,
+    "learning_rate": 0.00022654050495913495
+  },
+  {
+    "step": 118,
+    "epoch": 0.8,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435272192,
+    "loss": 0.5868,
+    "grad_norm": 8.318702697753906,
+    "learning_rate": 0.000225
+  },
+  {
+    "step": 119,
+    "epoch": 0.8067796610169492,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435442688,
+    "loss": 0.5089,
+    "grad_norm": 7.266751766204834,
+    "learning_rate": 0.00022344887446586865
+  },
+  {
+    "step": 120,
+    "epoch": 0.8135593220338984,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435175424,
+    "loss": 0.6088,
+    "grad_norm": 9.481569290161133,
+    "learning_rate": 0.00022188734800800852
+  },
+  {
+    "step": 121,
+    "epoch": 0.8203389830508474,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435203072,
+    "loss": 0.5676,
+    "grad_norm": 6.377309322357178,
+    "learning_rate": 0.00022031564175053754
+  },
+  {
+    "step": 122,
+    "epoch": 0.8271186440677966,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.43525376,
+    "loss": 0.534,
+    "grad_norm": 6.072964191436768,
+    "learning_rate": 0.00021873397825911153
+  },
+  {
+    "step": 123,
+    "epoch": 0.8338983050847457,
+    "cpu_mem": 1.520889856,
+    "gpu_mem": 4.435063296,
+    "loss": 0.6156,
+    "grad_norm": 11.986638069152832,
+    "learning_rate": 0.00021714258150940685
+  },
+  {
+    "step": 124,
+    "epoch": 0.8406779661016949,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.435505664,
+    "loss": 0.5422,
+    "grad_norm": 7.253986835479736,
+    "learning_rate": 0.0002155416768554039
+  },
+  {
+    "step": 125,
+    "epoch": 0.847457627118644,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.435232256,
+    "loss": 0.5348,
+    "grad_norm": 5.5946431159973145,
+    "learning_rate": 0.00021393149099747523
+  },
+  {
+    "step": 126,
+    "epoch": 0.8542372881355932,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.43511552,
+    "loss": 0.5548,
+    "grad_norm": 6.215771198272705,
+    "learning_rate": 0.00021231225195028297
+  },
+  {
+    "step": 127,
+    "epoch": 0.8610169491525423,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.435554816,
+    "loss": 0.5796,
+    "grad_norm": 6.4818949699401855,
+    "learning_rate": 0.00021068418901049025
+  },
+  {
+    "step": 128,
+    "epoch": 0.8677966101694915,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.43533056,
+    "loss": 0.5044,
+    "grad_norm": 5.554439067840576,
+    "learning_rate": 0.0002090475327242912
+  },
+  {
+    "step": 129,
+    "epoch": 0.8745762711864407,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.435370496,
+    "loss": 0.6452,
+    "grad_norm": 7.500157356262207,
+    "learning_rate": 0.00020740251485476345
+  },
+  {
+    "step": 130,
+    "epoch": 0.8813559322033898,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.435152384,
+    "loss": 0.6325,
+    "grad_norm": 4.366447925567627,
+    "learning_rate": 0.0002057493683490491
+  },
+  {
+    "step": 131,
+    "epoch": 0.888135593220339,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.435281408,
+    "loss": 0.5926,
+    "grad_norm": 4.608772277832031,
+    "learning_rate": 0.00020408832730536746
+  },
+  {
+    "step": 132,
+    "epoch": 0.8949152542372881,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.435362816,
+    "loss": 0.5173,
+    "grad_norm": 6.282446384429932,
+    "learning_rate": 0.00020241962693986476
+  },
+  {
+    "step": 133,
+    "epoch": 0.9016949152542373,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.43514624,
+    "loss": 0.5126,
+    "grad_norm": 5.64429235458374,
+    "learning_rate": 0.0002007435035533061
+  },
+  {
+    "step": 134,
+    "epoch": 0.9084745762711864,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.435279872,
+    "loss": 0.5368,
+    "grad_norm": 7.870690822601318,
+    "learning_rate": 0.00019906019449761325
+  },
+  {
+    "step": 135,
+    "epoch": 0.9152542372881356,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.435302912,
+    "loss": 0.563,
+    "grad_norm": 7.819180011749268,
+    "learning_rate": 0.00019736993814225374
+  },
+  {
+    "step": 136,
+    "epoch": 0.9220338983050848,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.435140096,
+    "loss": 0.5316,
+    "grad_norm": 7.46820592880249,
+    "learning_rate": 0.00019567297384048604
+  },
+  {
+    "step": 137,
+    "epoch": 0.9288135593220339,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.435020288,
+    "loss": 0.6543,
+    "grad_norm": 12.327062606811523,
+    "learning_rate": 0.0001939695418954653
+  },
+  {
+    "step": 138,
+    "epoch": 0.9355932203389831,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.435201536,
+    "loss": 0.5471,
+    "grad_norm": 7.879848957061768,
+    "learning_rate": 0.00019225988352621445
+  },
+  {
+    "step": 139,
+    "epoch": 0.9423728813559322,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.43510016,
+    "loss": 0.5362,
+    "grad_norm": 4.619059085845947,
+    "learning_rate": 0.00019054424083346592
+  },
+  {
+    "step": 140,
+    "epoch": 0.9491525423728814,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.435152384,
+    "loss": 0.5643,
+    "grad_norm": 8.845608711242676,
+    "learning_rate": 0.0001888228567653781
+  },
+  {
+    "step": 141,
+    "epoch": 0.9559322033898305,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.43518464,
+    "loss": 0.6876,
+    "grad_norm": 13.390810012817383,
+    "learning_rate": 0.0001870959750831323
+  },
+  {
+    "step": 142,
+    "epoch": 0.9627118644067797,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.435324416,
+    "loss": 0.5868,
+    "grad_norm": 14.687874794006348,
+    "learning_rate": 0.0001853638403264141
+  },
+  {
+    "step": 143,
+    "epoch": 0.9694915254237289,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.43530752,
+    "loss": 0.6293,
+    "grad_norm": 8.600774765014648,
+    "learning_rate": 0.00018362669777878453
+  },
+  {
+    "step": 144,
+    "epoch": 0.976271186440678,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.43549952,
+    "loss": 0.5202,
+    "grad_norm": 4.7456254959106445,
+    "learning_rate": 0.00018188479343294648
+  },
+  {
+    "step": 145,
+    "epoch": 0.9830508474576272,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.435210752,
+    "loss": 0.4946,
+    "grad_norm": 5.0703959465026855,
+    "learning_rate": 0.0001801383739559098
+  },
+  {
+    "step": 146,
+    "epoch": 0.9898305084745763,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.43524608,
+    "loss": 0.5895,
+    "grad_norm": 9.480493545532227,
+    "learning_rate": 0.0001783876866540615
+  },
+  {
+    "step": 147,
+    "epoch": 0.9966101694915255,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.435144704,
+    "loss": 0.6059,
+    "grad_norm": 10.175168991088867,
+    "learning_rate": 0.00017663297943814552
+  },
+  {
+    "step": 148,
+    "epoch": 1.0033898305084745,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440039936,
+    "loss": 0.9197,
+    "grad_norm": 15.65149211883545,
+    "learning_rate": 0.0001748745007881561
+  },
+  {
+    "step": 149,
+    "epoch": 1.0101694915254238,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439975424,
+    "loss": 0.4799,
+    "grad_norm": 6.842282295227051,
+    "learning_rate": 0.00017311249971815185
+  },
+  {
+    "step": 150,
+    "epoch": 1.0169491525423728,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439812608,
+    "loss": 0.4679,
+    "grad_norm": 4.733060359954834,
+    "learning_rate": 0.00017134722574099276
+  },
+  {
+    "step": 151,
+    "epoch": 1.023728813559322,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.4398848,
+    "loss": 0.5091,
+    "grad_norm": 4.760961055755615,
+    "learning_rate": 0.00016957892883300775
+  },
+  {
+    "step": 152,
+    "epoch": 1.0305084745762711,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439920128,
+    "loss": 0.4298,
+    "grad_norm": 6.078965187072754,
+    "learning_rate": 0.00016780785939859576
+  },
+  {
+    "step": 153,
+    "epoch": 1.0372881355932204,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439944704,
+    "loss": 0.5997,
+    "grad_norm": 11.618675231933594,
+    "learning_rate": 0.00016603426823476693
+  },
+  {
+    "step": 154,
+    "epoch": 1.0440677966101695,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439906304,
+    "loss": 0.4825,
+    "grad_norm": 8.579179763793945,
+    "learning_rate": 0.00016425840649562736
+  },
+  {
+    "step": 155,
+    "epoch": 1.0508474576271187,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440127488,
+    "loss": 0.4975,
+    "grad_norm": 6.922858238220215,
+    "learning_rate": 0.00016248052565681436
+  },
+  {
+    "step": 156,
+    "epoch": 1.0576271186440678,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440035328,
+    "loss": 0.463,
+    "grad_norm": 6.798807621002197,
+    "learning_rate": 0.00016070087747988482
+  },
+  {
+    "step": 157,
+    "epoch": 1.064406779661017,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439941632,
+    "loss": 0.4941,
+    "grad_norm": 8.606608390808105,
+    "learning_rate": 0.00015891971397666464
+  },
+  {
+    "step": 158,
+    "epoch": 1.071186440677966,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439867904,
+    "loss": 0.4964,
+    "grad_norm": 8.071224212646484,
+    "learning_rate": 0.00015713728737356137
+  },
+  {
+    "step": 159,
+    "epoch": 1.0779661016949154,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440216576,
+    "loss": 0.4681,
+    "grad_norm": 8.937488555908203,
+    "learning_rate": 0.00015535385007584706
+  },
+  {
+    "step": 160,
+    "epoch": 1.0847457627118644,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439811072,
+    "loss": 0.5065,
+    "grad_norm": 10.181997299194336,
+    "learning_rate": 0.0001535696546319161
+  },
+  {
+    "step": 161,
+    "epoch": 1.0915254237288137,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439757312,
+    "loss": 0.4166,
+    "grad_norm": 8.312394142150879,
+    "learning_rate": 0.00015178495369752213
+  },
+  {
+    "step": 162,
+    "epoch": 1.0983050847457627,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440532992,
+    "loss": 0.4284,
+    "grad_norm": 6.323586463928223,
+    "learning_rate": 0.00015
+  },
+  {
+    "step": 163,
+    "epoch": 1.1050847457627118,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440009216,
+    "loss": 0.4603,
+    "grad_norm": 5.615723133087158,
+    "learning_rate": 0.00014821504630247785
+  },
+  {
+    "step": 164,
+    "epoch": 1.111864406779661,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439921664,
+    "loss": 0.5386,
+    "grad_norm": 8.030407905578613,
+    "learning_rate": 0.00014643034536808387
+  },
+  {
+    "step": 165,
+    "epoch": 1.11864406779661,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439870976,
+    "loss": 0.398,
+    "grad_norm": 6.317580223083496,
+    "learning_rate": 0.00014464614992415294
+  },
+  {
+    "step": 166,
+    "epoch": 1.1254237288135593,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439966208,
+    "loss": 0.3961,
+    "grad_norm": 6.63275146484375,
+    "learning_rate": 0.00014286271262643866
+  },
+  {
+    "step": 167,
+    "epoch": 1.1322033898305084,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439883264,
+    "loss": 0.5081,
+    "grad_norm": 9.496688842773438,
+    "learning_rate": 0.00014108028602333536
+  },
+  {
+    "step": 168,
+    "epoch": 1.1389830508474577,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439901696,
+    "loss": 0.5905,
+    "grad_norm": 8.804789543151855,
+    "learning_rate": 0.00013929912252011516
+  },
+  {
+    "step": 169,
+    "epoch": 1.1457627118644067,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439989248,
+    "loss": 0.4639,
+    "grad_norm": 11.636336326599121,
+    "learning_rate": 0.00013751947434318564
+  },
+  {
+    "step": 170,
+    "epoch": 1.152542372881356,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439874048,
+    "loss": 0.4202,
+    "grad_norm": 11.674962997436523,
+    "learning_rate": 0.00013574159350437261
+  },
+  {
+    "step": 171,
+    "epoch": 1.159322033898305,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439937024,
+    "loss": 0.6057,
+    "grad_norm": 12.196778297424316,
+    "learning_rate": 0.0001339657317652331
+  },
+  {
+    "step": 172,
+    "epoch": 1.1661016949152543,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439844864,
+    "loss": 0.4275,
+    "grad_norm": 7.398139953613281,
+    "learning_rate": 0.00013219214060140424
+  },
+  {
+    "step": 173,
+    "epoch": 1.1728813559322033,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440144384,
+    "loss": 0.4748,
+    "grad_norm": 7.616250514984131,
+    "learning_rate": 0.00013042107116699228
+  },
+  {
+    "step": 174,
+    "epoch": 1.1796610169491526,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439867904,
+    "loss": 0.5088,
+    "grad_norm": 9.336560249328613,
+    "learning_rate": 0.00012865277425900724
+  },
+  {
+    "step": 175,
+    "epoch": 1.1864406779661016,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439834112,
+    "loss": 0.4445,
+    "grad_norm": 6.723942279815674,
+    "learning_rate": 0.00012688750028184818
+  },
+  {
+    "step": 176,
+    "epoch": 1.193220338983051,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439972352,
+    "loss": 0.4937,
+    "grad_norm": 11.034165382385254,
+    "learning_rate": 0.0001251254992118439
+  },
+  {
+    "step": 177,
+    "epoch": 1.2,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440070656,
+    "loss": 0.4964,
+    "grad_norm": 8.4466552734375,
+    "learning_rate": 0.00012336702056185453
+  },
+  {
+    "step": 178,
+    "epoch": 1.2067796610169492,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439817216,
+    "loss": 0.5088,
+    "grad_norm": 7.810116767883301,
+    "learning_rate": 0.00012161231334593851
+  },
+  {
+    "step": 179,
+    "epoch": 1.2135593220338983,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439917056,
+    "loss": 0.5458,
+    "grad_norm": 12.21668529510498,
+    "learning_rate": 0.00011986162604409015
+  },
+  {
+    "step": 180,
+    "epoch": 1.2203389830508475,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439889408,
+    "loss": 0.4019,
+    "grad_norm": 6.613919734954834,
+    "learning_rate": 0.00011811520656705348
+  },
+  {
+    "step": 181,
+    "epoch": 1.2271186440677966,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439826432,
+    "loss": 0.3733,
+    "grad_norm": 6.443704128265381,
+    "learning_rate": 0.00011637330222121543
+  },
+  {
+    "step": 182,
+    "epoch": 1.2338983050847459,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440044544,
+    "loss": 0.589,
+    "grad_norm": 11.146683692932129,
+    "learning_rate": 0.00011463615967358588
+  },
+  {
+    "step": 183,
+    "epoch": 1.240677966101695,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439941632,
+    "loss": 0.4674,
+    "grad_norm": 6.876543998718262,
+    "learning_rate": 0.00011290402491686766
+  },
+  {
+    "step": 184,
+    "epoch": 1.2474576271186442,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439889408,
+    "loss": 0.4037,
+    "grad_norm": 6.874241352081299,
+    "learning_rate": 0.00011117714323462186
+  },
+  {
+    "step": 185,
+    "epoch": 1.2542372881355932,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439867904,
+    "loss": 0.472,
+    "grad_norm": 8.139419555664062,
+    "learning_rate": 0.00010945575916653407
+  },
+  {
+    "step": 186,
+    "epoch": 1.2610169491525425,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.43987712,
+    "loss": 0.3631,
+    "grad_norm": 6.349623203277588,
+    "learning_rate": 0.00010774011647378553
+  },
+  {
+    "step": 187,
+    "epoch": 1.2677966101694915,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439809536,
+    "loss": 0.599,
+    "grad_norm": 10.569090843200684,
+    "learning_rate": 0.00010603045810453468
+  },
+  {
+    "step": 188,
+    "epoch": 1.2745762711864406,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439972352,
+    "loss": 0.387,
+    "grad_norm": 6.98309850692749,
+    "learning_rate": 0.00010432702615951396
+  },
+  {
+    "step": 189,
+    "epoch": 1.2813559322033898,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439841792,
+    "loss": 0.5156,
+    "grad_norm": 7.907104015350342,
+    "learning_rate": 0.00010263006185774627
+  },
+  {
+    "step": 190,
+    "epoch": 1.288135593220339,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.4399616,
+    "loss": 0.4805,
+    "grad_norm": 8.252829551696777,
+    "learning_rate": 0.00010093980550238675
+  },
+  {
+    "step": 191,
+    "epoch": 1.2949152542372881,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439780352,
+    "loss": 0.4315,
+    "grad_norm": 8.242355346679688,
+    "learning_rate": 9.925649644669391e-05
+  },
+  {
+    "step": 192,
+    "epoch": 1.3016949152542372,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439912448,
+    "loss": 0.4163,
+    "grad_norm": 13.093923568725586,
+    "learning_rate": 9.758037306013526e-05
+  },
+  {
+    "step": 193,
+    "epoch": 1.3084745762711865,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439886336,
+    "loss": 0.4201,
+    "grad_norm": 6.36842679977417,
+    "learning_rate": 9.591167269463255e-05
+  },
+  {
+    "step": 194,
+    "epoch": 1.3152542372881357,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439852544,
+    "loss": 0.4756,
+    "grad_norm": 6.584967613220215,
+    "learning_rate": 9.425063165095088e-05
+  },
+  {
+    "step": 195,
+    "epoch": 1.3220338983050848,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439956992,
+    "loss": 0.3519,
+    "grad_norm": 6.934253215789795,
+    "learning_rate": 9.259748514523653e-05
+  },
+  {
+    "step": 196,
+    "epoch": 1.3288135593220338,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439952384,
+    "loss": 0.3706,
+    "grad_norm": 7.2520880699157715,
+    "learning_rate": 9.095246727570879e-05
+  },
+  {
+    "step": 197,
+    "epoch": 1.335593220338983,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439811072,
+    "loss": 0.3915,
+    "grad_norm": 10.085671424865723,
+    "learning_rate": 8.931581098950973e-05
+  },
+  {
+    "step": 198,
+    "epoch": 1.3423728813559321,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440003072,
+    "loss": 0.4238,
+    "grad_norm": 6.6942572593688965,
+    "learning_rate": 8.768774804971705e-05
+  },
+  {
+    "step": 199,
+    "epoch": 1.3491525423728814,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.43985408,
+    "loss": 0.4759,
+    "grad_norm": 10.992294311523438,
+    "learning_rate": 8.606850900252478e-05
+  },
+  {
+    "step": 200,
+    "epoch": 1.3559322033898304,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439956992,
+    "loss": 0.4289,
+    "grad_norm": 9.138057708740234,
+    "learning_rate": 8.445832314459608e-05
+  },
+  {
+    "step": 201,
+    "epoch": 1.3627118644067797,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440159744,
+    "loss": 0.4458,
+    "grad_norm": 10.073532104492188,
+    "learning_rate": 8.285741849059311e-05
+  },
+  {
+    "step": 202,
+    "epoch": 1.3694915254237288,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.4399616,
+    "loss": 0.4531,
+    "grad_norm": 8.583861351013184,
+    "learning_rate": 8.126602174088843e-05
+  },
+  {
+    "step": 203,
+    "epoch": 1.376271186440678,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439847936,
+    "loss": 0.3384,
+    "grad_norm": 7.238431930541992,
+    "learning_rate": 7.968435824946242e-05
+  },
+  {
+    "step": 204,
+    "epoch": 1.383050847457627,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.43986176,
+    "loss": 0.3823,
+    "grad_norm": 7.507495403289795,
+    "learning_rate": 7.811265199199152e-05
+  },
+  {
+    "step": 205,
+    "epoch": 1.3898305084745763,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439906304,
+    "loss": 0.5007,
+    "grad_norm": 10.150589942932129,
+    "learning_rate": 7.655112553413135e-05
+  },
+  {
+    "step": 206,
+    "epoch": 1.3966101694915254,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439847936,
+    "loss": 0.383,
+    "grad_norm": 7.802042484283447,
+    "learning_rate": 7.500000000000002e-05
+  },
+  {
+    "step": 207,
+    "epoch": 1.4033898305084747,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440081408,
+    "loss": 0.4301,
+    "grad_norm": 10.05980396270752,
+    "learning_rate": 7.345949504086507e-05
+  },
+  {
+    "step": 208,
+    "epoch": 1.4101694915254237,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440112128,
+    "loss": 0.3913,
+    "grad_norm": 12.579257011413574,
+    "learning_rate": 7.192982880403917e-05
+  },
+  {
+    "step": 209,
+    "epoch": 1.4169491525423727,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.4400384,
+    "loss": 0.4284,
+    "grad_norm": 9.47547721862793,
+    "learning_rate": 7.041121790198881e-05
+  },
+  {
+    "step": 210,
+    "epoch": 1.423728813559322,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439926272,
+    "loss": 0.4698,
+    "grad_norm": 10.045745849609375,
+    "learning_rate": 6.890387738166041e-05
+  },
+  {
+    "step": 211,
+    "epoch": 1.4305084745762713,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439875584,
+    "loss": 0.3637,
+    "grad_norm": 8.85647964477539,
+    "learning_rate": 6.740802069402771e-05
+  },
+  {
+    "step": 212,
+    "epoch": 1.4372881355932203,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439844864,
+    "loss": 0.4818,
+    "grad_norm": 9.431801795959473,
+    "learning_rate": 6.592385966386588e-05
+  },
+  {
+    "step": 213,
+    "epoch": 1.4440677966101694,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439867904,
+    "loss": 0.5163,
+    "grad_norm": 10.95721435546875,
+    "learning_rate": 6.445160445975536e-05
+  },
+  {
+    "step": 214,
+    "epoch": 1.4508474576271186,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439950848,
+    "loss": 0.3934,
+    "grad_norm": 8.903146743774414,
+    "learning_rate": 6.299146356432029e-05
+  },
+  {
+    "step": 215,
+    "epoch": 1.457627118644068,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439878656,
+    "loss": 0.511,
+    "grad_norm": 13.773554801940918,
+    "learning_rate": 6.154364374470568e-05
+  },
+  {
+    "step": 216,
+    "epoch": 1.464406779661017,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440044544,
+    "loss": 0.4131,
+    "grad_norm": 8.684189796447754,
+    "learning_rate": 6.010835002329795e-05
+  },
+  {
+    "step": 217,
+    "epoch": 1.471186440677966,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439886336,
+    "loss": 0.4951,
+    "grad_norm": 12.174453735351562,
+    "learning_rate": 5.8685785648691894e-05
+  },
+  {
+    "step": 218,
+    "epoch": 1.4779661016949153,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439863296,
+    "loss": 0.4811,
+    "grad_norm": 9.615564346313477,
+    "learning_rate": 5.72761520669092e-05
+  },
+  {
+    "step": 219,
+    "epoch": 1.4847457627118645,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439989248,
+    "loss": 0.4467,
+    "grad_norm": 9.505061149597168,
+    "learning_rate": 5.587964889287218e-05
+  },
+  {
+    "step": 220,
+    "epoch": 1.4915254237288136,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.44002304,
+    "loss": 0.4204,
+    "grad_norm": 7.657809734344482,
+    "learning_rate": 5.449647388213678e-05
+  },
+  {
+    "step": 221,
+    "epoch": 1.4983050847457626,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439890944,
+    "loss": 0.4082,
+    "grad_norm": 8.211697578430176,
+    "learning_rate": 5.312682290288869e-05
+  },
+  {
+    "step": 222,
+    "epoch": 1.505084745762712,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440027648,
+    "loss": 0.3794,
+    "grad_norm": 8.586983680725098,
+    "learning_rate": 5.1770889908207245e-05
+  },
+  {
+    "step": 223,
+    "epoch": 1.5118644067796612,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439941632,
+    "loss": 0.409,
+    "grad_norm": 13.101545333862305,
+    "learning_rate": 5.0428866908599864e-05
+  },
+  {
+    "step": 224,
+    "epoch": 1.5186440677966102,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439906304,
+    "loss": 0.3994,
+    "grad_norm": 9.212977409362793,
+    "learning_rate": 4.9100943944812114e-05
+  },
+  {
+    "step": 225,
+    "epoch": 1.5254237288135593,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439870976,
+    "loss": 0.4026,
+    "grad_norm": 7.517599105834961,
+    "learning_rate": 4.778730906091632e-05
+  },
+  {
+    "step": 226,
+    "epoch": 1.5322033898305085,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440019968,
+    "loss": 0.3835,
+    "grad_norm": 9.062960624694824,
+    "learning_rate": 4.648814827768322e-05
+  },
+  {
+    "step": 227,
+    "epoch": 1.5389830508474578,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439909376,
+    "loss": 0.4001,
+    "grad_norm": 8.712440490722656,
+    "learning_rate": 4.5203645566239816e-05
+  },
+  {
+    "step": 228,
+    "epoch": 1.5457627118644068,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.43985408,
+    "loss": 0.5073,
+    "grad_norm": 8.387016296386719,
+    "learning_rate": 4.3933982822017876e-05
+  },
+  {
+    "step": 229,
+    "epoch": 1.5525423728813559,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439795712,
+    "loss": 0.3872,
+    "grad_norm": 7.554939270019531,
+    "learning_rate": 4.267933983899601e-05
+  },
+  {
+    "step": 230,
+    "epoch": 1.559322033898305,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439852544,
+    "loss": 0.3522,
+    "grad_norm": 9.5506010055542,
+    "learning_rate": 4.143989428423947e-05
+  },
+  {
+    "step": 231,
+    "epoch": 1.5661016949152542,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.44013056,
+    "loss": 0.4594,
+    "grad_norm": 9.223462104797363,
+    "learning_rate": 4.0215821672741213e-05
+  },
+  {
+    "step": 232,
+    "epoch": 1.5728813559322035,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.43985408,
+    "loss": 0.4328,
+    "grad_norm": 8.588035583496094,
+    "learning_rate": 3.900729534256745e-05
+  },
+  {
+    "step": 233,
+    "epoch": 1.5796610169491525,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440167424,
+    "loss": 0.344,
+    "grad_norm": 8.180073738098145,
+    "learning_rate": 3.781448643031187e-05
+  },
+  {
+    "step": 234,
+    "epoch": 1.5864406779661016,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440043008,
+    "loss": 0.3725,
+    "grad_norm": 11.422308921813965,
+    "learning_rate": 3.663756384686127e-05
+  },
+  {
+    "step": 235,
+    "epoch": 1.5932203389830508,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439798784,
+    "loss": 0.3704,
+    "grad_norm": 10.164522171020508,
+    "learning_rate": 3.547669425347647e-05
+  },
+  {
+    "step": 236,
+    "epoch": 1.6,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439858688,
+    "loss": 0.3934,
+    "grad_norm": 9.219282150268555,
+    "learning_rate": 3.433204203819185e-05
+  },
+  {
+    "step": 237,
+    "epoch": 1.6067796610169491,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439920128,
+    "loss": 0.4544,
+    "grad_norm": 9.418452262878418,
+    "learning_rate": 3.3203769292536764e-05
+  },
+  {
+    "step": 238,
+    "epoch": 1.6135593220338982,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439921664,
+    "loss": 0.4074,
+    "grad_norm": 8.276749610900879,
+    "learning_rate": 3.209203578858191e-05
+  },
+  {
+    "step": 239,
+    "epoch": 1.6203389830508474,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440175104,
+    "loss": 0.4748,
+    "grad_norm": 9.793453216552734,
+    "learning_rate": 3.099699895631474e-05
+  },
+  {
+    "step": 240,
+    "epoch": 1.6271186440677967,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439824896,
+    "loss": 0.6822,
+    "grad_norm": 12.661322593688965,
+    "learning_rate": 2.9918813861345952e-05
+  },
+  {
+    "step": 241,
+    "epoch": 1.6338983050847458,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440121344,
+    "loss": 0.4709,
+    "grad_norm": 10.91852855682373,
+    "learning_rate": 2.885763318295102e-05
+  },
+  {
+    "step": 242,
+    "epoch": 1.6406779661016948,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439983104,
+    "loss": 0.3463,
+    "grad_norm": 7.739848613739014,
+    "learning_rate": 2.781360719244964e-05
+  },
+  {
+    "step": 243,
+    "epoch": 1.647457627118644,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439835648,
+    "loss": 0.3889,
+    "grad_norm": 10.5182523727417,
+    "learning_rate": 2.6786883731926306e-05
+  },
+  {
+    "step": 244,
+    "epoch": 1.6542372881355933,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439975424,
+    "loss": 0.3747,
+    "grad_norm": 8.963127136230469,
+    "learning_rate": 2.5777608193294396e-05
+  },
+  {
+    "step": 245,
+    "epoch": 1.6610169491525424,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.43985408,
+    "loss": 0.4079,
+    "grad_norm": 7.328197479248047,
+    "learning_rate": 2.4785923497707956e-05
+  },
+  {
+    "step": 246,
+    "epoch": 1.6677966101694914,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439947776,
+    "loss": 0.4373,
+    "grad_norm": 8.23412799835205,
+    "learning_rate": 2.38119700753228e-05
+  },
+  {
+    "step": 247,
+    "epoch": 1.6745762711864407,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439966208,
+    "loss": 0.2948,
+    "grad_norm": 9.671659469604492,
+    "learning_rate": 2.285588584541047e-05
+  },
+  {
+    "step": 248,
+    "epoch": 1.68135593220339,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439918592,
+    "loss": 0.3806,
+    "grad_norm": 8.410043716430664,
+    "learning_rate": 2.1917806196827792e-05
+  },
+  {
+    "step": 249,
+    "epoch": 1.688135593220339,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439824896,
+    "loss": 0.3278,
+    "grad_norm": 8.285307884216309,
+    "learning_rate": 2.0997863968844914e-05
+  },
+  {
+    "step": 250,
+    "epoch": 1.694915254237288,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439917056,
+    "loss": 0.4161,
+    "grad_norm": 9.953680038452148,
+    "learning_rate": 2.009618943233419e-05
+  },
+  {
+    "step": 251,
+    "epoch": 1.7016949152542373,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439829504,
+    "loss": 0.3733,
+    "grad_norm": 10.67496395111084,
+    "learning_rate": 1.921291027132278e-05
+  },
+  {
+    "step": 252,
+    "epoch": 1.7084745762711866,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439872512,
+    "loss": 0.3729,
+    "grad_norm": 7.934334754943848,
+    "learning_rate": 1.834815156491165e-05
+  },
+  {
+    "step": 253,
+    "epoch": 1.7152542372881356,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440066048,
+    "loss": 0.4745,
+    "grad_norm": 10.878839492797852,
+    "learning_rate": 1.750203576956341e-05
+  },
+  {
+    "step": 254,
+    "epoch": 1.7220338983050847,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.43986176,
+    "loss": 0.4278,
+    "grad_norm": 9.874471664428711,
+    "learning_rate": 1.6674682701761493e-05
+  },
+  {
+    "step": 255,
+    "epoch": 1.7288135593220337,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440018432,
+    "loss": 0.4858,
+    "grad_norm": 10.910809516906738,
+    "learning_rate": 1.5866209521043304e-05
+  },
+  {
+    "step": 256,
+    "epoch": 1.735593220338983,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439844864,
+    "loss": 0.4624,
+    "grad_norm": 9.551429748535156,
+    "learning_rate": 1.5076730713409523e-05
+  },
+  {
+    "step": 257,
+    "epoch": 1.7423728813559323,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440258048,
+    "loss": 0.4309,
+    "grad_norm": 8.495111465454102,
+    "learning_rate": 1.4306358075111923e-05
+  },
+  {
+    "step": 258,
+    "epoch": 1.7491525423728813,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439917056,
+    "loss": 0.5244,
+    "grad_norm": 15.041364669799805,
+    "learning_rate": 1.3555200696822232e-05
+  },
+  {
+    "step": 259,
+    "epoch": 1.7559322033898304,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439834112,
+    "loss": 0.3973,
+    "grad_norm": 8.967065811157227,
+    "learning_rate": 1.2823364948184095e-05
+  },
+  {
+    "step": 260,
+    "epoch": 1.7627118644067796,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439950848,
+    "loss": 0.3158,
+    "grad_norm": 7.493344306945801,
+    "learning_rate": 1.2110954462750166e-05
+  },
+  {
+    "step": 261,
+    "epoch": 1.769491525423729,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439906304,
+    "loss": 0.326,
+    "grad_norm": 8.3322114944458,
+    "learning_rate": 1.1418070123306989e-05
+  },
+  {
+    "step": 262,
+    "epoch": 1.776271186440678,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439863296,
+    "loss": 0.3214,
+    "grad_norm": 7.001258850097656,
+    "learning_rate": 1.0744810047589115e-05
+  },
+  {
+    "step": 263,
+    "epoch": 1.783050847457627,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.43990016,
+    "loss": 0.3677,
+    "grad_norm": 7.1652116775512695,
+    "learning_rate": 1.0091269574384874e-05
+  },
+  {
+    "step": 264,
+    "epoch": 1.7898305084745763,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439987712,
+    "loss": 0.3891,
+    "grad_norm": 9.185032844543457,
+    "learning_rate": 9.45754125003576e-06
+  },
+  {
+    "step": 265,
+    "epoch": 1.7966101694915255,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439906304,
+    "loss": 0.4603,
+    "grad_norm": 9.57519245147705,
+    "learning_rate": 8.843714815330987e-06
+  },
+  {
+    "step": 266,
+    "epoch": 1.8033898305084746,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440121344,
+    "loss": 0.4583,
+    "grad_norm": 10.558945655822754,
+    "learning_rate": 8.249877192799731e-06
+  },
+  {
+    "step": 267,
+    "epoch": 1.8101694915254236,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439913984,
+    "loss": 0.4191,
+    "grad_norm": 10.343015670776367,
+    "learning_rate": 7.676112474402068e-06
+  },
+  {
+    "step": 268,
+    "epoch": 1.8169491525423729,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439918592,
+    "loss": 0.3232,
+    "grad_norm": 8.387310028076172,
+    "learning_rate": 7.122501909620926e-06
+  },
+  {
+    "step": 269,
+    "epoch": 1.8237288135593221,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439929344,
+    "loss": 0.4041,
+    "grad_norm": 11.825252532958984,
+    "learning_rate": 6.5891238939566275e-06
+  },
+  {
+    "step": 270,
+    "epoch": 1.8305084745762712,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439967744,
+    "loss": 0.365,
+    "grad_norm": 10.832479476928711,
+    "learning_rate": 6.076053957825411e-06
+  },
+  {
+    "step": 271,
+    "epoch": 1.8372881355932202,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440019968,
+    "loss": 0.4345,
+    "grad_norm": 10.482426643371582,
+    "learning_rate": 5.583364755863701e-06
+  },
+  {
+    "step": 272,
+    "epoch": 1.8440677966101695,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439878656,
+    "loss": 0.4865,
+    "grad_norm": 8.570451736450195,
+    "learning_rate": 5.11112605663977e-06
+  },
+  {
+    "step": 273,
+    "epoch": 1.8508474576271188,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439758848,
+    "loss": 0.425,
+    "grad_norm": 9.54675579071045,
+    "learning_rate": 4.659404732773908e-06
+  },
+  {
+    "step": 274,
+    "epoch": 1.8576271186440678,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439986176,
+    "loss": 0.4173,
+    "grad_norm": 8.669840812683105,
+    "learning_rate": 4.228264751468752e-06
+  },
+  {
+    "step": 275,
+    "epoch": 1.8644067796610169,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.4402304,
+    "loss": 0.3489,
+    "grad_norm": 7.460136890411377,
+    "learning_rate": 3.817767165451041e-06
+  },
+  {
+    "step": 276,
+    "epoch": 1.8711864406779661,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439890944,
+    "loss": 0.4088,
+    "grad_norm": 8.213056564331055,
+    "learning_rate": 3.4279701043260886e-06
+  },
+  {
+    "step": 277,
+    "epoch": 1.8779661016949154,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439837184,
+    "loss": 0.5171,
+    "grad_norm": 11.188621520996094,
+    "learning_rate": 3.0589287663461472e-06
+  },
+  {
+    "step": 278,
+    "epoch": 1.8847457627118644,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.44,
+    "loss": 0.4212,
+    "grad_norm": 8.904882431030273,
+    "learning_rate": 2.710695410593994e-06
+  },
+  {
+    "step": 279,
+    "epoch": 1.8915254237288135,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439940096,
+    "loss": 0.4414,
+    "grad_norm": 9.937751770019531,
+    "learning_rate": 2.3833193495825853e-06
+  },
+  {
+    "step": 280,
+    "epoch": 1.8983050847457628,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439920128,
+    "loss": 0.4172,
+    "grad_norm": 8.643899917602539,
+    "learning_rate": 2.076846942272026e-06
+  },
+  {
+    "step": 281,
+    "epoch": 1.905084745762712,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439855616,
+    "loss": 0.3792,
+    "grad_norm": 8.005178451538086,
+    "learning_rate": 1.791321587504768e-06
+  },
+  {
+    "step": 282,
+    "epoch": 1.911864406779661,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.44028416,
+    "loss": 0.3363,
+    "grad_norm": 10.754042625427246,
+    "learning_rate": 1.5267837178600972e-06
+  },
+  {
+    "step": 283,
+    "epoch": 1.9186440677966101,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439990784,
+    "loss": 0.3576,
+    "grad_norm": 9.666711807250977,
+    "learning_rate": 1.2832707939284427e-06
+  },
+  {
+    "step": 284,
+    "epoch": 1.9254237288135592,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.4398464,
+    "loss": 0.3908,
+    "grad_norm": 9.546576499938965,
+    "learning_rate": 1.0608172990067553e-06
+  },
+  {
+    "step": 285,
+    "epoch": 1.9322033898305084,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.43990016,
+    "loss": 0.4081,
+    "grad_norm": 8.594533920288086,
+    "learning_rate": 8.594547342153979e-07
+  },
+  {
+    "step": 286,
+    "epoch": 1.9389830508474577,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440317952,
+    "loss": 0.4529,
+    "grad_norm": 8.963507652282715,
+    "learning_rate": 6.792116140373116e-07
+  },
+  {
+    "step": 287,
+    "epoch": 1.9457627118644067,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.440087552,
+    "loss": 0.3062,
+    "grad_norm": 8.207653045654297,
+    "learning_rate": 5.201134622801473e-07
+  },
+  {
+    "step": 288,
+    "epoch": 1.9525423728813558,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439872512,
+    "loss": 0.5091,
+    "grad_norm": 11.16333293914795,
+    "learning_rate": 3.821828084619727e-07
+  },
+  {
+    "step": 289,
+    "epoch": 1.959322033898305,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439956992,
+    "loss": 0.4421,
+    "grad_norm": 8.138298988342285,
+    "learning_rate": 2.654391846207915e-07
+  },
+  {
+    "step": 290,
+    "epoch": 1.9661016949152543,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439881728,
+    "loss": 0.4773,
+    "grad_norm": 10.596417427062988,
+    "learning_rate": 1.6989912254880556e-07
+  },
+  {
+    "step": 291,
+    "epoch": 1.9728813559322034,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439917056,
+    "loss": 0.4893,
+    "grad_norm": 8.895089149475098,
+    "learning_rate": 9.557615145123765e-08
+  },
+  {
+    "step": 292,
+    "epoch": 1.9796610169491524,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.44,
+    "loss": 0.3718,
+    "grad_norm": 7.455649375915527,
+    "learning_rate": 4.248079603064724e-08
+  },
+  {
+    "step": 293,
+    "epoch": 1.9864406779661017,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439917056,
+    "loss": 0.5375,
+    "grad_norm": 9.682267189025879,
+    "learning_rate": 1.0620574996372811e-08
+  },
+  {
+    "step": 294,
+    "epoch": 1.993220338983051,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439943168,
+    "loss": 0.4929,
+    "grad_norm": 11.033202171325684,
+    "learning_rate": 0.0
+  },
+  {
+    "step": 294,
+    "epoch": 1.993220338983051,
+    "cpu_mem": 1.521086464,
+    "gpu_mem": 4.439943168,
+    "train_runtime": 4551.7265,
+    "train_samples_per_second": 4.142,
+    "train_steps_per_second": 0.065,
+    "total_flos": 0.0,
+    "train_loss": 0.6919786211381964
+  }
+]

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-boolq-r32-a2/adapter_config.json

@@ -0,0 +1,30 @@
+{
+  "alpha": 64,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "init_weight": "kaiming",
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "metric_tracking": false,
+  "modules_to_save": null,
+  "peft_type": "ABLATION",
+  "r": 32,
+  "revision": null,
+  "seed": 42,
+  "share_weights": false,
+  "target_modules": [
+    "up_proj",
+    "gate_proj",
+    "v_proj",
+    "k_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj"
+  ],
+  "task_type": null,
+  "track_n": 100,
+  "variant": "B"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-boolq-r32-a2/eval_results.json

@@ -0,0 +1,4 @@
+{
+    "task": "boolq",
+    "results": 0.7718654434250765
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-boolq-r32-a2/training_configuration.json

@@ -0,0 +1,38 @@
+{
+  "model_id": "TinyLlama/TinyLlama_v1.1",
+  "dataset": {
+    "name": "BOOLQ",
+    "dataset_id": "google/boolq",
+    "preprocess_id": "boolq_train_deepeval"
+  },
+  "peft_config": {
+    "method": "abl_B",
+    "rank": 32,
+    "alpha": 64,
+    "dropout": 0.0,
+    "bias": "none",
+    "target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "down_proj",
+      "up_proj"
+    ],
+    "trainable_parameter_count": 13009920
+  },
+  "training_config": {
+    "max_dataset_length": null,
+    "batch_size": 64,
+    "per_device_batch_size": 32,
+    "gradient_accumulation_steps": 2,
+    "learning_rate": 0.0003,
+    "num_epochs": 2,
+    "warmup_ratio": 0.1
+  },
+  "model_name": "TinyLlama_v1.1-abl_B-boolq-r32-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-boolq-r32-a2",
+  "seed": 42,
+  "timestamp": "2025-08-31T23:08:42.953080"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-boolq-r32-a2/training_logs.json

@@ -0,0 +1,2659 @@
+[
+  {
+    "step": 1,
+    "epoch": 0.006779661016949152,
+    "cpu_mem": 1.493827584,
+    "gpu_mem": 4.520274944,
+    "loss": 8.869,
+    "grad_norm": 13.0520601272583,
+    "learning_rate": 9.999999999999999e-06
+  },
+  {
+    "step": 2,
+    "epoch": 0.013559322033898305,
+    "cpu_mem": 1.494614016,
+    "gpu_mem": 4.624492032,
+    "loss": 8.9376,
+    "grad_norm": 13.392369270324707,
+    "learning_rate": 1.9999999999999998e-05
+  },
+  {
+    "step": 3,
+    "epoch": 0.020338983050847456,
+    "cpu_mem": 1.49520384,
+    "gpu_mem": 4.624410624,
+    "loss": 8.8022,
+    "grad_norm": 13.587163925170898,
+    "learning_rate": 2.9999999999999997e-05
+  },
+  {
+    "step": 4,
+    "epoch": 0.02711864406779661,
+    "cpu_mem": 1.495793664,
+    "gpu_mem": 4.624410624,
+    "loss": 8.4715,
+    "grad_norm": 13.895402908325195,
+    "learning_rate": 3.9999999999999996e-05
+  },
+  {
+    "step": 5,
+    "epoch": 0.03389830508474576,
+    "cpu_mem": 1.49618688,
+    "gpu_mem": 4.624346112,
+    "loss": 7.9105,
+    "grad_norm": 13.423394203186035,
+    "learning_rate": 4.9999999999999996e-05
+  },
+  {
+    "step": 6,
+    "epoch": 0.04067796610169491,
+    "cpu_mem": 1.496776704,
+    "gpu_mem": 4.62436608,
+    "loss": 7.5511,
+    "grad_norm": 12.783663749694824,
+    "learning_rate": 5.9999999999999995e-05
+  },
+  {
+    "step": 7,
+    "epoch": 0.04745762711864407,
+    "cpu_mem": 1.497366528,
+    "gpu_mem": 4.624418304,
+    "loss": 6.6874,
+    "grad_norm": 13.757925987243652,
+    "learning_rate": 7e-05
+  },
+  {
+    "step": 8,
+    "epoch": 0.05423728813559322,
+    "cpu_mem": 1.497759744,
+    "gpu_mem": 4.62450432,
+    "loss": 5.8091,
+    "grad_norm": 13.570146560668945,
+    "learning_rate": 7.999999999999999e-05
+  },
+  {
+    "step": 9,
+    "epoch": 0.061016949152542375,
+    "cpu_mem": 1.49815296,
+    "gpu_mem": 4.62441216,
+    "loss": 4.8142,
+    "grad_norm": 12.81284236907959,
+    "learning_rate": 8.999999999999999e-05
+  },
+  {
+    "step": 10,
+    "epoch": 0.06779661016949153,
+    "cpu_mem": 1.498546176,
+    "gpu_mem": 4.62431232,
+    "loss": 3.9045,
+    "grad_norm": 12.29382610321045,
+    "learning_rate": 9.999999999999999e-05
+  },
+  {
+    "step": 11,
+    "epoch": 0.07457627118644068,
+    "cpu_mem": 1.499136,
+    "gpu_mem": 4.624416768,
+    "loss": 2.7592,
+    "grad_norm": 10.226658821105957,
+    "learning_rate": 0.00010999999999999998
+  },
+  {
+    "step": 12,
+    "epoch": 0.08135593220338982,
+    "cpu_mem": 1.499529216,
+    "gpu_mem": 4.62478848,
+    "loss": 2.0805,
+    "grad_norm": 7.122352600097656,
+    "learning_rate": 0.00011999999999999999
+  },
+  {
+    "step": 13,
+    "epoch": 0.08813559322033898,
+    "cpu_mem": 1.50011904,
+    "gpu_mem": 4.624392192,
+    "loss": 1.7604,
+    "grad_norm": 4.6575140953063965,
+    "learning_rate": 0.00013
+  },
+  {
+    "step": 14,
+    "epoch": 0.09491525423728814,
+    "cpu_mem": 1.500512256,
+    "gpu_mem": 4.624369152,
+    "loss": 1.1904,
+    "grad_norm": 2.9982383251190186,
+    "learning_rate": 0.00014
+  },
+  {
+    "step": 15,
+    "epoch": 0.1016949152542373,
+    "cpu_mem": 1.500708864,
+    "gpu_mem": 4.624307712,
+    "loss": 1.148,
+    "grad_norm": 2.116892099380493,
+    "learning_rate": 0.00015
+  },
+  {
+    "step": 16,
+    "epoch": 0.10847457627118644,
+    "cpu_mem": 1.50110208,
+    "gpu_mem": 4.624392192,
+    "loss": 0.9269,
+    "grad_norm": 1.62022066116333,
+    "learning_rate": 0.00015999999999999999
+  },
+  {
+    "step": 17,
+    "epoch": 0.1152542372881356,
+    "cpu_mem": 1.501495296,
+    "gpu_mem": 4.624432128,
+    "loss": 0.8529,
+    "grad_norm": 1.9826555252075195,
+    "learning_rate": 0.00016999999999999999
+  },
+  {
+    "step": 18,
+    "epoch": 0.12203389830508475,
+    "cpu_mem": 1.501691904,
+    "gpu_mem": 4.624495104,
+    "loss": 0.7057,
+    "grad_norm": 0.8324311971664429,
+    "learning_rate": 0.00017999999999999998
+  },
+  {
+    "step": 19,
+    "epoch": 0.1288135593220339,
+    "cpu_mem": 1.50208512,
+    "gpu_mem": 4.624332288,
+    "loss": 0.7184,
+    "grad_norm": 0.6730552911758423,
+    "learning_rate": 0.00018999999999999998
+  },
+  {
+    "step": 20,
+    "epoch": 0.13559322033898305,
+    "cpu_mem": 1.502478336,
+    "gpu_mem": 4.624444416,
+    "loss": 0.7847,
+    "grad_norm": 3.902228593826294,
+    "learning_rate": 0.00019999999999999998
+  },
+  {
+    "step": 21,
+    "epoch": 0.1423728813559322,
+    "cpu_mem": 1.502674944,
+    "gpu_mem": 4.624602624,
+    "loss": 0.6757,
+    "grad_norm": 1.8044828176498413,
+    "learning_rate": 0.00020999999999999998
+  },
+  {
+    "step": 22,
+    "epoch": 0.14915254237288136,
+    "cpu_mem": 1.50306816,
+    "gpu_mem": 4.624495104,
+    "loss": 0.8293,
+    "grad_norm": 3.8413350582122803,
+    "learning_rate": 0.00021999999999999995
+  },
+  {
+    "step": 23,
+    "epoch": 0.15593220338983052,
+    "cpu_mem": 1.503264768,
+    "gpu_mem": 4.624467456,
+    "loss": 0.8114,
+    "grad_norm": 3.3119895458221436,
+    "learning_rate": 0.00023
+  },
+  {
+    "step": 24,
+    "epoch": 0.16271186440677965,
+    "cpu_mem": 1.503657984,
+    "gpu_mem": 4.624524288,
+    "loss": 0.6281,
+    "grad_norm": 1.0927964448928833,
+    "learning_rate": 0.00023999999999999998
+  },
+  {
+    "step": 25,
+    "epoch": 0.1694915254237288,
+    "cpu_mem": 1.503854592,
+    "gpu_mem": 4.624309248,
+    "loss": 0.6868,
+    "grad_norm": 1.6278636455535889,
+    "learning_rate": 0.00025
+  },
+  {
+    "step": 26,
+    "epoch": 0.17627118644067796,
+    "cpu_mem": 1.5040512,
+    "gpu_mem": 4.624364544,
+    "loss": 0.6885,
+    "grad_norm": 1.0237977504730225,
+    "learning_rate": 0.00026
+  },
+  {
+    "step": 27,
+    "epoch": 0.18305084745762712,
+    "cpu_mem": 1.504247808,
+    "gpu_mem": 4.624656384,
+    "loss": 0.6395,
+    "grad_norm": 1.4592667818069458,
+    "learning_rate": 0.00027
+  },
+  {
+    "step": 28,
+    "epoch": 0.18983050847457628,
+    "cpu_mem": 1.504641024,
+    "gpu_mem": 4.62433536,
+    "loss": 0.8711,
+    "grad_norm": 3.5112271308898926,
+    "learning_rate": 0.00028
+  },
+  {
+    "step": 29,
+    "epoch": 0.19661016949152543,
+    "cpu_mem": 1.50503424,
+    "gpu_mem": 4.624399872,
+    "loss": 0.7362,
+    "grad_norm": 2.1286396980285645,
+    "learning_rate": 0.00029
+  },
+  {
+    "step": 30,
+    "epoch": 0.2033898305084746,
+    "cpu_mem": 1.505230848,
+    "gpu_mem": 4.624478208,
+    "loss": 0.7353,
+    "grad_norm": 1.9305531978607178,
+    "learning_rate": 0.0003
+  },
+  {
+    "step": 31,
+    "epoch": 0.21016949152542372,
+    "cpu_mem": 1.505427456,
+    "gpu_mem": 4.6242816,
+    "loss": 0.6234,
+    "grad_norm": 0.9298433661460876,
+    "learning_rate": 0.0002999893794250036
+  },
+  {
+    "step": 32,
+    "epoch": 0.21694915254237288,
+    "cpu_mem": 1.505624064,
+    "gpu_mem": 4.624395264,
+    "loss": 0.6989,
+    "grad_norm": 0.8718124032020569,
+    "learning_rate": 0.00029995751920396937
+  },
+  {
+    "step": 33,
+    "epoch": 0.22372881355932203,
+    "cpu_mem": 1.505820672,
+    "gpu_mem": 4.624633344,
+    "loss": 0.6819,
+    "grad_norm": 0.6011456847190857,
+    "learning_rate": 0.00029990442384854874
+  },
+  {
+    "step": 34,
+    "epoch": 0.2305084745762712,
+    "cpu_mem": 1.50601728,
+    "gpu_mem": 4.62433536,
+    "loss": 0.589,
+    "grad_norm": 0.5000481009483337,
+    "learning_rate": 0.0002998301008774512
+  },
+  {
+    "step": 35,
+    "epoch": 0.23728813559322035,
+    "cpu_mem": 1.506213888,
+    "gpu_mem": 4.624545792,
+    "loss": 0.7987,
+    "grad_norm": 1.9291328191757202,
+    "learning_rate": 0.0002997345608153792
+  },
+  {
+    "step": 36,
+    "epoch": 0.2440677966101695,
+    "cpu_mem": 1.506410496,
+    "gpu_mem": 4.62449664,
+    "loss": 0.7698,
+    "grad_norm": 1.7601375579833984,
+    "learning_rate": 0.000299617817191538
+  },
+  {
+    "step": 37,
+    "epoch": 0.25084745762711863,
+    "cpu_mem": 1.506607104,
+    "gpu_mem": 4.624307712,
+    "loss": 0.6271,
+    "grad_norm": 0.5662587285041809,
+    "learning_rate": 0.0002994798865377198
+  },
+  {
+    "step": 38,
+    "epoch": 0.2576271186440678,
+    "cpu_mem": 1.506803712,
+    "gpu_mem": 4.624555008,
+    "loss": 0.7151,
+    "grad_norm": 0.951106071472168,
+    "learning_rate": 0.0002993207883859627
+  },
+  {
+    "step": 39,
+    "epoch": 0.26440677966101694,
+    "cpu_mem": 1.50700032,
+    "gpu_mem": 4.6249344,
+    "loss": 0.7289,
+    "grad_norm": 1.2373714447021484,
+    "learning_rate": 0.0002991405452657846
+  },
+  {
+    "step": 40,
+    "epoch": 0.2711864406779661,
+    "cpu_mem": 1.507196928,
+    "gpu_mem": 4.62450432,
+    "loss": 0.6913,
+    "grad_norm": 1.2386014461517334,
+    "learning_rate": 0.00029893918270099324
+  },
+  {
+    "step": 41,
+    "epoch": 0.27796610169491526,
+    "cpu_mem": 1.507393536,
+    "gpu_mem": 4.624731648,
+    "loss": 0.6724,
+    "grad_norm": 0.6577208638191223,
+    "learning_rate": 0.00029871672920607153
+  },
+  {
+    "step": 42,
+    "epoch": 0.2847457627118644,
+    "cpu_mem": 1.507590144,
+    "gpu_mem": 4.624628736,
+    "loss": 0.6555,
+    "grad_norm": 0.25326287746429443,
+    "learning_rate": 0.0002984732162821399
+  },
+  {
+    "step": 43,
+    "epoch": 0.29152542372881357,
+    "cpu_mem": 1.507786752,
+    "gpu_mem": 4.62445056,
+    "loss": 0.6011,
+    "grad_norm": 0.29461658000946045,
+    "learning_rate": 0.0002982086784124952
+  },
+  {
+    "step": 44,
+    "epoch": 0.2983050847457627,
+    "cpu_mem": 1.507786752,
+    "gpu_mem": 4.624593408,
+    "loss": 0.6203,
+    "grad_norm": 0.5558605790138245,
+    "learning_rate": 0.00029792315305772796
+  },
+  {
+    "step": 45,
+    "epoch": 0.3050847457627119,
+    "cpu_mem": 1.507786752,
+    "gpu_mem": 4.62437376,
+    "loss": 0.8654,
+    "grad_norm": 2.026333808898926,
+    "learning_rate": 0.0002976166806504174
+  },
+  {
+    "step": 46,
+    "epoch": 0.31186440677966104,
+    "cpu_mem": 1.50798336,
+    "gpu_mem": 4.624616448,
+    "loss": 0.6738,
+    "grad_norm": 1.057332992553711,
+    "learning_rate": 0.00029728930458940595
+  },
+  {
+    "step": 47,
+    "epoch": 0.31864406779661014,
+    "cpu_mem": 1.508179968,
+    "gpu_mem": 4.624339968,
+    "loss": 0.5998,
+    "grad_norm": 0.4258666932582855,
+    "learning_rate": 0.00029694107123365385
+  },
+  {
+    "step": 48,
+    "epoch": 0.3254237288135593,
+    "cpu_mem": 1.508376576,
+    "gpu_mem": 4.624416768,
+    "loss": 0.6132,
+    "grad_norm": 1.7531728744506836,
+    "learning_rate": 0.00029657202989567393
+  },
+  {
+    "step": 49,
+    "epoch": 0.33220338983050846,
+    "cpu_mem": 1.508573184,
+    "gpu_mem": 4.624433664,
+    "loss": 0.6978,
+    "grad_norm": 0.3271001875400543,
+    "learning_rate": 0.00029618223283454893
+  },
+  {
+    "step": 50,
+    "epoch": 0.3389830508474576,
+    "cpu_mem": 1.508573184,
+    "gpu_mem": 4.624372224,
+    "loss": 0.6478,
+    "grad_norm": 1.044687032699585,
+    "learning_rate": 0.00029577173524853123
+  },
+  {
+    "step": 51,
+    "epoch": 0.34576271186440677,
+    "cpu_mem": 1.508769792,
+    "gpu_mem": 4.624376832,
+    "loss": 0.6598,
+    "grad_norm": 1.3430612087249756,
+    "learning_rate": 0.0002953405952672261
+  },
+  {
+    "step": 52,
+    "epoch": 0.3525423728813559,
+    "cpu_mem": 1.508769792,
+    "gpu_mem": 4.624456704,
+    "loss": 0.6723,
+    "grad_norm": 0.5483831763267517,
+    "learning_rate": 0.0002948888739433602
+  },
+  {
+    "step": 53,
+    "epoch": 0.3593220338983051,
+    "cpu_mem": 1.508769792,
+    "gpu_mem": 4.624479744,
+    "loss": 0.573,
+    "grad_norm": 0.3019401729106903,
+    "learning_rate": 0.0002944166352441363
+  },
+  {
+    "step": 54,
+    "epoch": 0.36610169491525424,
+    "cpu_mem": 1.5089664,
+    "gpu_mem": 4.624407552,
+    "loss": 0.7986,
+    "grad_norm": 1.5273728370666504,
+    "learning_rate": 0.0002939239460421746
+  },
+  {
+    "step": 55,
+    "epoch": 0.3728813559322034,
+    "cpu_mem": 1.5089664,
+    "gpu_mem": 4.624677888,
+    "loss": 0.6091,
+    "grad_norm": 0.4779277741909027,
+    "learning_rate": 0.00029341087610604337
+  },
+  {
+    "step": 56,
+    "epoch": 0.37966101694915255,
+    "cpu_mem": 1.509163008,
+    "gpu_mem": 4.624464384,
+    "loss": 0.6733,
+    "grad_norm": 1.1983107328414917,
+    "learning_rate": 0.00029287749809037904
+  },
+  {
+    "step": 57,
+    "epoch": 0.3864406779661017,
+    "cpu_mem": 1.509163008,
+    "gpu_mem": 4.62445824,
+    "loss": 0.6289,
+    "grad_norm": 1.399798035621643,
+    "learning_rate": 0.0002923238875255979
+  },
+  {
+    "step": 58,
+    "epoch": 0.39322033898305087,
+    "cpu_mem": 1.509163008,
+    "gpu_mem": 4.624353792,
+    "loss": 0.6225,
+    "grad_norm": 1.6279141902923584,
+    "learning_rate": 0.00029175012280720024
+  },
+  {
+    "step": 59,
+    "epoch": 0.4,
+    "cpu_mem": 1.509359616,
+    "gpu_mem": 4.624370688,
+    "loss": 0.6482,
+    "grad_norm": 0.6617344617843628,
+    "learning_rate": 0.000291156285184669
+  },
+  {
+    "step": 60,
+    "epoch": 0.4067796610169492,
+    "cpu_mem": 1.509556224,
+    "gpu_mem": 4.624464384,
+    "loss": 0.5742,
+    "grad_norm": 0.8502650856971741,
+    "learning_rate": 0.00029054245874996426
+  },
+  {
+    "step": 61,
+    "epoch": 0.4135593220338983,
+    "cpu_mem": 1.509556224,
+    "gpu_mem": 4.624475136,
+    "loss": 0.5745,
+    "grad_norm": 0.5171988010406494,
+    "learning_rate": 0.0002899087304256151
+  },
+  {
+    "step": 62,
+    "epoch": 0.42033898305084744,
+    "cpu_mem": 1.509556224,
+    "gpu_mem": 4.624462848,
+    "loss": 0.8279,
+    "grad_norm": 1.8439751863479614,
+    "learning_rate": 0.0002892551899524109
+  },
+  {
+    "step": 63,
+    "epoch": 0.4271186440677966,
+    "cpu_mem": 1.509556224,
+    "gpu_mem": 4.624455168,
+    "loss": 0.4921,
+    "grad_norm": 0.5544669032096863,
+    "learning_rate": 0.000288581929876693
+  },
+  {
+    "step": 64,
+    "epoch": 0.43389830508474575,
+    "cpu_mem": 1.509752832,
+    "gpu_mem": 4.624384512,
+    "loss": 0.5735,
+    "grad_norm": 0.6035580635070801,
+    "learning_rate": 0.0002878890455372498
+  },
+  {
+    "step": 65,
+    "epoch": 0.4406779661016949,
+    "cpu_mem": 1.509752832,
+    "gpu_mem": 4.624429056,
+    "loss": 0.5902,
+    "grad_norm": 0.5334798097610474,
+    "learning_rate": 0.0002871766350518159
+  },
+  {
+    "step": 66,
+    "epoch": 0.44745762711864406,
+    "cpu_mem": 1.50994944,
+    "gpu_mem": 4.624622592,
+    "loss": 0.6281,
+    "grad_norm": 1.107169270515442,
+    "learning_rate": 0.00028644479930317775
+  },
+  {
+    "step": 67,
+    "epoch": 0.4542372881355932,
+    "cpu_mem": 1.510146048,
+    "gpu_mem": 4.624332288,
+    "loss": 0.6952,
+    "grad_norm": 0.9980673789978027,
+    "learning_rate": 0.00028569364192488803
+  },
+  {
+    "step": 68,
+    "epoch": 0.4610169491525424,
+    "cpu_mem": 1.510146048,
+    "gpu_mem": 4.624300032,
+    "loss": 0.7016,
+    "grad_norm": 0.480561226606369,
+    "learning_rate": 0.00028492326928659045
+  },
+  {
+    "step": 69,
+    "epoch": 0.46779661016949153,
+    "cpu_mem": 1.510146048,
+    "gpu_mem": 4.62436608,
+    "loss": 0.5867,
+    "grad_norm": 0.3215237855911255,
+    "learning_rate": 0.00028413379047895665
+  },
+  {
+    "step": 70,
+    "epoch": 0.4745762711864407,
+    "cpu_mem": 1.510146048,
+    "gpu_mem": 4.624359936,
+    "loss": 0.5704,
+    "grad_norm": 0.5735366940498352,
+    "learning_rate": 0.0002833253172982385
+  },
+  {
+    "step": 71,
+    "epoch": 0.48135593220338985,
+    "cpu_mem": 1.510146048,
+    "gpu_mem": 4.6245888,
+    "loss": 0.5306,
+    "grad_norm": 0.38639822602272034,
+    "learning_rate": 0.0002824979642304366
+  },
+  {
+    "step": 72,
+    "epoch": 0.488135593220339,
+    "cpu_mem": 1.510146048,
+    "gpu_mem": 4.62458112,
+    "loss": 0.566,
+    "grad_norm": 0.44350624084472656,
+    "learning_rate": 0.0002816518484350883
+  },
+  {
+    "step": 73,
+    "epoch": 0.49491525423728816,
+    "cpu_mem": 1.510342656,
+    "gpu_mem": 4.624547328,
+    "loss": 0.6532,
+    "grad_norm": 1.009778380393982,
+    "learning_rate": 0.0002807870897286772
+  },
+  {
+    "step": 74,
+    "epoch": 0.5016949152542373,
+    "cpu_mem": 1.510342656,
+    "gpu_mem": 4.624407552,
+    "loss": 0.569,
+    "grad_norm": 0.530914306640625,
+    "learning_rate": 0.0002799038105676658
+  },
+  {
+    "step": 75,
+    "epoch": 0.5084745762711864,
+    "cpu_mem": 1.510342656,
+    "gpu_mem": 4.624332288,
+    "loss": 0.5485,
+    "grad_norm": 1.2175366878509521,
+    "learning_rate": 0.000279002136031155
+  },
+  {
+    "step": 76,
+    "epoch": 0.5152542372881356,
+    "cpu_mem": 1.510342656,
+    "gpu_mem": 4.624272384,
+    "loss": 0.6296,
+    "grad_norm": 1.2137932777404785,
+    "learning_rate": 0.00027808219380317216
+  },
+  {
+    "step": 77,
+    "epoch": 0.5220338983050847,
+    "cpu_mem": 1.510342656,
+    "gpu_mem": 4.624346112,
+    "loss": 0.5439,
+    "grad_norm": 1.0348937511444092,
+    "learning_rate": 0.0002771441141545895
+  },
+  {
+    "step": 78,
+    "epoch": 0.5288135593220339,
+    "cpu_mem": 1.510539264,
+    "gpu_mem": 4.624398336,
+    "loss": 0.7299,
+    "grad_norm": 2.1830639839172363,
+    "learning_rate": 0.0002761880299246772
+  },
+  {
+    "step": 79,
+    "epoch": 0.535593220338983,
+    "cpu_mem": 1.510539264,
+    "gpu_mem": 4.624530432,
+    "loss": 0.7099,
+    "grad_norm": 1.5999380350112915,
+    "learning_rate": 0.000275214076502292
+  },
+  {
+    "step": 80,
+    "epoch": 0.5423728813559322,
+    "cpu_mem": 1.510539264,
+    "gpu_mem": 4.624421376,
+    "loss": 0.6399,
+    "grad_norm": 1.355906367301941,
+    "learning_rate": 0.0002742223918067056
+  },
+  {
+    "step": 81,
+    "epoch": 0.5491525423728814,
+    "cpu_mem": 1.510539264,
+    "gpu_mem": 4.624301568,
+    "loss": 0.6151,
+    "grad_norm": 1.0865662097930908,
+    "learning_rate": 0.00027321311626807374
+  },
+  {
+    "step": 82,
+    "epoch": 0.5559322033898305,
+    "cpu_mem": 1.510539264,
+    "gpu_mem": 4.624370688,
+    "loss": 0.6127,
+    "grad_norm": 0.4206037223339081,
+    "learning_rate": 0.0002721863928075503
+  },
+  {
+    "step": 83,
+    "epoch": 0.5627118644067797,
+    "cpu_mem": 1.510539264,
+    "gpu_mem": 4.624470528,
+    "loss": 0.6713,
+    "grad_norm": 0.580093264579773,
+    "learning_rate": 0.000271142366817049
+  },
+  {
+    "step": 84,
+    "epoch": 0.5694915254237288,
+    "cpu_mem": 1.510539264,
+    "gpu_mem": 4.624433664,
+    "loss": 0.7334,
+    "grad_norm": 1.9020066261291504,
+    "learning_rate": 0.00027008118613865406
+  },
+  {
+    "step": 85,
+    "epoch": 0.576271186440678,
+    "cpu_mem": 1.510539264,
+    "gpu_mem": 4.62446592,
+    "loss": 0.7021,
+    "grad_norm": 1.7552995681762695,
+    "learning_rate": 0.00026900300104368524
+  },
+  {
+    "step": 86,
+    "epoch": 0.5830508474576271,
+    "cpu_mem": 1.510735872,
+    "gpu_mem": 4.624416768,
+    "loss": 0.6841,
+    "grad_norm": 1.2111093997955322,
+    "learning_rate": 0.00026790796421141813
+  },
+  {
+    "step": 87,
+    "epoch": 0.5898305084745763,
+    "cpu_mem": 1.510735872,
+    "gpu_mem": 4.624424448,
+    "loss": 0.5347,
+    "grad_norm": 0.44297388195991516,
+    "learning_rate": 0.00026679623070746325
+  },
+  {
+    "step": 88,
+    "epoch": 0.5966101694915255,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.624568832,
+    "loss": 0.5532,
+    "grad_norm": 0.36787739396095276,
+    "learning_rate": 0.0002656679579618081
+  },
+  {
+    "step": 89,
+    "epoch": 0.6033898305084746,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.62435072,
+    "loss": 0.6981,
+    "grad_norm": 1.192941665649414,
+    "learning_rate": 0.0002645233057465235
+  },
+  {
+    "step": 90,
+    "epoch": 0.6101694915254238,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.62440448,
+    "loss": 0.5992,
+    "grad_norm": 0.9635607004165649,
+    "learning_rate": 0.00026336243615313873
+  },
+  {
+    "step": 91,
+    "epoch": 0.6169491525423729,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.624372224,
+    "loss": 0.5141,
+    "grad_norm": 0.5678079128265381,
+    "learning_rate": 0.00026218551356968814
+  },
+  {
+    "step": 92,
+    "epoch": 0.6237288135593221,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.624453632,
+    "loss": 0.7089,
+    "grad_norm": 1.4068773984909058,
+    "learning_rate": 0.00026099270465743254
+  },
+  {
+    "step": 93,
+    "epoch": 0.6305084745762712,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.624257024,
+    "loss": 0.7013,
+    "grad_norm": 0.5640556216239929,
+    "learning_rate": 0.0002597841783272588
+  },
+  {
+    "step": 94,
+    "epoch": 0.6372881355932203,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.624370688,
+    "loss": 0.6064,
+    "grad_norm": 0.9994253516197205,
+    "learning_rate": 0.0002585601057157605
+  },
+  {
+    "step": 95,
+    "epoch": 0.6440677966101694,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.624390656,
+    "loss": 0.5994,
+    "grad_norm": 0.8690909743309021,
+    "learning_rate": 0.00025732066016100394
+  },
+  {
+    "step": 96,
+    "epoch": 0.6508474576271186,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.624429056,
+    "loss": 0.5682,
+    "grad_norm": 1.0052908658981323,
+    "learning_rate": 0.00025606601717798207
+  },
+  {
+    "step": 97,
+    "epoch": 0.6576271186440678,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.624413696,
+    "loss": 0.6538,
+    "grad_norm": 0.5900957584381104,
+    "learning_rate": 0.0002547963544337602
+  },
+  {
+    "step": 98,
+    "epoch": 0.6644067796610169,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.624326144,
+    "loss": 0.5676,
+    "grad_norm": 0.3594554662704468,
+    "learning_rate": 0.0002535118517223168
+  },
+  {
+    "step": 99,
+    "epoch": 0.6711864406779661,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.624275456,
+    "loss": 0.5491,
+    "grad_norm": 0.4727354347705841,
+    "learning_rate": 0.00025221269093908365
+  },
+  {
+    "step": 100,
+    "epoch": 0.6779661016949152,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.624392192,
+    "loss": 0.5544,
+    "grad_norm": 0.36728915572166443,
+    "learning_rate": 0.0002508990560551879
+  },
+  {
+    "step": 101,
+    "epoch": 0.6847457627118644,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.624424448,
+    "loss": 0.574,
+    "grad_norm": 0.6338084936141968,
+    "learning_rate": 0.0002495711330914001
+  },
+  {
+    "step": 102,
+    "epoch": 0.6915254237288135,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.62445824,
+    "loss": 0.609,
+    "grad_norm": 0.601031482219696,
+    "learning_rate": 0.00024822911009179276
+  },
+  {
+    "step": 103,
+    "epoch": 0.6983050847457627,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.624508928,
+    "loss": 0.6165,
+    "grad_norm": 0.6266305446624756,
+    "learning_rate": 0.0002468731770971113
+  },
+  {
+    "step": 104,
+    "epoch": 0.7050847457627119,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.624413696,
+    "loss": 0.5317,
+    "grad_norm": 0.5269937515258789,
+    "learning_rate": 0.0002455035261178632
+  },
+  {
+    "step": 105,
+    "epoch": 0.711864406779661,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.624515072,
+    "loss": 0.6078,
+    "grad_norm": 0.5082085132598877,
+    "learning_rate": 0.0002441203511071278
+  },
+  {
+    "step": 106,
+    "epoch": 0.7186440677966102,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.62446592,
+    "loss": 0.6005,
+    "grad_norm": 0.8072856664657593,
+    "learning_rate": 0.00024272384793309077
+  },
+  {
+    "step": 107,
+    "epoch": 0.7254237288135593,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.624353792,
+    "loss": 0.524,
+    "grad_norm": 0.751579225063324,
+    "learning_rate": 0.00024131421435130807
+  },
+  {
+    "step": 108,
+    "epoch": 0.7322033898305085,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.624538112,
+    "loss": 0.5979,
+    "grad_norm": 0.6362266540527344,
+    "learning_rate": 0.00023989164997670202
+  },
+  {
+    "step": 109,
+    "epoch": 0.7389830508474576,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.624392192,
+    "loss": 0.6221,
+    "grad_norm": 0.570347011089325,
+    "learning_rate": 0.0002384563562552943
+  },
+  {
+    "step": 110,
+    "epoch": 0.7457627118644068,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.624395264,
+    "loss": 0.5563,
+    "grad_norm": 0.6132338047027588,
+    "learning_rate": 0.0002370085364356797
+  },
+  {
+    "step": 111,
+    "epoch": 0.752542372881356,
+    "cpu_mem": 1.51093248,
+    "gpu_mem": 4.624364544,
+    "loss": 0.5369,
+    "grad_norm": 0.6124526858329773,
+    "learning_rate": 0.0002355483955402446
+  },
+  {
+    "step": 112,
+    "epoch": 0.7593220338983051,
+    "cpu_mem": 1.511129088,
+    "gpu_mem": 4.624410624,
+    "loss": 0.4849,
+    "grad_norm": 0.37720710039138794,
+    "learning_rate": 0.00023407614033613407
+  },
+  {
+    "step": 113,
+    "epoch": 0.7661016949152543,
+    "cpu_mem": 1.511129088,
+    "gpu_mem": 4.624401408,
+    "loss": 0.5588,
+    "grad_norm": 0.5072793364524841,
+    "learning_rate": 0.0002325919793059723
+  },
+  {
+    "step": 114,
+    "epoch": 0.7728813559322034,
+    "cpu_mem": 1.511129088,
+    "gpu_mem": 4.624382976,
+    "loss": 0.5045,
+    "grad_norm": 0.44536927342414856,
+    "learning_rate": 0.00023109612261833963
+  },
+  {
+    "step": 115,
+    "epoch": 0.7796610169491526,
+    "cpu_mem": 1.511129088,
+    "gpu_mem": 4.62445824,
+    "loss": 0.5212,
+    "grad_norm": 0.38146018981933594,
+    "learning_rate": 0.0002295887820980112
+  },
+  {
+    "step": 116,
+    "epoch": 0.7864406779661017,
+    "cpu_mem": 1.511129088,
+    "gpu_mem": 4.624378368,
+    "loss": 0.5656,
+    "grad_norm": 0.6866984367370605,
+    "learning_rate": 0.0002280701711959608
+  },
+  {
+    "step": 117,
+    "epoch": 0.7932203389830509,
+    "cpu_mem": 1.511129088,
+    "gpu_mem": 4.624269312,
+    "loss": 0.5171,
+    "grad_norm": 0.45652973651885986,
+    "learning_rate": 0.00022654050495913495
+  },
+  {
+    "step": 118,
+    "epoch": 0.8,
+    "cpu_mem": 1.511129088,
+    "gpu_mem": 4.624507392,
+    "loss": 0.5348,
+    "grad_norm": 0.5435035824775696,
+    "learning_rate": 0.000225
+  },
+  {
+    "step": 119,
+    "epoch": 0.8067796610169492,
+    "cpu_mem": 1.511129088,
+    "gpu_mem": 4.624677888,
+    "loss": 0.4785,
+    "grad_norm": 0.4610842168331146,
+    "learning_rate": 0.00022344887446586865
+  },
+  {
+    "step": 120,
+    "epoch": 0.8135593220338984,
+    "cpu_mem": 1.511129088,
+    "gpu_mem": 4.624410624,
+    "loss": 0.5077,
+    "grad_norm": 0.5721786618232727,
+    "learning_rate": 0.00022188734800800852
+  },
+  {
+    "step": 121,
+    "epoch": 0.8203389830508474,
+    "cpu_mem": 1.511129088,
+    "gpu_mem": 4.624438272,
+    "loss": 0.5366,
+    "grad_norm": 0.4264529049396515,
+    "learning_rate": 0.00022031564175053754
+  },
+  {
+    "step": 122,
+    "epoch": 0.8271186440677966,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.62448896,
+    "loss": 0.4827,
+    "grad_norm": 0.4972001910209656,
+    "learning_rate": 0.00021873397825911153
+  },
+  {
+    "step": 123,
+    "epoch": 0.8338983050847457,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.624298496,
+    "loss": 0.5555,
+    "grad_norm": 1.0725321769714355,
+    "learning_rate": 0.00021714258150940685
+  },
+  {
+    "step": 124,
+    "epoch": 0.8406779661016949,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.624740864,
+    "loss": 0.476,
+    "grad_norm": 0.7154480218887329,
+    "learning_rate": 0.0002155416768554039
+  },
+  {
+    "step": 125,
+    "epoch": 0.847457627118644,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.624467456,
+    "loss": 0.5311,
+    "grad_norm": 0.47886914014816284,
+    "learning_rate": 0.00021393149099747523
+  },
+  {
+    "step": 126,
+    "epoch": 0.8542372881355932,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.62435072,
+    "loss": 0.5198,
+    "grad_norm": 0.5867719650268555,
+    "learning_rate": 0.00021231225195028297
+  },
+  {
+    "step": 127,
+    "epoch": 0.8610169491525423,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.624790016,
+    "loss": 0.5319,
+    "grad_norm": 0.6425331234931946,
+    "learning_rate": 0.00021068418901049025
+  },
+  {
+    "step": 128,
+    "epoch": 0.8677966101694915,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.62456576,
+    "loss": 0.4434,
+    "grad_norm": 0.6586298942565918,
+    "learning_rate": 0.0002090475327242912
+  },
+  {
+    "step": 129,
+    "epoch": 0.8745762711864407,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.624605696,
+    "loss": 0.5039,
+    "grad_norm": 0.5399751663208008,
+    "learning_rate": 0.00020740251485476345
+  },
+  {
+    "step": 130,
+    "epoch": 0.8813559322033898,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.624387584,
+    "loss": 0.6539,
+    "grad_norm": 0.5986157655715942,
+    "learning_rate": 0.0002057493683490491
+  },
+  {
+    "step": 131,
+    "epoch": 0.888135593220339,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.624516608,
+    "loss": 0.4843,
+    "grad_norm": 0.49613156914711,
+    "learning_rate": 0.00020408832730536746
+  },
+  {
+    "step": 132,
+    "epoch": 0.8949152542372881,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.624598016,
+    "loss": 0.4115,
+    "grad_norm": 0.8342369794845581,
+    "learning_rate": 0.00020241962693986476
+  },
+  {
+    "step": 133,
+    "epoch": 0.9016949152542373,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.62438144,
+    "loss": 0.4729,
+    "grad_norm": 0.633710503578186,
+    "learning_rate": 0.0002007435035533061
+  },
+  {
+    "step": 134,
+    "epoch": 0.9084745762711864,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.624515072,
+    "loss": 0.449,
+    "grad_norm": 1.1560707092285156,
+    "learning_rate": 0.00019906019449761325
+  },
+  {
+    "step": 135,
+    "epoch": 0.9152542372881356,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.624538112,
+    "loss": 0.501,
+    "grad_norm": 0.5545433163642883,
+    "learning_rate": 0.00019736993814225374
+  },
+  {
+    "step": 136,
+    "epoch": 0.9220338983050848,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.624375296,
+    "loss": 0.4472,
+    "grad_norm": 0.6953291893005371,
+    "learning_rate": 0.00019567297384048604
+  },
+  {
+    "step": 137,
+    "epoch": 0.9288135593220339,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.624255488,
+    "loss": 0.5768,
+    "grad_norm": 1.2485830783843994,
+    "learning_rate": 0.0001939695418954653
+  },
+  {
+    "step": 138,
+    "epoch": 0.9355932203389831,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.624436736,
+    "loss": 0.5167,
+    "grad_norm": 1.273888349533081,
+    "learning_rate": 0.00019225988352621445
+  },
+  {
+    "step": 139,
+    "epoch": 0.9423728813559322,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.62433536,
+    "loss": 0.4855,
+    "grad_norm": 1.0167266130447388,
+    "learning_rate": 0.00019054424083346592
+  },
+  {
+    "step": 140,
+    "epoch": 0.9491525423728814,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.624387584,
+    "loss": 0.4915,
+    "grad_norm": 0.6369791626930237,
+    "learning_rate": 0.0001888228567653781
+  },
+  {
+    "step": 141,
+    "epoch": 0.9559322033898305,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.62441984,
+    "loss": 0.5621,
+    "grad_norm": 0.7908885478973389,
+    "learning_rate": 0.0001870959750831323
+  },
+  {
+    "step": 142,
+    "epoch": 0.9627118644067797,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.624559616,
+    "loss": 0.5112,
+    "grad_norm": 1.217617154121399,
+    "learning_rate": 0.0001853638403264141
+  },
+  {
+    "step": 143,
+    "epoch": 0.9694915254237289,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.62454272,
+    "loss": 0.6221,
+    "grad_norm": 1.0842041969299316,
+    "learning_rate": 0.00018362669777878453
+  },
+  {
+    "step": 144,
+    "epoch": 0.976271186440678,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.62473472,
+    "loss": 0.504,
+    "grad_norm": 0.8124033212661743,
+    "learning_rate": 0.00018188479343294648
+  },
+  {
+    "step": 145,
+    "epoch": 0.9830508474576272,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.624445952,
+    "loss": 0.512,
+    "grad_norm": 1.1542987823486328,
+    "learning_rate": 0.0001801383739559098
+  },
+  {
+    "step": 146,
+    "epoch": 0.9898305084745763,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.62448128,
+    "loss": 0.4851,
+    "grad_norm": 0.4650062024593353,
+    "learning_rate": 0.0001783876866540615
+  },
+  {
+    "step": 147,
+    "epoch": 0.9966101694915255,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.624379904,
+    "loss": 0.5265,
+    "grad_norm": 0.7359176278114319,
+    "learning_rate": 0.00017663297943814552
+  },
+  {
+    "step": 148,
+    "epoch": 1.0033898305084745,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676583936,
+    "loss": 0.7821,
+    "grad_norm": 1.846174716949463,
+    "learning_rate": 0.0001748745007881561
+  },
+  {
+    "step": 149,
+    "epoch": 1.0101694915254238,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676519424,
+    "loss": 0.4444,
+    "grad_norm": 0.9513587951660156,
+    "learning_rate": 0.00017311249971815185
+  },
+  {
+    "step": 150,
+    "epoch": 1.0169491525423728,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676356608,
+    "loss": 0.4795,
+    "grad_norm": 0.9790411591529846,
+    "learning_rate": 0.00017134722574099276
+  },
+  {
+    "step": 151,
+    "epoch": 1.023728813559322,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.6764288,
+    "loss": 0.6015,
+    "grad_norm": 1.2339156866073608,
+    "learning_rate": 0.00016957892883300775
+  },
+  {
+    "step": 152,
+    "epoch": 1.0305084745762711,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676464128,
+    "loss": 0.4506,
+    "grad_norm": 0.6777993440628052,
+    "learning_rate": 0.00016780785939859576
+  },
+  {
+    "step": 153,
+    "epoch": 1.0372881355932204,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676488704,
+    "loss": 0.5975,
+    "grad_norm": 0.7841456532478333,
+    "learning_rate": 0.00016603426823476693
+  },
+  {
+    "step": 154,
+    "epoch": 1.0440677966101695,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676450304,
+    "loss": 0.4571,
+    "grad_norm": 0.8193861842155457,
+    "learning_rate": 0.00016425840649562736
+  },
+  {
+    "step": 155,
+    "epoch": 1.0508474576271187,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676671488,
+    "loss": 0.4829,
+    "grad_norm": 0.938800573348999,
+    "learning_rate": 0.00016248052565681436
+  },
+  {
+    "step": 156,
+    "epoch": 1.0576271186440678,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676579328,
+    "loss": 0.5072,
+    "grad_norm": 1.5463695526123047,
+    "learning_rate": 0.00016070087747988482
+  },
+  {
+    "step": 157,
+    "epoch": 1.064406779661017,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676485632,
+    "loss": 0.4785,
+    "grad_norm": 0.6637791395187378,
+    "learning_rate": 0.00015891971397666464
+  },
+  {
+    "step": 158,
+    "epoch": 1.071186440677966,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676411904,
+    "loss": 0.4213,
+    "grad_norm": 0.8457335829734802,
+    "learning_rate": 0.00015713728737356137
+  },
+  {
+    "step": 159,
+    "epoch": 1.0779661016949154,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676760576,
+    "loss": 0.3644,
+    "grad_norm": 0.5662264823913574,
+    "learning_rate": 0.00015535385007584706
+  },
+  {
+    "step": 160,
+    "epoch": 1.0847457627118644,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676355072,
+    "loss": 0.4047,
+    "grad_norm": 0.6654481291770935,
+    "learning_rate": 0.0001535696546319161
+  },
+  {
+    "step": 161,
+    "epoch": 1.0915254237288137,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676301312,
+    "loss": 0.4,
+    "grad_norm": 0.5183922052383423,
+    "learning_rate": 0.00015178495369752213
+  },
+  {
+    "step": 162,
+    "epoch": 1.0983050847457627,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.677076992,
+    "loss": 0.4024,
+    "grad_norm": 0.5879871845245361,
+    "learning_rate": 0.00015
+  },
+  {
+    "step": 163,
+    "epoch": 1.1050847457627118,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676553216,
+    "loss": 0.537,
+    "grad_norm": 1.4586920738220215,
+    "learning_rate": 0.00014821504630247785
+  },
+  {
+    "step": 164,
+    "epoch": 1.111864406779661,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676465664,
+    "loss": 0.5474,
+    "grad_norm": 1.072028398513794,
+    "learning_rate": 0.00014643034536808387
+  },
+  {
+    "step": 165,
+    "epoch": 1.11864406779661,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676414976,
+    "loss": 0.4617,
+    "grad_norm": 1.0854562520980835,
+    "learning_rate": 0.00014464614992415294
+  },
+  {
+    "step": 166,
+    "epoch": 1.1254237288135593,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676510208,
+    "loss": 0.418,
+    "grad_norm": 1.0139094591140747,
+    "learning_rate": 0.00014286271262643866
+  },
+  {
+    "step": 167,
+    "epoch": 1.1322033898305084,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676427264,
+    "loss": 0.4613,
+    "grad_norm": 0.7832831144332886,
+    "learning_rate": 0.00014108028602333536
+  },
+  {
+    "step": 168,
+    "epoch": 1.1389830508474577,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676445696,
+    "loss": 0.4621,
+    "grad_norm": 0.6947904825210571,
+    "learning_rate": 0.00013929912252011516
+  },
+  {
+    "step": 169,
+    "epoch": 1.1457627118644067,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676533248,
+    "loss": 0.4963,
+    "grad_norm": 1.5868031978607178,
+    "learning_rate": 0.00013751947434318564
+  },
+  {
+    "step": 170,
+    "epoch": 1.152542372881356,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676418048,
+    "loss": 0.5307,
+    "grad_norm": 1.127265214920044,
+    "learning_rate": 0.00013574159350437261
+  },
+  {
+    "step": 171,
+    "epoch": 1.159322033898305,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676481024,
+    "loss": 0.5001,
+    "grad_norm": 0.7554064989089966,
+    "learning_rate": 0.0001339657317652331
+  },
+  {
+    "step": 172,
+    "epoch": 1.1661016949152543,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676388864,
+    "loss": 0.3924,
+    "grad_norm": 0.740827202796936,
+    "learning_rate": 0.00013219214060140424
+  },
+  {
+    "step": 173,
+    "epoch": 1.1728813559322033,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676688384,
+    "loss": 0.4615,
+    "grad_norm": 0.6514274477958679,
+    "learning_rate": 0.00013042107116699228
+  },
+  {
+    "step": 174,
+    "epoch": 1.1796610169491526,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676411904,
+    "loss": 0.4797,
+    "grad_norm": 0.7258147597312927,
+    "learning_rate": 0.00012865277425900724
+  },
+  {
+    "step": 175,
+    "epoch": 1.1864406779661016,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676378112,
+    "loss": 0.4992,
+    "grad_norm": 0.7582775950431824,
+    "learning_rate": 0.00012688750028184818
+  },
+  {
+    "step": 176,
+    "epoch": 1.193220338983051,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676516352,
+    "loss": 0.3651,
+    "grad_norm": 0.7316360473632812,
+    "learning_rate": 0.0001251254992118439
+  },
+  {
+    "step": 177,
+    "epoch": 1.2,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676614656,
+    "loss": 0.4881,
+    "grad_norm": 0.7436537146568298,
+    "learning_rate": 0.00012336702056185453
+  },
+  {
+    "step": 178,
+    "epoch": 1.2067796610169492,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676361216,
+    "loss": 0.4471,
+    "grad_norm": 0.9715722799301147,
+    "learning_rate": 0.00012161231334593851
+  },
+  {
+    "step": 179,
+    "epoch": 1.2135593220338983,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676461056,
+    "loss": 0.4749,
+    "grad_norm": 0.6974865794181824,
+    "learning_rate": 0.00011986162604409015
+  },
+  {
+    "step": 180,
+    "epoch": 1.2203389830508475,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676433408,
+    "loss": 0.481,
+    "grad_norm": 1.586224913597107,
+    "learning_rate": 0.00011811520656705348
+  },
+  {
+    "step": 181,
+    "epoch": 1.2271186440677966,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676370432,
+    "loss": 0.3847,
+    "grad_norm": 0.6357300877571106,
+    "learning_rate": 0.00011637330222121543
+  },
+  {
+    "step": 182,
+    "epoch": 1.2338983050847459,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676588544,
+    "loss": 0.4714,
+    "grad_norm": 1.1362266540527344,
+    "learning_rate": 0.00011463615967358588
+  },
+  {
+    "step": 183,
+    "epoch": 1.240677966101695,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676485632,
+    "loss": 0.4662,
+    "grad_norm": 0.8635218143463135,
+    "learning_rate": 0.00011290402491686766
+  },
+  {
+    "step": 184,
+    "epoch": 1.2474576271186442,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676433408,
+    "loss": 0.4476,
+    "grad_norm": 0.8836385607719421,
+    "learning_rate": 0.00011117714323462186
+  },
+  {
+    "step": 185,
+    "epoch": 1.2542372881355932,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676411904,
+    "loss": 0.4304,
+    "grad_norm": 0.7257511615753174,
+    "learning_rate": 0.00010945575916653407
+  },
+  {
+    "step": 186,
+    "epoch": 1.2610169491525425,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.67642112,
+    "loss": 0.3666,
+    "grad_norm": 0.6456829905509949,
+    "learning_rate": 0.00010774011647378553
+  },
+  {
+    "step": 187,
+    "epoch": 1.2677966101694915,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676353536,
+    "loss": 0.5261,
+    "grad_norm": 0.7594860792160034,
+    "learning_rate": 0.00010603045810453468
+  },
+  {
+    "step": 188,
+    "epoch": 1.2745762711864406,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676516352,
+    "loss": 0.3928,
+    "grad_norm": 0.6978077292442322,
+    "learning_rate": 0.00010432702615951396
+  },
+  {
+    "step": 189,
+    "epoch": 1.2813559322033898,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676385792,
+    "loss": 0.5247,
+    "grad_norm": 0.7300601601600647,
+    "learning_rate": 0.00010263006185774627
+  },
+  {
+    "step": 190,
+    "epoch": 1.288135593220339,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.6765056,
+    "loss": 0.4131,
+    "grad_norm": 0.7177391648292542,
+    "learning_rate": 0.00010093980550238675
+  },
+  {
+    "step": 191,
+    "epoch": 1.2949152542372881,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676324352,
+    "loss": 0.3503,
+    "grad_norm": 0.8216970562934875,
+    "learning_rate": 9.925649644669391e-05
+  },
+  {
+    "step": 192,
+    "epoch": 1.3016949152542372,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676456448,
+    "loss": 0.3357,
+    "grad_norm": 0.9710990786552429,
+    "learning_rate": 9.758037306013526e-05
+  },
+  {
+    "step": 193,
+    "epoch": 1.3084745762711865,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676430336,
+    "loss": 0.4152,
+    "grad_norm": 0.6677502393722534,
+    "learning_rate": 9.591167269463255e-05
+  },
+  {
+    "step": 194,
+    "epoch": 1.3152542372881357,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676396544,
+    "loss": 0.3694,
+    "grad_norm": 0.6760179400444031,
+    "learning_rate": 9.425063165095088e-05
+  },
+  {
+    "step": 195,
+    "epoch": 1.3220338983050848,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676500992,
+    "loss": 0.3101,
+    "grad_norm": 0.7037206888198853,
+    "learning_rate": 9.259748514523653e-05
+  },
+  {
+    "step": 196,
+    "epoch": 1.3288135593220338,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676496384,
+    "loss": 0.4392,
+    "grad_norm": 0.7807204723358154,
+    "learning_rate": 9.095246727570879e-05
+  },
+  {
+    "step": 197,
+    "epoch": 1.335593220338983,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676355072,
+    "loss": 0.4013,
+    "grad_norm": 1.1068989038467407,
+    "learning_rate": 8.931581098950973e-05
+  },
+  {
+    "step": 198,
+    "epoch": 1.3423728813559321,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676547072,
+    "loss": 0.4122,
+    "grad_norm": 0.7010363340377808,
+    "learning_rate": 8.768774804971705e-05
+  },
+  {
+    "step": 199,
+    "epoch": 1.3491525423728814,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.67639808,
+    "loss": 0.4062,
+    "grad_norm": 0.7628048658370972,
+    "learning_rate": 8.606850900252478e-05
+  },
+  {
+    "step": 200,
+    "epoch": 1.3559322033898304,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676500992,
+    "loss": 0.3826,
+    "grad_norm": 0.933638334274292,
+    "learning_rate": 8.445832314459608e-05
+  },
+  {
+    "step": 201,
+    "epoch": 1.3627118644067797,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676703744,
+    "loss": 0.3256,
+    "grad_norm": 0.9295195937156677,
+    "learning_rate": 8.285741849059311e-05
+  },
+  {
+    "step": 202,
+    "epoch": 1.3694915254237288,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.6765056,
+    "loss": 0.4441,
+    "grad_norm": 0.8123700022697449,
+    "learning_rate": 8.126602174088843e-05
+  },
+  {
+    "step": 203,
+    "epoch": 1.376271186440678,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676391936,
+    "loss": 0.4034,
+    "grad_norm": 0.8742663860321045,
+    "learning_rate": 7.968435824946242e-05
+  },
+  {
+    "step": 204,
+    "epoch": 1.383050847457627,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.67640576,
+    "loss": 0.3757,
+    "grad_norm": 0.953644335269928,
+    "learning_rate": 7.811265199199152e-05
+  },
+  {
+    "step": 205,
+    "epoch": 1.3898305084745763,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676450304,
+    "loss": 0.4845,
+    "grad_norm": 1.084182620048523,
+    "learning_rate": 7.655112553413135e-05
+  },
+  {
+    "step": 206,
+    "epoch": 1.3966101694915254,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676391936,
+    "loss": 0.342,
+    "grad_norm": 0.9180456399917603,
+    "learning_rate": 7.500000000000002e-05
+  },
+  {
+    "step": 207,
+    "epoch": 1.4033898305084747,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676625408,
+    "loss": 0.4528,
+    "grad_norm": 1.341009497642517,
+    "learning_rate": 7.345949504086507e-05
+  },
+  {
+    "step": 208,
+    "epoch": 1.4101694915254237,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676656128,
+    "loss": 0.4201,
+    "grad_norm": 1.7378958463668823,
+    "learning_rate": 7.192982880403917e-05
+  },
+  {
+    "step": 209,
+    "epoch": 1.4169491525423727,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.6765824,
+    "loss": 0.4907,
+    "grad_norm": 0.8371671438217163,
+    "learning_rate": 7.041121790198881e-05
+  },
+  {
+    "step": 210,
+    "epoch": 1.423728813559322,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676470272,
+    "loss": 0.4671,
+    "grad_norm": 1.0010132789611816,
+    "learning_rate": 6.890387738166041e-05
+  },
+  {
+    "step": 211,
+    "epoch": 1.4305084745762713,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676419584,
+    "loss": 0.3792,
+    "grad_norm": 1.223456621170044,
+    "learning_rate": 6.740802069402771e-05
+  },
+  {
+    "step": 212,
+    "epoch": 1.4372881355932203,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676388864,
+    "loss": 0.4226,
+    "grad_norm": 0.8292936086654663,
+    "learning_rate": 6.592385966386588e-05
+  },
+  {
+    "step": 213,
+    "epoch": 1.4440677966101694,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676411904,
+    "loss": 0.57,
+    "grad_norm": 1.3903371095657349,
+    "learning_rate": 6.445160445975536e-05
+  },
+  {
+    "step": 214,
+    "epoch": 1.4508474576271186,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676494848,
+    "loss": 0.3324,
+    "grad_norm": 0.8603353500366211,
+    "learning_rate": 6.299146356432029e-05
+  },
+  {
+    "step": 215,
+    "epoch": 1.457627118644068,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676422656,
+    "loss": 0.5364,
+    "grad_norm": 1.7915467023849487,
+    "learning_rate": 6.154364374470568e-05
+  },
+  {
+    "step": 216,
+    "epoch": 1.464406779661017,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676588544,
+    "loss": 0.4089,
+    "grad_norm": 0.8353825807571411,
+    "learning_rate": 6.010835002329795e-05
+  },
+  {
+    "step": 217,
+    "epoch": 1.471186440677966,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676430336,
+    "loss": 0.4557,
+    "grad_norm": 1.2738498449325562,
+    "learning_rate": 5.8685785648691894e-05
+  },
+  {
+    "step": 218,
+    "epoch": 1.4779661016949153,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676407296,
+    "loss": 0.4089,
+    "grad_norm": 0.9500507116317749,
+    "learning_rate": 5.72761520669092e-05
+  },
+  {
+    "step": 219,
+    "epoch": 1.4847457627118645,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676533248,
+    "loss": 0.4562,
+    "grad_norm": 0.8485879898071289,
+    "learning_rate": 5.587964889287218e-05
+  },
+  {
+    "step": 220,
+    "epoch": 1.4915254237288136,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.67656704,
+    "loss": 0.4296,
+    "grad_norm": 0.8201168179512024,
+    "learning_rate": 5.449647388213678e-05
+  },
+  {
+    "step": 221,
+    "epoch": 1.4983050847457626,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676434944,
+    "loss": 0.4364,
+    "grad_norm": 0.8443863391876221,
+    "learning_rate": 5.312682290288869e-05
+  },
+  {
+    "step": 222,
+    "epoch": 1.505084745762712,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676571648,
+    "loss": 0.4266,
+    "grad_norm": 0.8384501338005066,
+    "learning_rate": 5.1770889908207245e-05
+  },
+  {
+    "step": 223,
+    "epoch": 1.5118644067796612,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676485632,
+    "loss": 0.3531,
+    "grad_norm": 1.1277318000793457,
+    "learning_rate": 5.0428866908599864e-05
+  },
+  {
+    "step": 224,
+    "epoch": 1.5186440677966102,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676450304,
+    "loss": 0.3186,
+    "grad_norm": 0.6930432319641113,
+    "learning_rate": 4.9100943944812114e-05
+  },
+  {
+    "step": 225,
+    "epoch": 1.5254237288135593,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676414976,
+    "loss": 0.3449,
+    "grad_norm": 0.739639401435852,
+    "learning_rate": 4.778730906091632e-05
+  },
+  {
+    "step": 226,
+    "epoch": 1.5322033898305085,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676563968,
+    "loss": 0.4041,
+    "grad_norm": 0.9897132515907288,
+    "learning_rate": 4.648814827768322e-05
+  },
+  {
+    "step": 227,
+    "epoch": 1.5389830508474578,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676453376,
+    "loss": 0.4365,
+    "grad_norm": 0.8921437859535217,
+    "learning_rate": 4.5203645566239816e-05
+  },
+  {
+    "step": 228,
+    "epoch": 1.5457627118644068,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.67639808,
+    "loss": 0.4615,
+    "grad_norm": 0.6736249327659607,
+    "learning_rate": 4.3933982822017876e-05
+  },
+  {
+    "step": 229,
+    "epoch": 1.5525423728813559,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676339712,
+    "loss": 0.3326,
+    "grad_norm": 0.6789183020591736,
+    "learning_rate": 4.267933983899601e-05
+  },
+  {
+    "step": 230,
+    "epoch": 1.559322033898305,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676396544,
+    "loss": 0.3678,
+    "grad_norm": 0.7079548835754395,
+    "learning_rate": 4.143989428423947e-05
+  },
+  {
+    "step": 231,
+    "epoch": 1.5661016949152542,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.67667456,
+    "loss": 0.3894,
+    "grad_norm": 0.9309094548225403,
+    "learning_rate": 4.0215821672741213e-05
+  },
+  {
+    "step": 232,
+    "epoch": 1.5728813559322035,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.67639808,
+    "loss": 0.4916,
+    "grad_norm": 1.011915922164917,
+    "learning_rate": 3.900729534256745e-05
+  },
+  {
+    "step": 233,
+    "epoch": 1.5796610169491525,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676711424,
+    "loss": 0.3778,
+    "grad_norm": 0.9273942112922668,
+    "learning_rate": 3.781448643031187e-05
+  },
+  {
+    "step": 234,
+    "epoch": 1.5864406779661016,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676587008,
+    "loss": 0.408,
+    "grad_norm": 1.0228288173675537,
+    "learning_rate": 3.663756384686127e-05
+  },
+  {
+    "step": 235,
+    "epoch": 1.5932203389830508,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676342784,
+    "loss": 0.3153,
+    "grad_norm": 0.8042024970054626,
+    "learning_rate": 3.547669425347647e-05
+  },
+  {
+    "step": 236,
+    "epoch": 1.6,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676402688,
+    "loss": 0.4871,
+    "grad_norm": 1.011894702911377,
+    "learning_rate": 3.433204203819185e-05
+  },
+  {
+    "step": 237,
+    "epoch": 1.6067796610169491,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676464128,
+    "loss": 0.4899,
+    "grad_norm": 1.044426441192627,
+    "learning_rate": 3.3203769292536764e-05
+  },
+  {
+    "step": 238,
+    "epoch": 1.6135593220338982,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676465664,
+    "loss": 0.4463,
+    "grad_norm": 1.1736929416656494,
+    "learning_rate": 3.209203578858191e-05
+  },
+  {
+    "step": 239,
+    "epoch": 1.6203389830508474,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676719104,
+    "loss": 0.4268,
+    "grad_norm": 0.8223713040351868,
+    "learning_rate": 3.099699895631474e-05
+  },
+  {
+    "step": 240,
+    "epoch": 1.6271186440677967,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676368896,
+    "loss": 0.5236,
+    "grad_norm": 0.9299846887588501,
+    "learning_rate": 2.9918813861345952e-05
+  },
+  {
+    "step": 241,
+    "epoch": 1.6338983050847458,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676665344,
+    "loss": 0.4318,
+    "grad_norm": 0.9577910900115967,
+    "learning_rate": 2.885763318295102e-05
+  },
+  {
+    "step": 242,
+    "epoch": 1.6406779661016948,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676527104,
+    "loss": 0.4158,
+    "grad_norm": 0.8543640971183777,
+    "learning_rate": 2.781360719244964e-05
+  },
+  {
+    "step": 243,
+    "epoch": 1.647457627118644,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676379648,
+    "loss": 0.567,
+    "grad_norm": 1.039746642112732,
+    "learning_rate": 2.6786883731926306e-05
+  },
+  {
+    "step": 244,
+    "epoch": 1.6542372881355933,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676519424,
+    "loss": 0.342,
+    "grad_norm": 0.99267578125,
+    "learning_rate": 2.5777608193294396e-05
+  },
+  {
+    "step": 245,
+    "epoch": 1.6610169491525424,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.67639808,
+    "loss": 0.4102,
+    "grad_norm": 0.6880216002464294,
+    "learning_rate": 2.4785923497707956e-05
+  },
+  {
+    "step": 246,
+    "epoch": 1.6677966101694914,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676491776,
+    "loss": 0.4772,
+    "grad_norm": 0.7552581429481506,
+    "learning_rate": 2.38119700753228e-05
+  },
+  {
+    "step": 247,
+    "epoch": 1.6745762711864407,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676510208,
+    "loss": 0.3039,
+    "grad_norm": 0.5652927756309509,
+    "learning_rate": 2.285588584541047e-05
+  },
+  {
+    "step": 248,
+    "epoch": 1.68135593220339,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676462592,
+    "loss": 0.3855,
+    "grad_norm": 0.6893491744995117,
+    "learning_rate": 2.1917806196827792e-05
+  },
+  {
+    "step": 249,
+    "epoch": 1.688135593220339,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676368896,
+    "loss": 0.2917,
+    "grad_norm": 0.860058069229126,
+    "learning_rate": 2.0997863968844914e-05
+  },
+  {
+    "step": 250,
+    "epoch": 1.694915254237288,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676461056,
+    "loss": 0.4473,
+    "grad_norm": 0.9282390475273132,
+    "learning_rate": 2.009618943233419e-05
+  },
+  {
+    "step": 251,
+    "epoch": 1.7016949152542373,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676373504,
+    "loss": 0.2981,
+    "grad_norm": 0.6758578419685364,
+    "learning_rate": 1.921291027132278e-05
+  },
+  {
+    "step": 252,
+    "epoch": 1.7084745762711866,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676416512,
+    "loss": 0.45,
+    "grad_norm": 0.8535842299461365,
+    "learning_rate": 1.834815156491165e-05
+  },
+  {
+    "step": 253,
+    "epoch": 1.7152542372881356,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676610048,
+    "loss": 0.4595,
+    "grad_norm": 0.8331772089004517,
+    "learning_rate": 1.750203576956341e-05
+  },
+  {
+    "step": 254,
+    "epoch": 1.7220338983050847,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.67640576,
+    "loss": 0.3657,
+    "grad_norm": 0.8118672370910645,
+    "learning_rate": 1.6674682701761493e-05
+  },
+  {
+    "step": 255,
+    "epoch": 1.7288135593220337,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676562432,
+    "loss": 0.5165,
+    "grad_norm": 1.3498203754425049,
+    "learning_rate": 1.5866209521043304e-05
+  },
+  {
+    "step": 256,
+    "epoch": 1.735593220338983,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676388864,
+    "loss": 0.3667,
+    "grad_norm": 0.6675718426704407,
+    "learning_rate": 1.5076730713409523e-05
+  },
+  {
+    "step": 257,
+    "epoch": 1.7423728813559323,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676802048,
+    "loss": 0.47,
+    "grad_norm": 0.7657194137573242,
+    "learning_rate": 1.4306358075111923e-05
+  },
+  {
+    "step": 258,
+    "epoch": 1.7491525423728813,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676461056,
+    "loss": 0.4638,
+    "grad_norm": 1.069642186164856,
+    "learning_rate": 1.3555200696822232e-05
+  },
+  {
+    "step": 259,
+    "epoch": 1.7559322033898304,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676378112,
+    "loss": 0.3957,
+    "grad_norm": 1.1308850049972534,
+    "learning_rate": 1.2823364948184095e-05
+  },
+  {
+    "step": 260,
+    "epoch": 1.7627118644067796,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676494848,
+    "loss": 0.3246,
+    "grad_norm": 0.8047335147857666,
+    "learning_rate": 1.2110954462750166e-05
+  },
+  {
+    "step": 261,
+    "epoch": 1.769491525423729,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676450304,
+    "loss": 0.2867,
+    "grad_norm": 0.6239379644393921,
+    "learning_rate": 1.1418070123306989e-05
+  },
+  {
+    "step": 262,
+    "epoch": 1.776271186440678,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676407296,
+    "loss": 0.3148,
+    "grad_norm": 0.6769895553588867,
+    "learning_rate": 1.0744810047589115e-05
+  },
+  {
+    "step": 263,
+    "epoch": 1.783050847457627,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.67644416,
+    "loss": 0.3841,
+    "grad_norm": 0.7769913077354431,
+    "learning_rate": 1.0091269574384874e-05
+  },
+  {
+    "step": 264,
+    "epoch": 1.7898305084745763,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676531712,
+    "loss": 0.3774,
+    "grad_norm": 0.8806053400039673,
+    "learning_rate": 9.45754125003576e-06
+  },
+  {
+    "step": 265,
+    "epoch": 1.7966101694915255,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676450304,
+    "loss": 0.4358,
+    "grad_norm": 0.8880800008773804,
+    "learning_rate": 8.843714815330987e-06
+  },
+  {
+    "step": 266,
+    "epoch": 1.8033898305084746,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676665344,
+    "loss": 0.4605,
+    "grad_norm": 0.7958164811134338,
+    "learning_rate": 8.249877192799731e-06
+  },
+  {
+    "step": 267,
+    "epoch": 1.8101694915254236,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676457984,
+    "loss": 0.4532,
+    "grad_norm": 1.444252371788025,
+    "learning_rate": 7.676112474402068e-06
+  },
+  {
+    "step": 268,
+    "epoch": 1.8169491525423729,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676462592,
+    "loss": 0.3734,
+    "grad_norm": 0.9235910773277283,
+    "learning_rate": 7.122501909620926e-06
+  },
+  {
+    "step": 269,
+    "epoch": 1.8237288135593221,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676473344,
+    "loss": 0.4529,
+    "grad_norm": 1.036499261856079,
+    "learning_rate": 6.5891238939566275e-06
+  },
+  {
+    "step": 270,
+    "epoch": 1.8305084745762712,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676511744,
+    "loss": 0.4235,
+    "grad_norm": 1.073111653327942,
+    "learning_rate": 6.076053957825411e-06
+  },
+  {
+    "step": 271,
+    "epoch": 1.8372881355932202,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676563968,
+    "loss": 0.4308,
+    "grad_norm": 1.0276501178741455,
+    "learning_rate": 5.583364755863701e-06
+  },
+  {
+    "step": 272,
+    "epoch": 1.8440677966101695,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676422656,
+    "loss": 0.4068,
+    "grad_norm": 0.6079270839691162,
+    "learning_rate": 5.11112605663977e-06
+  },
+  {
+    "step": 273,
+    "epoch": 1.8508474576271188,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676302848,
+    "loss": 0.414,
+    "grad_norm": 0.7414558529853821,
+    "learning_rate": 4.659404732773908e-06
+  },
+  {
+    "step": 274,
+    "epoch": 1.8576271186440678,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676530176,
+    "loss": 0.3892,
+    "grad_norm": 0.9337151050567627,
+    "learning_rate": 4.228264751468752e-06
+  },
+  {
+    "step": 275,
+    "epoch": 1.8644067796610169,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.6767744,
+    "loss": 0.4114,
+    "grad_norm": 1.0385849475860596,
+    "learning_rate": 3.817767165451041e-06
+  },
+  {
+    "step": 276,
+    "epoch": 1.8711864406779661,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676434944,
+    "loss": 0.42,
+    "grad_norm": 0.8132132291793823,
+    "learning_rate": 3.4279701043260886e-06
+  },
+  {
+    "step": 277,
+    "epoch": 1.8779661016949154,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676381184,
+    "loss": 0.4333,
+    "grad_norm": 0.9133860468864441,
+    "learning_rate": 3.0589287663461472e-06
+  },
+  {
+    "step": 278,
+    "epoch": 1.8847457627118644,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676544,
+    "loss": 0.5146,
+    "grad_norm": 0.8138352036476135,
+    "learning_rate": 2.710695410593994e-06
+  },
+  {
+    "step": 279,
+    "epoch": 1.8915254237288135,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676484096,
+    "loss": 0.4438,
+    "grad_norm": 0.8574395775794983,
+    "learning_rate": 2.3833193495825853e-06
+  },
+  {
+    "step": 280,
+    "epoch": 1.8983050847457628,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676464128,
+    "loss": 0.4087,
+    "grad_norm": 0.8839820027351379,
+    "learning_rate": 2.076846942272026e-06
+  },
+  {
+    "step": 281,
+    "epoch": 1.905084745762712,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676399616,
+    "loss": 0.4411,
+    "grad_norm": 0.9226896166801453,
+    "learning_rate": 1.791321587504768e-06
+  },
+  {
+    "step": 282,
+    "epoch": 1.911864406779661,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.67682816,
+    "loss": 0.3647,
+    "grad_norm": 1.2175434827804565,
+    "learning_rate": 1.5267837178600972e-06
+  },
+  {
+    "step": 283,
+    "epoch": 1.9186440677966101,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676534784,
+    "loss": 0.3445,
+    "grad_norm": 0.8091426491737366,
+    "learning_rate": 1.2832707939284427e-06
+  },
+  {
+    "step": 284,
+    "epoch": 1.9254237288135592,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.6763904,
+    "loss": 0.3911,
+    "grad_norm": 0.6109596490859985,
+    "learning_rate": 1.0608172990067553e-06
+  },
+  {
+    "step": 285,
+    "epoch": 1.9322033898305084,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.67644416,
+    "loss": 0.4041,
+    "grad_norm": 0.8254313468933105,
+    "learning_rate": 8.594547342153979e-07
+  },
+  {
+    "step": 286,
+    "epoch": 1.9389830508474577,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676861952,
+    "loss": 0.3275,
+    "grad_norm": 0.6968205571174622,
+    "learning_rate": 6.792116140373116e-07
+  },
+  {
+    "step": 287,
+    "epoch": 1.9457627118644067,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676631552,
+    "loss": 0.3366,
+    "grad_norm": 0.6965417861938477,
+    "learning_rate": 5.201134622801473e-07
+  },
+  {
+    "step": 288,
+    "epoch": 1.9525423728813558,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676416512,
+    "loss": 0.5175,
+    "grad_norm": 1.0540348291397095,
+    "learning_rate": 3.821828084619727e-07
+  },
+  {
+    "step": 289,
+    "epoch": 1.959322033898305,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676500992,
+    "loss": 0.4007,
+    "grad_norm": 0.7388603091239929,
+    "learning_rate": 2.654391846207915e-07
+  },
+  {
+    "step": 290,
+    "epoch": 1.9661016949152543,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676425728,
+    "loss": 0.4716,
+    "grad_norm": 1.206581473350525,
+    "learning_rate": 1.6989912254880556e-07
+  },
+  {
+    "step": 291,
+    "epoch": 1.9728813559322034,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676461056,
+    "loss": 0.4941,
+    "grad_norm": 0.9969897270202637,
+    "learning_rate": 9.557615145123765e-08
+  },
+  {
+    "step": 292,
+    "epoch": 1.9796610169491524,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676544,
+    "loss": 0.4308,
+    "grad_norm": 0.7124324440956116,
+    "learning_rate": 4.248079603064724e-08
+  },
+  {
+    "step": 293,
+    "epoch": 1.9864406779661017,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676461056,
+    "loss": 0.4941,
+    "grad_norm": 1.315065622329712,
+    "learning_rate": 1.0620574996372811e-08
+  },
+  {
+    "step": 294,
+    "epoch": 1.993220338983051,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676487168,
+    "loss": 0.4635,
+    "grad_norm": 0.8283355832099915,
+    "learning_rate": 0.0
+  },
+  {
+    "step": 294,
+    "epoch": 1.993220338983051,
+    "cpu_mem": 1.511325696,
+    "gpu_mem": 4.676487168,
+    "train_runtime": 4573.7736,
+    "train_samples_per_second": 4.122,
+    "train_steps_per_second": 0.064,
+    "total_flos": 0.0,
+    "train_loss": 0.7646797082456601
+  }
+]

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-boolq-r8-a2/adapter_config.json

@@ -0,0 +1,30 @@
+{
+  "alpha": 16,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "init_weight": "kaiming",
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "metric_tracking": false,
+  "modules_to_save": null,
+  "peft_type": "ABLATION",
+  "r": 8,
+  "revision": null,
+  "seed": 42,
+  "share_weights": false,
+  "target_modules": [
+    "up_proj",
+    "gate_proj",
+    "v_proj",
+    "k_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj"
+  ],
+  "task_type": null,
+  "track_n": 100,
+  "variant": "B"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-boolq-r8-a2/training_configuration.json

@@ -0,0 +1,38 @@
+{
+  "model_id": "TinyLlama/TinyLlama_v1.1",
+  "dataset": {
+    "name": "BOOLQ",
+    "dataset_id": "google/boolq",
+    "preprocess_id": "boolq_train_deepeval"
+  },
+  "peft_config": {
+    "method": "abl_B",
+    "rank": 8,
+    "alpha": 16,
+    "dropout": 0.0,
+    "bias": "none",
+    "target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "down_proj",
+      "up_proj"
+    ],
+    "trainable_parameter_count": 3548160
+  },
+  "training_config": {
+    "max_dataset_length": null,
+    "batch_size": 64,
+    "per_device_batch_size": 32,
+    "gradient_accumulation_steps": 2,
+    "learning_rate": 0.0003,
+    "num_epochs": 2,
+    "warmup_ratio": 0.1
+  },
+  "model_name": "TinyLlama_v1.1-abl_B-boolq-r8-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-boolq-r8-a2",
+  "seed": 42,
+  "timestamp": "2025-08-31T16:07:01.748900"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-hellaswag-r2-a2/adapter_config.json

@@ -0,0 +1,30 @@
+{
+  "alpha": 4,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "init_weight": "kaiming",
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "metric_tracking": false,
+  "modules_to_save": null,
+  "peft_type": "ABLATION",
+  "r": 2,
+  "revision": null,
+  "seed": 42,
+  "share_weights": false,
+  "target_modules": [
+    "up_proj",
+    "gate_proj",
+    "v_proj",
+    "k_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj"
+  ],
+  "task_type": null,
+  "track_n": 100,
+  "variant": "B"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-hellaswag-r2-a2/eval_results.json

@@ -0,0 +1,4 @@
+{
+    "task": "hellaswag",
+    "results": 0.7999402509460267
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-hellaswag-r2-a2/training_configuration.json

@@ -0,0 +1,38 @@
+{
+  "model_id": "TinyLlama/TinyLlama_v1.1",
+  "dataset": {
+    "name": "HELLASWAG",
+    "dataset_id": "Rowan/hellaswag",
+    "preprocess_id": "hellaswag_train_deepeval"
+  },
+  "peft_config": {
+    "method": "abl_B",
+    "rank": 2,
+    "alpha": 4,
+    "dropout": 0.0,
+    "bias": "none",
+    "target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "down_proj",
+      "up_proj"
+    ],
+    "trainable_parameter_count": 1182720
+  },
+  "training_config": {
+    "max_dataset_length": null,
+    "batch_size": 64,
+    "per_device_batch_size": 32,
+    "gradient_accumulation_steps": 2,
+    "learning_rate": 0.0003,
+    "num_epochs": 1,
+    "warmup_ratio": 0.1
+  },
+  "model_name": "TinyLlama_v1.1-abl_B-hellaswag-r2-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-hellaswag-r2-a2",
+  "seed": 42,
+  "timestamp": "2025-08-31T13:39:06.562417"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-hellaswag-r8-a2/adapter_config.json

@@ -0,0 +1,30 @@
+{
+  "alpha": 16,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "init_weight": "kaiming",
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "metric_tracking": false,
+  "modules_to_save": null,
+  "peft_type": "ABLATION",
+  "r": 8,
+  "revision": null,
+  "seed": 42,
+  "share_weights": false,
+  "target_modules": [
+    "up_proj",
+    "gate_proj",
+    "v_proj",
+    "k_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj"
+  ],
+  "task_type": null,
+  "track_n": 100,
+  "variant": "B"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-hellaswag-r8-a2/eval_results.json

@@ -0,0 +1,4 @@
+{
+    "task": "hellaswag",
+    "results": 0.8323043218482374
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-hellaswag-r8-a2/training_configuration.json

@@ -0,0 +1,38 @@
+{
+  "model_id": "TinyLlama/TinyLlama_v1.1",
+  "dataset": {
+    "name": "HELLASWAG",
+    "dataset_id": "Rowan/hellaswag",
+    "preprocess_id": "hellaswag_train_deepeval"
+  },
+  "peft_config": {
+    "method": "abl_B",
+    "rank": 8,
+    "alpha": 16,
+    "dropout": 0.0,
+    "bias": "none",
+    "target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "down_proj",
+      "up_proj"
+    ],
+    "trainable_parameter_count": 3548160
+  },
+  "training_config": {
+    "max_dataset_length": null,
+    "batch_size": 64,
+    "per_device_batch_size": 32,
+    "gradient_accumulation_steps": 2,
+    "learning_rate": 0.0003,
+    "num_epochs": 1,
+    "warmup_ratio": 0.1
+  },
+  "model_name": "TinyLlama_v1.1-abl_B-hellaswag-r8-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-hellaswag-r8-a2",
+  "seed": 42,
+  "timestamp": "2025-08-31T20:40:36.678485"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-logiqa-r32-a2/adapter_config.json

@@ -0,0 +1,30 @@
+{
+  "alpha": 64,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "init_weight": "kaiming",
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "metric_tracking": false,
+  "modules_to_save": null,
+  "peft_type": "ABLATION",
+  "r": 32,
+  "revision": null,
+  "seed": 42,
+  "share_weights": false,
+  "target_modules": [
+    "up_proj",
+    "gate_proj",
+    "v_proj",
+    "k_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj"
+  ],
+  "task_type": null,
+  "track_n": 100,
+  "variant": "B"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-logiqa-r32-a2/eval_results.json

@@ -0,0 +1,4 @@
+{
+    "task": "logiqa",
+    "results": 0.29394753150175584
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-logiqa-r32-a2/training_configuration.json

@@ -0,0 +1,38 @@
+{
+  "model_id": "TinyLlama/TinyLlama_v1.1",
+  "dataset": {
+    "name": "LOGIQA",
+    "dataset_id": "data/logiqa_train",
+    "preprocess_id": "logiqa_train_deepeval"
+  },
+  "peft_config": {
+    "method": "abl_B",
+    "rank": 32,
+    "alpha": 64,
+    "dropout": 0.0,
+    "bias": "none",
+    "target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "down_proj",
+      "up_proj"
+    ],
+    "trainable_parameter_count": 13009920
+  },
+  "training_config": {
+    "max_dataset_length": null,
+    "batch_size": 64,
+    "per_device_batch_size": 32,
+    "gradient_accumulation_steps": 2,
+    "learning_rate": 0.0003,
+    "num_epochs": 3,
+    "warmup_ratio": 0.1
+  },
+  "model_name": "TinyLlama_v1.1-abl_B-logiqa-r32-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-logiqa-r32-a2",
+  "seed": 42,
+  "timestamp": "2025-09-01T00:29:31.648142"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-logiqa-r8-a2/adapter_config.json

@@ -0,0 +1,30 @@
+{
+  "alpha": 16,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "init_weight": "kaiming",
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "metric_tracking": false,
+  "modules_to_save": null,
+  "peft_type": "ABLATION",
+  "r": 8,
+  "revision": null,
+  "seed": 42,
+  "share_weights": false,
+  "target_modules": [
+    "up_proj",
+    "gate_proj",
+    "v_proj",
+    "k_proj",
+    "q_proj",
+    "down_proj",
+    "o_proj"
+  ],
+  "task_type": null,
+  "track_n": 100,
+  "variant": "B"
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-logiqa-r8-a2/eval_results.json

@@ -0,0 +1,4 @@
+{
+    "task": "logiqa",
+    "results": 0.3850444123115059
+}

TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-logiqa-r8-a2/training_configuration.json

@@ -0,0 +1,38 @@
+{
+  "model_id": "TinyLlama/TinyLlama_v1.1",
+  "dataset": {
+    "name": "LOGIQA",
+    "dataset_id": "data/logiqa_train",
+    "preprocess_id": "logiqa_train_deepeval"
+  },
+  "peft_config": {
+    "method": "abl_B",
+    "rank": 8,
+    "alpha": 16,
+    "dropout": 0.0,
+    "bias": "none",
+    "target_modules": [
+      "q_proj",
+      "k_proj",
+      "v_proj",
+      "o_proj",
+      "gate_proj",
+      "down_proj",
+      "up_proj"
+    ],
+    "trainable_parameter_count": 3548160
+  },
+  "training_config": {
+    "max_dataset_length": null,
+    "batch_size": 64,
+    "per_device_batch_size": 32,
+    "gradient_accumulation_steps": 2,
+    "learning_rate": 0.0003,
+    "num_epochs": 3,
+    "warmup_ratio": 0.1
+  },
+  "model_name": "TinyLlama_v1.1-abl_B-logiqa-r8-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-abl_B/TinyLlama_v1.1-abl_B-logiqa-r8-a2",
+  "seed": 42,
+  "timestamp": "2025-08-31T17:26:03.159670"
+}