Add files using upload-large-folder tool

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-arc_c-r2-a2/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha": 4,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "enabled_mlp": true,
+  "enabled_qkv": [
+    "q",
+    "k",
+    "v"
+  ],
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "mixture": false,
+  "modules_to_preserve_errors": null,
+  "modules_to_quantize": null,
+  "modules_to_save": null,
+  "onnx_export": false,
+  "optimization_level": 3,
+  "orthogonal_init": false,
+  "peft_type": "MARS",
+  "quant_n_bits": 4,
+  "r": 2,
+  "revision": null,
+  "seed": 42,
+  "shared_r": 2,
+  "target_modules": [
+    "o_proj",
+    "k_proj",
+    "down_proj",
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "q_proj"
+  ],
+  "task_type": null
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-arc_c-r2-a2/eval_results.json

@@ -0,0 +1,4 @@
+{
+    "task": "arc_c",
+    "results": 0.5691126279863481
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-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": "mars",
+    "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": 1307064
+  },
+  "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-mars-arc_c-r2-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-mars-opt3-q4-new/TinyLlama_v1.1-mars-arc_c-r2-a2",
+  "seed": 42,
+  "timestamp": "2025-08-29T18:23:17.489113"
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-arc_c-r2-a2/training_logs.json

@@ -0,0 +1,625 @@
+[
+  {
+    "step": 1,
+    "epoch": 0.05714285714285714,
+    "cpu_mem": 2.580422656,
+    "gpu_mem": 1.065693696,
+    "loss": 4.6127,
+    "grad_norm": 381.2337341308594,
+    "learning_rate": 4.285714285714285e-05
+  },
+  {
+    "step": 2,
+    "epoch": 0.11428571428571428,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076182016,
+    "loss": 4.674,
+    "grad_norm": 388.8489685058594,
+    "learning_rate": 8.57142857142857e-05
+  },
+  {
+    "step": 3,
+    "epoch": 0.17142857142857143,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076212736,
+    "loss": 1.8432,
+    "grad_norm": 72.13705444335938,
+    "learning_rate": 0.00012857142857142855
+  },
+  {
+    "step": 4,
+    "epoch": 0.22857142857142856,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076178944,
+    "loss": 1.6594,
+    "grad_norm": 50.97885513305664,
+    "learning_rate": 0.0001714285714285714
+  },
+  {
+    "step": 5,
+    "epoch": 0.2857142857142857,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076166656,
+    "loss": 1.4504,
+    "grad_norm": 25.90673065185547,
+    "learning_rate": 0.00021428571428571427
+  },
+  {
+    "step": 6,
+    "epoch": 0.34285714285714286,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076229632,
+    "loss": 1.4254,
+    "grad_norm": 21.743545532226562,
+    "learning_rate": 0.0002571428571428571
+  },
+  {
+    "step": 7,
+    "epoch": 0.4,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076235776,
+    "loss": 1.4606,
+    "grad_norm": 25.414306640625,
+    "learning_rate": 0.0003
+  },
+  {
+    "step": 8,
+    "epoch": 0.45714285714285713,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076194304,
+    "loss": 1.5571,
+    "grad_norm": 35.37538146972656,
+    "learning_rate": 0.00029980111348272456
+  },
+  {
+    "step": 9,
+    "epoch": 0.5142857142857142,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076189696,
+    "loss": 1.3306,
+    "grad_norm": 12.570324897766113,
+    "learning_rate": 0.00029920498134218835
+  },
+  {
+    "step": 10,
+    "epoch": 0.5714285714285714,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076178944,
+    "loss": 1.6494,
+    "grad_norm": 30.424644470214844,
+    "learning_rate": 0.0002982131844136615
+  },
+  {
+    "step": 11,
+    "epoch": 0.6285714285714286,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076189696,
+    "loss": 1.4528,
+    "grad_norm": 14.515549659729004,
+    "learning_rate": 0.0002968283527643036
+  },
+  {
+    "step": 12,
+    "epoch": 0.6857142857142857,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076214272,
+    "loss": 1.443,
+    "grad_norm": 14.896547317504883,
+    "learning_rate": 0.000295054158718698
+  },
+  {
+    "step": 13,
+    "epoch": 0.7428571428571429,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076214272,
+    "loss": 1.3921,
+    "grad_norm": 30.263879776000977,
+    "learning_rate": 0.00029289530712050735
+  },
+  {
+    "step": 14,
+    "epoch": 0.8,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076162048,
+    "loss": 1.7549,
+    "grad_norm": 41.95735168457031,
+    "learning_rate": 0.000290357522856074
+  },
+  {
+    "step": 15,
+    "epoch": 0.8571428571428571,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076237312,
+    "loss": 1.5313,
+    "grad_norm": 22.88416290283203,
+    "learning_rate": 0.0002874475356730507
+  },
+  {
+    "step": 16,
+    "epoch": 0.9142857142857143,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076231168,
+    "loss": 1.4738,
+    "grad_norm": 30.19275665283203,
+    "learning_rate": 0.0002841730623343193
+  },
+  {
+    "step": 17,
+    "epoch": 0.9714285714285714,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076235776,
+    "loss": 1.5348,
+    "grad_norm": 30.131532669067383,
+    "learning_rate": 0.00028054278615452326
+  },
+  {
+    "step": 18,
+    "epoch": 1.0285714285714285,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081441792,
+    "loss": 2.0798,
+    "grad_norm": 17.52899742126465,
+    "learning_rate": 0.0002765663339734778
+  },
+  {
+    "step": 19,
+    "epoch": 1.0857142857142856,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081440256,
+    "loss": 1.4488,
+    "grad_norm": 20.202335357666016,
+    "learning_rate": 0.00027225425062752165
+  },
+  {
+    "step": 20,
+    "epoch": 1.1428571428571428,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.08141568,
+    "loss": 1.2952,
+    "grad_norm": 13.343443870544434,
+    "learning_rate": 0.0002676179709865066
+  },
+  {
+    "step": 21,
+    "epoch": 1.2,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.08142336,
+    "loss": 1.4289,
+    "grad_norm": 17.170682907104492,
+    "learning_rate": 0.0002626697896305779
+  },
+  {
+    "step": 22,
+    "epoch": 1.2571428571428571,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081452544,
+    "loss": 1.391,
+    "grad_norm": 17.611751556396484,
+    "learning_rate": 0.000257422828247159
+  },
+  {
+    "step": 23,
+    "epoch": 1.3142857142857143,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081481728,
+    "loss": 1.3177,
+    "grad_norm": 11.4840087890625,
+    "learning_rate": 0.00025189100083459397
+  },
+  {
+    "step": 24,
+    "epoch": 1.3714285714285714,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081424896,
+    "loss": 1.5163,
+    "grad_norm": 30.58349609375,
+    "learning_rate": 0.0002460889768047263
+  },
+  {
+    "step": 25,
+    "epoch": 1.4285714285714286,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081494016,
+    "loss": 1.3967,
+    "grad_norm": 14.075834274291992,
+    "learning_rate": 0.00024003214208225522
+  },
+  {
+    "step": 26,
+    "epoch": 1.4857142857142858,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081451008,
+    "loss": 1.3574,
+    "grad_norm": 14.51030158996582,
+    "learning_rate": 0.00023373655830402968
+  },
+  {
+    "step": 27,
+    "epoch": 1.5428571428571427,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081409536,
+    "loss": 1.3929,
+    "grad_norm": 13.834196090698242,
+    "learning_rate": 0.00022721892022647462
+  },
+  {
+    "step": 28,
+    "epoch": 1.6,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081455616,
+    "loss": 1.6196,
+    "grad_norm": 46.8470344543457,
+    "learning_rate": 0.000220496511454098
+  },
+  {
+    "step": 29,
+    "epoch": 1.657142857142857,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081451008,
+    "loss": 1.3476,
+    "grad_norm": 7.685541152954102,
+    "learning_rate": 0.0002135871586064791
+  },
+  {
+    "step": 30,
+    "epoch": 1.7142857142857144,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081440256,
+    "loss": 1.3697,
+    "grad_norm": 9.882088661193848,
+    "learning_rate": 0.00020650918404527775
+  },
+  {
+    "step": 31,
+    "epoch": 1.7714285714285714,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081470976,
+    "loss": 1.331,
+    "grad_norm": 13.1102294921875,
+    "learning_rate": 0.00019928135728662522
+  },
+  {
+    "step": 32,
+    "epoch": 1.8285714285714287,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081480192,
+    "loss": 1.3643,
+    "grad_norm": 9.88525676727295,
+    "learning_rate": 0.00019192284522774142
+  },
+  {
+    "step": 33,
+    "epoch": 1.8857142857142857,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081460224,
+    "loss": 1.3741,
+    "grad_norm": 8.89490795135498,
+    "learning_rate": 0.00018445316131976934
+  },
+  {
+    "step": 34,
+    "epoch": 1.9428571428571428,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.08143872,
+    "loss": 1.3514,
+    "grad_norm": 11.753167152404785,
+    "learning_rate": 0.00017689211382161034
+  },
+  {
+    "step": 35,
+    "epoch": 2.0,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081326592,
+    "loss": 2.0743,
+    "grad_norm": 16.071163177490234,
+    "learning_rate": 0.00016925975327198266
+  },
+  {
+    "step": 36,
+    "epoch": 2.057142857142857,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076208128,
+    "loss": 1.3093,
+    "grad_norm": 8.304706573486328,
+    "learning_rate": 0.00016157631931899697
+  },
+  {
+    "step": 37,
+    "epoch": 2.1142857142857143,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076217344,
+    "loss": 1.4079,
+    "grad_norm": 17.81264877319336,
+    "learning_rate": 0.0001538621870482483
+  },
+  {
+    "step": 38,
+    "epoch": 2.1714285714285713,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.07618816,
+    "loss": 1.3492,
+    "grad_norm": 16.122692108154297,
+    "learning_rate": 0.00014613781295175172
+  },
+  {
+    "step": 39,
+    "epoch": 2.2285714285714286,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076206592,
+    "loss": 1.3398,
+    "grad_norm": 14.483545303344727,
+    "learning_rate": 0.00013842368068100303
+  },
+  {
+    "step": 40,
+    "epoch": 2.2857142857142856,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076183552,
+    "loss": 1.2991,
+    "grad_norm": 7.728339195251465,
+    "learning_rate": 0.00013074024672801731
+  },
+  {
+    "step": 41,
+    "epoch": 2.342857142857143,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076185088,
+    "loss": 1.3073,
+    "grad_norm": 8.475275039672852,
+    "learning_rate": 0.00012310788617838966
+  },
+  {
+    "step": 42,
+    "epoch": 2.4,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076214272,
+    "loss": 1.4702,
+    "grad_norm": 66.37582397460938,
+    "learning_rate": 0.00011554683868023067
+  },
+  {
+    "step": 43,
+    "epoch": 2.4571428571428573,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076229632,
+    "loss": 1.2906,
+    "grad_norm": 13.710612297058105,
+    "learning_rate": 0.00010807715477225858
+  },
+  {
+    "step": 44,
+    "epoch": 2.5142857142857142,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076248064,
+    "loss": 1.3712,
+    "grad_norm": 212.30665588378906,
+    "learning_rate": 0.00010071864271337478
+  },
+  {
+    "step": 45,
+    "epoch": 2.571428571428571,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076201984,
+    "loss": 1.2661,
+    "grad_norm": 96.19535827636719,
+    "learning_rate": 9.34908159547222e-05
+  },
+  {
+    "step": 46,
+    "epoch": 2.6285714285714286,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.07619584,
+    "loss": 1.2963,
+    "grad_norm": 16.374256134033203,
+    "learning_rate": 8.641284139352091e-05
+  },
+  {
+    "step": 47,
+    "epoch": 2.685714285714286,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076189696,
+    "loss": 1.293,
+    "grad_norm": 18.95760154724121,
+    "learning_rate": 7.950348854590204e-05
+  },
+  {
+    "step": 48,
+    "epoch": 2.742857142857143,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076194304,
+    "loss": 1.211,
+    "grad_norm": 10.877612113952637,
+    "learning_rate": 7.278107977352543e-05
+  },
+  {
+    "step": 49,
+    "epoch": 2.8,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076185088,
+    "loss": 1.2268,
+    "grad_norm": 15.633013725280762,
+    "learning_rate": 6.626344169597031e-05
+  },
+  {
+    "step": 50,
+    "epoch": 2.857142857142857,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076166656,
+    "loss": 1.1838,
+    "grad_norm": 8.313477516174316,
+    "learning_rate": 5.996785791774478e-05
+  },
+  {
+    "step": 51,
+    "epoch": 2.914285714285714,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.076191232,
+    "loss": 1.2146,
+    "grad_norm": 11.858922004699707,
+    "learning_rate": 5.391102319527373e-05
+  },
+  {
+    "step": 52,
+    "epoch": 2.9714285714285715,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.07621888,
+    "loss": 1.305,
+    "grad_norm": 9.989842414855957,
+    "learning_rate": 4.8108999165406026e-05
+  },
+  {
+    "step": 53,
+    "epoch": 3.0285714285714285,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081437184,
+    "loss": 1.8208,
+    "grad_norm": 18.719863891601562,
+    "learning_rate": 4.257717175284103e-05
+  },
+  {
+    "step": 54,
+    "epoch": 3.085714285714286,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081406464,
+    "loss": 1.2094,
+    "grad_norm": 13.842973709106445,
+    "learning_rate": 3.733021036942205e-05
+  },
+  {
+    "step": 55,
+    "epoch": 3.142857142857143,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081440256,
+    "loss": 1.2424,
+    "grad_norm": 14.39765739440918,
+    "learning_rate": 3.238202901349345e-05
+  },
+  {
+    "step": 56,
+    "epoch": 3.2,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081513984,
+    "loss": 1.151,
+    "grad_norm": 12.968193054199219,
+    "learning_rate": 2.774574937247831e-05
+  },
+  {
+    "step": 57,
+    "epoch": 3.257142857142857,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081457152,
+    "loss": 1.1003,
+    "grad_norm": 13.164222717285156,
+    "learning_rate": 2.3433666026522153e-05
+  },
+  {
+    "step": 58,
+    "epoch": 3.314285714285714,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081451008,
+    "loss": 1.0009,
+    "grad_norm": 10.711393356323242,
+    "learning_rate": 1.945721384547671e-05
+  },
+  {
+    "step": 59,
+    "epoch": 3.3714285714285714,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081501696,
+    "loss": 1.0536,
+    "grad_norm": 13.167465209960938,
+    "learning_rate": 1.5826937665680693e-05
+  },
+  {
+    "step": 60,
+    "epoch": 3.4285714285714284,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081427968,
+    "loss": 1.1045,
+    "grad_norm": 23.531604766845703,
+    "learning_rate": 1.2552464326949302e-05
+  },
+  {
+    "step": 61,
+    "epoch": 3.4857142857142858,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081441792,
+    "loss": 1.1359,
+    "grad_norm": 17.696874618530273,
+    "learning_rate": 9.64247714392597e-06
+  },
+  {
+    "step": 62,
+    "epoch": 3.5428571428571427,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081443328,
+    "loss": 1.0889,
+    "grad_norm": 15.676621437072754,
+    "learning_rate": 7.104692879492624e-06
+  },
+  {
+    "step": 63,
+    "epoch": 3.6,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081432576,
+    "loss": 1.178,
+    "grad_norm": 18.50305938720703,
+    "learning_rate": 4.945841281301943e-06
+  },
+  {
+    "step": 64,
+    "epoch": 3.657142857142857,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081449472,
+    "loss": 1.1488,
+    "grad_norm": 18.644163131713867,
+    "learning_rate": 3.1716472356963286e-06
+  },
+  {
+    "step": 65,
+    "epoch": 3.7142857142857144,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081470976,
+    "loss": 1.0864,
+    "grad_norm": 16.862136840820312,
+    "learning_rate": 1.7868155863384415e-06
+  },
+  {
+    "step": 66,
+    "epoch": 3.7714285714285714,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.08146176,
+    "loss": 1.012,
+    "grad_norm": 18.95318031311035,
+    "learning_rate": 7.950186578116413e-07
+  },
+  {
+    "step": 67,
+    "epoch": 3.8285714285714287,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.081487872,
+    "loss": 1.0928,
+    "grad_norm": 18.284706115722656,
+    "learning_rate": 1.988865172754206e-07
+  },
+  {
+    "step": 68,
+    "epoch": 3.8857142857142857,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.08143872,
+    "loss": 1.1189,
+    "grad_norm": 15.364293098449707,
+    "learning_rate": 0.0
+  },
+  {
+    "step": 68,
+    "epoch": 3.8857142857142857,
+    "cpu_mem": 2.580619264,
+    "gpu_mem": 1.08143872,
+    "train_runtime": 374.0598,
+    "train_samples_per_second": 11.966,
+    "train_steps_per_second": 0.182,
+    "total_flos": 0.0,
+    "train_loss": 1.4615876972675323
+  }
+]

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-arc_c-r32-a2/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha": 64,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "enabled_mlp": true,
+  "enabled_qkv": [
+    "q",
+    "k",
+    "v"
+  ],
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "mixture": false,
+  "modules_to_preserve_errors": null,
+  "modules_to_quantize": null,
+  "modules_to_save": null,
+  "onnx_export": false,
+  "optimization_level": 3,
+  "orthogonal_init": false,
+  "peft_type": "MARS",
+  "quant_n_bits": 4,
+  "r": 32,
+  "revision": null,
+  "seed": 42,
+  "shared_r": 32,
+  "target_modules": [
+    "o_proj",
+    "k_proj",
+    "down_proj",
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "q_proj"
+  ],
+  "task_type": null
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-arc_c-r32-a2/eval_results.json

@@ -0,0 +1,4 @@
+{
+    "task": "arc_c",
+    "results": 0.5503412969283277
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-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": "mars",
+    "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": 21018624
+  },
+  "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-mars-arc_c-r32-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-mars-opt3-q4-new/TinyLlama_v1.1-mars-arc_c-r32-a2",
+  "seed": 42,
+  "timestamp": "2025-08-30T08:36:40.140752"
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-arc_c-r32-a2/training_logs.json

@@ -0,0 +1,625 @@
+[
+  {
+    "step": 1,
+    "epoch": 0.05714285714285714,
+    "cpu_mem": 2.438529024,
+    "gpu_mem": 1.153432064,
+    "loss": 4.6127,
+    "grad_norm": 83.32475280761719,
+    "learning_rate": 4.285714285714285e-05
+  },
+  {
+    "step": 2,
+    "epoch": 0.11428571428571428,
+    "cpu_mem": 2.438725632,
+    "gpu_mem": 1.321571328,
+    "loss": 4.674,
+    "grad_norm": 85.61141204833984,
+    "learning_rate": 8.57142857142857e-05
+  },
+  {
+    "step": 3,
+    "epoch": 0.17142857142857143,
+    "cpu_mem": 2.43892224,
+    "gpu_mem": 1.321602048,
+    "loss": 2.0451,
+    "grad_norm": 22.385805130004883,
+    "learning_rate": 0.00012857142857142855
+  },
+  {
+    "step": 4,
+    "epoch": 0.22857142857142856,
+    "cpu_mem": 2.43892224,
+    "gpu_mem": 1.321568256,
+    "loss": 1.5792,
+    "grad_norm": 8.16244125366211,
+    "learning_rate": 0.0001714285714285714
+  },
+  {
+    "step": 5,
+    "epoch": 0.2857142857142857,
+    "cpu_mem": 2.43892224,
+    "gpu_mem": 1.321555968,
+    "loss": 1.4964,
+    "grad_norm": 8.01338005065918,
+    "learning_rate": 0.00021428571428571427
+  },
+  {
+    "step": 6,
+    "epoch": 0.34285714285714286,
+    "cpu_mem": 2.43892224,
+    "gpu_mem": 1.321618944,
+    "loss": 1.448,
+    "grad_norm": 5.569636821746826,
+    "learning_rate": 0.0002571428571428571
+  },
+  {
+    "step": 7,
+    "epoch": 0.4,
+    "cpu_mem": 2.43892224,
+    "gpu_mem": 1.321625088,
+    "loss": 1.4577,
+    "grad_norm": 5.084090232849121,
+    "learning_rate": 0.0003
+  },
+  {
+    "step": 8,
+    "epoch": 0.45714285714285713,
+    "cpu_mem": 2.43892224,
+    "gpu_mem": 1.321583616,
+    "loss": 1.4895,
+    "grad_norm": 4.238208293914795,
+    "learning_rate": 0.00029980111348272456
+  },
+  {
+    "step": 9,
+    "epoch": 0.5142857142857142,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.321579008,
+    "loss": 1.3082,
+    "grad_norm": 1.7702213525772095,
+    "learning_rate": 0.00029920498134218835
+  },
+  {
+    "step": 10,
+    "epoch": 0.5714285714285714,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.321568256,
+    "loss": 1.5789,
+    "grad_norm": 5.247867107391357,
+    "learning_rate": 0.0002982131844136615
+  },
+  {
+    "step": 11,
+    "epoch": 0.6285714285714286,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.321579008,
+    "loss": 1.4814,
+    "grad_norm": 3.4710798263549805,
+    "learning_rate": 0.0002968283527643036
+  },
+  {
+    "step": 12,
+    "epoch": 0.6857142857142857,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.321603584,
+    "loss": 1.3923,
+    "grad_norm": 1.7332303524017334,
+    "learning_rate": 0.000295054158718698
+  },
+  {
+    "step": 13,
+    "epoch": 0.7428571428571429,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.321603584,
+    "loss": 1.3625,
+    "grad_norm": 3.513798713684082,
+    "learning_rate": 0.00029289530712050735
+  },
+  {
+    "step": 14,
+    "epoch": 0.8,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.32155136,
+    "loss": 1.653,
+    "grad_norm": 4.981041431427002,
+    "learning_rate": 0.000290357522856074
+  },
+  {
+    "step": 15,
+    "epoch": 0.8571428571428571,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.321626624,
+    "loss": 1.5649,
+    "grad_norm": 3.986966133117676,
+    "learning_rate": 0.0002874475356730507
+  },
+  {
+    "step": 16,
+    "epoch": 0.9142857142857143,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.32162048,
+    "loss": 1.418,
+    "grad_norm": 2.171356678009033,
+    "learning_rate": 0.0002841730623343193
+  },
+  {
+    "step": 17,
+    "epoch": 0.9714285714285714,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.321625088,
+    "loss": 1.4328,
+    "grad_norm": 2.804058790206909,
+    "learning_rate": 0.00028054278615452326
+  },
+  {
+    "step": 18,
+    "epoch": 1.0285714285714285,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405656576,
+    "loss": 2.0652,
+    "grad_norm": 2.6565699577331543,
+    "learning_rate": 0.0002765663339734778
+  },
+  {
+    "step": 19,
+    "epoch": 1.0857142857142856,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.40565504,
+    "loss": 1.3863,
+    "grad_norm": 1.8464845418930054,
+    "learning_rate": 0.00027225425062752165
+  },
+  {
+    "step": 20,
+    "epoch": 1.1428571428571428,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405630464,
+    "loss": 1.3204,
+    "grad_norm": 1.9601354598999023,
+    "learning_rate": 0.0002676179709865066
+  },
+  {
+    "step": 21,
+    "epoch": 1.2,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405638144,
+    "loss": 1.384,
+    "grad_norm": 2.193528413772583,
+    "learning_rate": 0.0002626697896305779
+  },
+  {
+    "step": 22,
+    "epoch": 1.2571428571428571,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405667328,
+    "loss": 1.3968,
+    "grad_norm": 3.0995216369628906,
+    "learning_rate": 0.000257422828247159
+  },
+  {
+    "step": 23,
+    "epoch": 1.3142857142857143,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405696512,
+    "loss": 1.3764,
+    "grad_norm": 3.8095834255218506,
+    "learning_rate": 0.00025189100083459397
+  },
+  {
+    "step": 24,
+    "epoch": 1.3714285714285714,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.40563968,
+    "loss": 1.3585,
+    "grad_norm": 2.3663456439971924,
+    "learning_rate": 0.0002460889768047263
+  },
+  {
+    "step": 25,
+    "epoch": 1.4285714285714286,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.4057088,
+    "loss": 1.3727,
+    "grad_norm": 2.6865601539611816,
+    "learning_rate": 0.00024003214208225522
+  },
+  {
+    "step": 26,
+    "epoch": 1.4857142857142858,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405665792,
+    "loss": 1.3964,
+    "grad_norm": 2.392038583755493,
+    "learning_rate": 0.00023373655830402968
+  },
+  {
+    "step": 27,
+    "epoch": 1.5428571428571427,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.40562432,
+    "loss": 1.4306,
+    "grad_norm": 2.2583298683166504,
+    "learning_rate": 0.00022721892022647462
+  },
+  {
+    "step": 28,
+    "epoch": 1.6,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.4056704,
+    "loss": 1.641,
+    "grad_norm": 4.678727149963379,
+    "learning_rate": 0.000220496511454098
+  },
+  {
+    "step": 29,
+    "epoch": 1.657142857142857,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405665792,
+    "loss": 1.3692,
+    "grad_norm": 1.6019768714904785,
+    "learning_rate": 0.0002135871586064791
+  },
+  {
+    "step": 30,
+    "epoch": 1.7142857142857144,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.40565504,
+    "loss": 1.3711,
+    "grad_norm": 1.569014310836792,
+    "learning_rate": 0.00020650918404527775
+  },
+  {
+    "step": 31,
+    "epoch": 1.7714285714285714,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.40568576,
+    "loss": 1.346,
+    "grad_norm": 1.4607075452804565,
+    "learning_rate": 0.00019928135728662522
+  },
+  {
+    "step": 32,
+    "epoch": 1.8285714285714287,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405694976,
+    "loss": 1.4061,
+    "grad_norm": 2.075629949569702,
+    "learning_rate": 0.00019192284522774142
+  },
+  {
+    "step": 33,
+    "epoch": 1.8857142857142857,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405675008,
+    "loss": 1.4244,
+    "grad_norm": 1.9367634057998657,
+    "learning_rate": 0.00018445316131976934
+  },
+  {
+    "step": 34,
+    "epoch": 1.9428571428571428,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405653504,
+    "loss": 1.3788,
+    "grad_norm": 1.2187618017196655,
+    "learning_rate": 0.00017689211382161034
+  },
+  {
+    "step": 35,
+    "epoch": 2.0,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405541376,
+    "loss": 2.0491,
+    "grad_norm": 1.391384243965149,
+    "learning_rate": 0.00016925975327198266
+  },
+  {
+    "step": 36,
+    "epoch": 2.057142857142857,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.32159744,
+    "loss": 1.3294,
+    "grad_norm": 0.9405258297920227,
+    "learning_rate": 0.00016157631931899697
+  },
+  {
+    "step": 37,
+    "epoch": 2.1142857142857143,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.321606656,
+    "loss": 1.4945,
+    "grad_norm": 2.8952982425689697,
+    "learning_rate": 0.0001538621870482483
+  },
+  {
+    "step": 38,
+    "epoch": 2.1714285714285713,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.321577472,
+    "loss": 1.4347,
+    "grad_norm": 2.3792271614074707,
+    "learning_rate": 0.00014613781295175172
+  },
+  {
+    "step": 39,
+    "epoch": 2.2285714285714286,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.321595904,
+    "loss": 1.4687,
+    "grad_norm": 2.7014079093933105,
+    "learning_rate": 0.00013842368068100303
+  },
+  {
+    "step": 40,
+    "epoch": 2.2857142857142856,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.321572864,
+    "loss": 1.3671,
+    "grad_norm": 1.4931557178497314,
+    "learning_rate": 0.00013074024672801731
+  },
+  {
+    "step": 41,
+    "epoch": 2.342857142857143,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.3215744,
+    "loss": 1.372,
+    "grad_norm": 1.171431541442871,
+    "learning_rate": 0.00012310788617838966
+  },
+  {
+    "step": 42,
+    "epoch": 2.4,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.321603584,
+    "loss": 1.3052,
+    "grad_norm": 1.5233906507492065,
+    "learning_rate": 0.00011554683868023067
+  },
+  {
+    "step": 43,
+    "epoch": 2.4571428571428573,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.321618944,
+    "loss": 1.3462,
+    "grad_norm": 1.46059250831604,
+    "learning_rate": 0.00010807715477225858
+  },
+  {
+    "step": 44,
+    "epoch": 2.5142857142857142,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.321637376,
+    "loss": 1.2995,
+    "grad_norm": 1.066210150718689,
+    "learning_rate": 0.00010071864271337478
+  },
+  {
+    "step": 45,
+    "epoch": 2.571428571428571,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.321591296,
+    "loss": 1.292,
+    "grad_norm": 0.8745197057723999,
+    "learning_rate": 9.34908159547222e-05
+  },
+  {
+    "step": 46,
+    "epoch": 2.6285714285714286,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.321585152,
+    "loss": 1.2877,
+    "grad_norm": 1.2903093099594116,
+    "learning_rate": 8.641284139352091e-05
+  },
+  {
+    "step": 47,
+    "epoch": 2.685714285714286,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.321579008,
+    "loss": 1.2495,
+    "grad_norm": 0.9918281435966492,
+    "learning_rate": 7.950348854590204e-05
+  },
+  {
+    "step": 48,
+    "epoch": 2.742857142857143,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.321583616,
+    "loss": 1.1968,
+    "grad_norm": 0.9396869540214539,
+    "learning_rate": 7.278107977352543e-05
+  },
+  {
+    "step": 49,
+    "epoch": 2.8,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.3215744,
+    "loss": 1.2387,
+    "grad_norm": 0.9950159788131714,
+    "learning_rate": 6.626344169597031e-05
+  },
+  {
+    "step": 50,
+    "epoch": 2.857142857142857,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.321555968,
+    "loss": 1.2554,
+    "grad_norm": 1.2974423170089722,
+    "learning_rate": 5.996785791774478e-05
+  },
+  {
+    "step": 51,
+    "epoch": 2.914285714285714,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.321580544,
+    "loss": 1.2675,
+    "grad_norm": 1.546997308731079,
+    "learning_rate": 5.391102319527373e-05
+  },
+  {
+    "step": 52,
+    "epoch": 2.9714285714285715,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.321608192,
+    "loss": 1.3211,
+    "grad_norm": 1.2576476335525513,
+    "learning_rate": 4.8108999165406026e-05
+  },
+  {
+    "step": 53,
+    "epoch": 3.0285714285714285,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405651968,
+    "loss": 1.7943,
+    "grad_norm": 1.8999438285827637,
+    "learning_rate": 4.257717175284103e-05
+  },
+  {
+    "step": 54,
+    "epoch": 3.085714285714286,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405621248,
+    "loss": 1.2363,
+    "grad_norm": 1.7934603691101074,
+    "learning_rate": 3.733021036942205e-05
+  },
+  {
+    "step": 55,
+    "epoch": 3.142857142857143,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.40565504,
+    "loss": 1.2356,
+    "grad_norm": 1.768385887145996,
+    "learning_rate": 3.238202901349345e-05
+  },
+  {
+    "step": 56,
+    "epoch": 3.2,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405728768,
+    "loss": 1.2013,
+    "grad_norm": 1.5940515995025635,
+    "learning_rate": 2.774574937247831e-05
+  },
+  {
+    "step": 57,
+    "epoch": 3.257142857142857,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405671936,
+    "loss": 1.1569,
+    "grad_norm": 1.4316647052764893,
+    "learning_rate": 2.3433666026522153e-05
+  },
+  {
+    "step": 58,
+    "epoch": 3.314285714285714,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405665792,
+    "loss": 1.0889,
+    "grad_norm": 1.7646082639694214,
+    "learning_rate": 1.945721384547671e-05
+  },
+  {
+    "step": 59,
+    "epoch": 3.3714285714285714,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.40571648,
+    "loss": 1.1441,
+    "grad_norm": 1.6605567932128906,
+    "learning_rate": 1.5826937665680693e-05
+  },
+  {
+    "step": 60,
+    "epoch": 3.4285714285714284,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405642752,
+    "loss": 1.194,
+    "grad_norm": 2.2929937839508057,
+    "learning_rate": 1.2552464326949302e-05
+  },
+  {
+    "step": 61,
+    "epoch": 3.4857142857142858,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405656576,
+    "loss": 1.1754,
+    "grad_norm": 2.0269861221313477,
+    "learning_rate": 9.64247714392597e-06
+  },
+  {
+    "step": 62,
+    "epoch": 3.5428571428571427,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405658112,
+    "loss": 1.1437,
+    "grad_norm": 2.0639264583587646,
+    "learning_rate": 7.104692879492624e-06
+  },
+  {
+    "step": 63,
+    "epoch": 3.6,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.40564736,
+    "loss": 1.181,
+    "grad_norm": 1.9590049982070923,
+    "learning_rate": 4.945841281301943e-06
+  },
+  {
+    "step": 64,
+    "epoch": 3.657142857142857,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405664256,
+    "loss": 1.14,
+    "grad_norm": 2.042527198791504,
+    "learning_rate": 3.1716472356963286e-06
+  },
+  {
+    "step": 65,
+    "epoch": 3.7142857142857144,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.40568576,
+    "loss": 1.1584,
+    "grad_norm": 2.447296380996704,
+    "learning_rate": 1.7868155863384415e-06
+  },
+  {
+    "step": 66,
+    "epoch": 3.7714285714285714,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405676544,
+    "loss": 1.058,
+    "grad_norm": 2.189685583114624,
+    "learning_rate": 7.950186578116413e-07
+  },
+  {
+    "step": 67,
+    "epoch": 3.8285714285714287,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405702656,
+    "loss": 1.1381,
+    "grad_norm": 1.9674383401870728,
+    "learning_rate": 1.988865172754206e-07
+  },
+  {
+    "step": 68,
+    "epoch": 3.8857142857142857,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405653504,
+    "loss": 1.1528,
+    "grad_norm": 2.103919267654419,
+    "learning_rate": 0.0
+  },
+  {
+    "step": 68,
+    "epoch": 3.8857142857142857,
+    "cpu_mem": 2.439118848,
+    "gpu_mem": 1.405653504,
+    "train_runtime": 378.0821,
+    "train_samples_per_second": 11.839,
+    "train_steps_per_second": 0.18,
+    "total_flos": 0.0,
+    "train_loss": 1.4734139126889847
+  }
+]

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-arc_c-r8-a2/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha": 16,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "enabled_mlp": true,
+  "enabled_qkv": [
+    "q",
+    "k",
+    "v"
+  ],
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "mixture": false,
+  "modules_to_preserve_errors": null,
+  "modules_to_quantize": null,
+  "modules_to_save": null,
+  "onnx_export": false,
+  "optimization_level": 3,
+  "orthogonal_init": false,
+  "peft_type": "MARS",
+  "quant_n_bits": 4,
+  "r": 8,
+  "revision": null,
+  "seed": 42,
+  "shared_r": 8,
+  "target_modules": [
+    "o_proj",
+    "k_proj",
+    "down_proj",
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "q_proj"
+  ],
+  "task_type": null
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-arc_c-r8-a2/eval_results.json

@@ -0,0 +1,4 @@
+{
+    "task": "arc_c",
+    "results": 0.5674061433447098
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-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": "mars",
+    "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": 5233536
+  },
+  "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-mars-arc_c-r8-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-mars-opt3-q4-new/TinyLlama_v1.1-mars-arc_c-r8-a2",
+  "seed": 42,
+  "timestamp": "2025-08-30T01:30:22.564244"
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-arc_c-r8-a2/training_logs.json

@@ -0,0 +1,625 @@
+[
+  {
+    "step": 1,
+    "epoch": 0.05714285714285714,
+    "cpu_mem": 2.590892032,
+    "gpu_mem": 1.079811584,
+    "loss": 4.6127,
+    "grad_norm": 190.1083221435547,
+    "learning_rate": 4.285714285714285e-05
+  },
+  {
+    "step": 2,
+    "epoch": 0.11428571428571428,
+    "cpu_mem": 2.59108864,
+    "gpu_mem": 1.121681408,
+    "loss": 4.674,
+    "grad_norm": 195.60543823242188,
+    "learning_rate": 8.57142857142857e-05
+  },
+  {
+    "step": 3,
+    "epoch": 0.17142857142857143,
+    "cpu_mem": 2.59108864,
+    "gpu_mem": 1.121712128,
+    "loss": 1.9423,
+    "grad_norm": 42.55434036254883,
+    "learning_rate": 0.00012857142857142855
+  },
+  {
+    "step": 4,
+    "epoch": 0.22857142857142856,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121678336,
+    "loss": 1.6033,
+    "grad_norm": 18.729658126831055,
+    "learning_rate": 0.0001714285714285714
+  },
+  {
+    "step": 5,
+    "epoch": 0.2857142857142857,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121666048,
+    "loss": 1.502,
+    "grad_norm": 17.119783401489258,
+    "learning_rate": 0.00021428571428571427
+  },
+  {
+    "step": 6,
+    "epoch": 0.34285714285714286,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121729024,
+    "loss": 1.4153,
+    "grad_norm": 8.702713012695312,
+    "learning_rate": 0.0002571428571428571
+  },
+  {
+    "step": 7,
+    "epoch": 0.4,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121735168,
+    "loss": 1.4484,
+    "grad_norm": 9.63271713256836,
+    "learning_rate": 0.0003
+  },
+  {
+    "step": 8,
+    "epoch": 0.45714285714285713,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121693696,
+    "loss": 1.4964,
+    "grad_norm": 8.020484924316406,
+    "learning_rate": 0.00029980111348272456
+  },
+  {
+    "step": 9,
+    "epoch": 0.5142857142857142,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121689088,
+    "loss": 1.3058,
+    "grad_norm": 5.498310089111328,
+    "learning_rate": 0.00029920498134218835
+  },
+  {
+    "step": 10,
+    "epoch": 0.5714285714285714,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121678336,
+    "loss": 1.5191,
+    "grad_norm": 12.341209411621094,
+    "learning_rate": 0.0002982131844136615
+  },
+  {
+    "step": 11,
+    "epoch": 0.6285714285714286,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121689088,
+    "loss": 1.4231,
+    "grad_norm": 8.037294387817383,
+    "learning_rate": 0.0002968283527643036
+  },
+  {
+    "step": 12,
+    "epoch": 0.6857142857142857,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121713664,
+    "loss": 1.4298,
+    "grad_norm": 5.25429630279541,
+    "learning_rate": 0.000295054158718698
+  },
+  {
+    "step": 13,
+    "epoch": 0.7428571428571429,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121713664,
+    "loss": 1.3615,
+    "grad_norm": 7.831629753112793,
+    "learning_rate": 0.00029289530712050735
+  },
+  {
+    "step": 14,
+    "epoch": 0.8,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.12166144,
+    "loss": 1.6705,
+    "grad_norm": 10.625100135803223,
+    "learning_rate": 0.000290357522856074
+  },
+  {
+    "step": 15,
+    "epoch": 0.8571428571428571,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121736704,
+    "loss": 1.6489,
+    "grad_norm": 9.55422306060791,
+    "learning_rate": 0.0002874475356730507
+  },
+  {
+    "step": 16,
+    "epoch": 0.9142857142857143,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.12173056,
+    "loss": 1.4015,
+    "grad_norm": 2.770230770111084,
+    "learning_rate": 0.0002841730623343193
+  },
+  {
+    "step": 17,
+    "epoch": 0.9714285714285714,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121735168,
+    "loss": 1.4158,
+    "grad_norm": 4.04749870300293,
+    "learning_rate": 0.00028054278615452326
+  },
+  {
+    "step": 18,
+    "epoch": 1.0285714285714285,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142631936,
+    "loss": 2.0539,
+    "grad_norm": 5.462876319885254,
+    "learning_rate": 0.0002765663339734778
+  },
+  {
+    "step": 19,
+    "epoch": 1.0857142857142856,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.1426304,
+    "loss": 1.4073,
+    "grad_norm": 4.310235023498535,
+    "learning_rate": 0.00027225425062752165
+  },
+  {
+    "step": 20,
+    "epoch": 1.1428571428571428,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142605824,
+    "loss": 1.3455,
+    "grad_norm": 4.525611877441406,
+    "learning_rate": 0.0002676179709865066
+  },
+  {
+    "step": 21,
+    "epoch": 1.2,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142613504,
+    "loss": 1.4146,
+    "grad_norm": 4.8981218338012695,
+    "learning_rate": 0.0002626697896305779
+  },
+  {
+    "step": 22,
+    "epoch": 1.2571428571428571,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142642688,
+    "loss": 1.3934,
+    "grad_norm": 6.351171493530273,
+    "learning_rate": 0.000257422828247159
+  },
+  {
+    "step": 23,
+    "epoch": 1.3142857142857143,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142671872,
+    "loss": 1.3354,
+    "grad_norm": 5.1450886726379395,
+    "learning_rate": 0.00025189100083459397
+  },
+  {
+    "step": 24,
+    "epoch": 1.3714285714285714,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.14261504,
+    "loss": 1.394,
+    "grad_norm": 6.1006340980529785,
+    "learning_rate": 0.0002460889768047263
+  },
+  {
+    "step": 25,
+    "epoch": 1.4285714285714286,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.14268416,
+    "loss": 1.3449,
+    "grad_norm": 4.99368143081665,
+    "learning_rate": 0.00024003214208225522
+  },
+  {
+    "step": 26,
+    "epoch": 1.4857142857142858,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142641152,
+    "loss": 1.3966,
+    "grad_norm": 5.3322882652282715,
+    "learning_rate": 0.00023373655830402968
+  },
+  {
+    "step": 27,
+    "epoch": 1.5428571428571427,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.14259968,
+    "loss": 1.4423,
+    "grad_norm": 5.347881317138672,
+    "learning_rate": 0.00022721892022647462
+  },
+  {
+    "step": 28,
+    "epoch": 1.6,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.14264576,
+    "loss": 1.5757,
+    "grad_norm": 7.814599514007568,
+    "learning_rate": 0.000220496511454098
+  },
+  {
+    "step": 29,
+    "epoch": 1.657142857142857,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142641152,
+    "loss": 1.353,
+    "grad_norm": 2.587792158126831,
+    "learning_rate": 0.0002135871586064791
+  },
+  {
+    "step": 30,
+    "epoch": 1.7142857142857144,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.1426304,
+    "loss": 1.3714,
+    "grad_norm": 3.4897141456604004,
+    "learning_rate": 0.00020650918404527775
+  },
+  {
+    "step": 31,
+    "epoch": 1.7714285714285714,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.14266112,
+    "loss": 1.3522,
+    "grad_norm": 2.2583186626434326,
+    "learning_rate": 0.00019928135728662522
+  },
+  {
+    "step": 32,
+    "epoch": 1.8285714285714287,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142670336,
+    "loss": 1.3582,
+    "grad_norm": 2.4103989601135254,
+    "learning_rate": 0.00019192284522774142
+  },
+  {
+    "step": 33,
+    "epoch": 1.8857142857142857,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142650368,
+    "loss": 1.3951,
+    "grad_norm": 3.2258849143981934,
+    "learning_rate": 0.00018445316131976934
+  },
+  {
+    "step": 34,
+    "epoch": 1.9428571428571428,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142628864,
+    "loss": 1.3668,
+    "grad_norm": 2.021745204925537,
+    "learning_rate": 0.00017689211382161034
+  },
+  {
+    "step": 35,
+    "epoch": 2.0,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142516736,
+    "loss": 2.0591,
+    "grad_norm": 2.1437292098999023,
+    "learning_rate": 0.00016925975327198266
+  },
+  {
+    "step": 36,
+    "epoch": 2.057142857142857,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.12170752,
+    "loss": 1.3225,
+    "grad_norm": 1.7468737363815308,
+    "learning_rate": 0.00016157631931899697
+  },
+  {
+    "step": 37,
+    "epoch": 2.1142857142857143,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121716736,
+    "loss": 1.4179,
+    "grad_norm": 3.716902494430542,
+    "learning_rate": 0.0001538621870482483
+  },
+  {
+    "step": 38,
+    "epoch": 2.1714285714285713,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121687552,
+    "loss": 1.3573,
+    "grad_norm": 2.5589308738708496,
+    "learning_rate": 0.00014613781295175172
+  },
+  {
+    "step": 39,
+    "epoch": 2.2285714285714286,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121705984,
+    "loss": 1.3358,
+    "grad_norm": 2.8327057361602783,
+    "learning_rate": 0.00013842368068100303
+  },
+  {
+    "step": 40,
+    "epoch": 2.2857142857142856,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121682944,
+    "loss": 1.3182,
+    "grad_norm": 2.0415289402008057,
+    "learning_rate": 0.00013074024672801731
+  },
+  {
+    "step": 41,
+    "epoch": 2.342857142857143,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.12168448,
+    "loss": 1.3512,
+    "grad_norm": 2.6037545204162598,
+    "learning_rate": 0.00012310788617838966
+  },
+  {
+    "step": 42,
+    "epoch": 2.4,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121713664,
+    "loss": 1.3009,
+    "grad_norm": 3.407050132751465,
+    "learning_rate": 0.00011554683868023067
+  },
+  {
+    "step": 43,
+    "epoch": 2.4571428571428573,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121729024,
+    "loss": 1.375,
+    "grad_norm": 4.440658092498779,
+    "learning_rate": 0.00010807715477225858
+  },
+  {
+    "step": 44,
+    "epoch": 2.5142857142857142,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121747456,
+    "loss": 1.2674,
+    "grad_norm": 2.6665470600128174,
+    "learning_rate": 0.00010071864271337478
+  },
+  {
+    "step": 45,
+    "epoch": 2.571428571428571,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121701376,
+    "loss": 1.2365,
+    "grad_norm": 3.8944313526153564,
+    "learning_rate": 9.34908159547222e-05
+  },
+  {
+    "step": 46,
+    "epoch": 2.6285714285714286,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121695232,
+    "loss": 1.2408,
+    "grad_norm": 2.97174072265625,
+    "learning_rate": 8.641284139352091e-05
+  },
+  {
+    "step": 47,
+    "epoch": 2.685714285714286,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121689088,
+    "loss": 1.24,
+    "grad_norm": 3.1578757762908936,
+    "learning_rate": 7.950348854590204e-05
+  },
+  {
+    "step": 48,
+    "epoch": 2.742857142857143,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121693696,
+    "loss": 1.1557,
+    "grad_norm": 3.0455143451690674,
+    "learning_rate": 7.278107977352543e-05
+  },
+  {
+    "step": 49,
+    "epoch": 2.8,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.12168448,
+    "loss": 1.2344,
+    "grad_norm": 3.0066888332366943,
+    "learning_rate": 6.626344169597031e-05
+  },
+  {
+    "step": 50,
+    "epoch": 2.857142857142857,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121666048,
+    "loss": 1.2225,
+    "grad_norm": 3.4496467113494873,
+    "learning_rate": 5.996785791774478e-05
+  },
+  {
+    "step": 51,
+    "epoch": 2.914285714285714,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121690624,
+    "loss": 1.2478,
+    "grad_norm": 4.3415117263793945,
+    "learning_rate": 5.391102319527373e-05
+  },
+  {
+    "step": 52,
+    "epoch": 2.9714285714285715,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.121718272,
+    "loss": 1.2959,
+    "grad_norm": 3.4093315601348877,
+    "learning_rate": 4.8108999165406026e-05
+  },
+  {
+    "step": 53,
+    "epoch": 3.0285714285714285,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142627328,
+    "loss": 1.7767,
+    "grad_norm": 5.994507312774658,
+    "learning_rate": 4.257717175284103e-05
+  },
+  {
+    "step": 54,
+    "epoch": 3.085714285714286,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142596608,
+    "loss": 1.2086,
+    "grad_norm": 4.536657333374023,
+    "learning_rate": 3.733021036942205e-05
+  },
+  {
+    "step": 55,
+    "epoch": 3.142857142857143,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.1426304,
+    "loss": 1.1557,
+    "grad_norm": 4.9032063484191895,
+    "learning_rate": 3.238202901349345e-05
+  },
+  {
+    "step": 56,
+    "epoch": 3.2,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142704128,
+    "loss": 1.1267,
+    "grad_norm": 4.643110275268555,
+    "learning_rate": 2.774574937247831e-05
+  },
+  {
+    "step": 57,
+    "epoch": 3.257142857142857,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142647296,
+    "loss": 1.159,
+    "grad_norm": 3.9950804710388184,
+    "learning_rate": 2.3433666026522153e-05
+  },
+  {
+    "step": 58,
+    "epoch": 3.314285714285714,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142641152,
+    "loss": 1.035,
+    "grad_norm": 4.866928577423096,
+    "learning_rate": 1.945721384547671e-05
+  },
+  {
+    "step": 59,
+    "epoch": 3.3714285714285714,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.14269184,
+    "loss": 1.0495,
+    "grad_norm": 4.91159725189209,
+    "learning_rate": 1.5826937665680693e-05
+  },
+  {
+    "step": 60,
+    "epoch": 3.4285714285714284,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142618112,
+    "loss": 1.128,
+    "grad_norm": 6.490297794342041,
+    "learning_rate": 1.2552464326949302e-05
+  },
+  {
+    "step": 61,
+    "epoch": 3.4857142857142858,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142631936,
+    "loss": 1.0631,
+    "grad_norm": 5.946484088897705,
+    "learning_rate": 9.64247714392597e-06
+  },
+  {
+    "step": 62,
+    "epoch": 3.5428571428571427,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142633472,
+    "loss": 1.122,
+    "grad_norm": 6.1243577003479,
+    "learning_rate": 7.104692879492624e-06
+  },
+  {
+    "step": 63,
+    "epoch": 3.6,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.14262272,
+    "loss": 1.1157,
+    "grad_norm": 5.4304728507995605,
+    "learning_rate": 4.945841281301943e-06
+  },
+  {
+    "step": 64,
+    "epoch": 3.657142857142857,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142639616,
+    "loss": 1.0507,
+    "grad_norm": 5.880957126617432,
+    "learning_rate": 3.1716472356963286e-06
+  },
+  {
+    "step": 65,
+    "epoch": 3.7142857142857144,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.14266112,
+    "loss": 1.0811,
+    "grad_norm": 6.808589458465576,
+    "learning_rate": 1.7868155863384415e-06
+  },
+  {
+    "step": 66,
+    "epoch": 3.7714285714285714,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142651904,
+    "loss": 0.9973,
+    "grad_norm": 6.240514278411865,
+    "learning_rate": 7.950186578116413e-07
+  },
+  {
+    "step": 67,
+    "epoch": 3.8285714285714287,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142678016,
+    "loss": 1.0386,
+    "grad_norm": 5.9608283042907715,
+    "learning_rate": 1.988865172754206e-07
+  },
+  {
+    "step": 68,
+    "epoch": 3.8857142857142857,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142628864,
+    "loss": 1.0538,
+    "grad_norm": 6.074984550476074,
+    "learning_rate": 0.0
+  },
+  {
+    "step": 68,
+    "epoch": 3.8857142857142857,
+    "cpu_mem": 2.591285248,
+    "gpu_mem": 1.142628864,
+    "train_runtime": 374.9512,
+    "train_samples_per_second": 11.938,
+    "train_steps_per_second": 0.181,
+    "total_flos": 0.0,
+    "train_loss": 1.4457515811218935
+  }
+]

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-arc_e-r32-a2/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha": 64,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "enabled_mlp": true,
+  "enabled_qkv": [
+    "q",
+    "k",
+    "v"
+  ],
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "mixture": false,
+  "modules_to_preserve_errors": null,
+  "modules_to_quantize": null,
+  "modules_to_save": null,
+  "onnx_export": false,
+  "optimization_level": 3,
+  "orthogonal_init": false,
+  "peft_type": "MARS",
+  "quant_n_bits": 4,
+  "r": 32,
+  "revision": null,
+  "seed": 42,
+  "shared_r": 32,
+  "target_modules": [
+    "o_proj",
+    "k_proj",
+    "down_proj",
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "q_proj"
+  ],
+  "task_type": null
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-arc_e-r32-a2/eval_results.json

@@ -0,0 +1,4 @@
+{
+    "task": "arc_e",
+    "results": 0.7121212121212122
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-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": "mars",
+    "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": 21018624
+  },
+  "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-mars-arc_e-r32-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-mars-opt3-q4-new/TinyLlama_v1.1-mars-arc_e-r32-a2",
+  "seed": 42,
+  "timestamp": "2025-08-30T07:55:18.577303"
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-arc_e-r32-a2/training_logs.json

@@ -0,0 +1,1273 @@
+[
+  {
+    "step": 1,
+    "epoch": 0.028169014084507043,
+    "cpu_mem": 2.556440576,
+    "gpu_mem": 1.143415296,
+    "loss": 4.6689,
+    "grad_norm": 80.02104187011719,
+    "learning_rate": 2.1428571428571425e-05
+  },
+  {
+    "step": 2,
+    "epoch": 0.056338028169014086,
+    "cpu_mem": 2.556637184,
+    "gpu_mem": 1.311625216,
+    "loss": 4.5086,
+    "grad_norm": 83.96923065185547,
+    "learning_rate": 4.285714285714285e-05
+  },
+  {
+    "step": 3,
+    "epoch": 0.08450704225352113,
+    "cpu_mem": 2.556833792,
+    "gpu_mem": 1.311603712,
+    "loss": 3.0059,
+    "grad_norm": 49.818973541259766,
+    "learning_rate": 6.428571428571427e-05
+  },
+  {
+    "step": 4,
+    "epoch": 0.11267605633802817,
+    "cpu_mem": 2.556833792,
+    "gpu_mem": 1.311582208,
+    "loss": 2.0779,
+    "grad_norm": 18.78441619873047,
+    "learning_rate": 8.57142857142857e-05
+  },
+  {
+    "step": 5,
+    "epoch": 0.14084507042253522,
+    "cpu_mem": 2.5570304,
+    "gpu_mem": 1.31162368,
+    "loss": 1.5397,
+    "grad_norm": 5.960499286651611,
+    "learning_rate": 0.00010714285714285714
+  },
+  {
+    "step": 6,
+    "epoch": 0.16901408450704225,
+    "cpu_mem": 2.557227008,
+    "gpu_mem": 1.311599104,
+    "loss": 1.4567,
+    "grad_norm": 8.546557426452637,
+    "learning_rate": 0.00012857142857142855
+  },
+  {
+    "step": 7,
+    "epoch": 0.19718309859154928,
+    "cpu_mem": 2.557227008,
+    "gpu_mem": 1.311622144,
+    "loss": 1.6186,
+    "grad_norm": 14.599489212036133,
+    "learning_rate": 0.00015
+  },
+  {
+    "step": 8,
+    "epoch": 0.22535211267605634,
+    "cpu_mem": 2.557227008,
+    "gpu_mem": 1.311580672,
+    "loss": 1.4127,
+    "grad_norm": 6.496516227722168,
+    "learning_rate": 0.0001714285714285714
+  },
+  {
+    "step": 9,
+    "epoch": 0.2535211267605634,
+    "cpu_mem": 2.557227008,
+    "gpu_mem": 1.311582208,
+    "loss": 1.4107,
+    "grad_norm": 4.811722278594971,
+    "learning_rate": 0.00019285714285714286
+  },
+  {
+    "step": 10,
+    "epoch": 0.28169014084507044,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.3115776,
+    "loss": 1.531,
+    "grad_norm": 7.236036777496338,
+    "learning_rate": 0.00021428571428571427
+  },
+  {
+    "step": 11,
+    "epoch": 0.30985915492957744,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311655936,
+    "loss": 1.3515,
+    "grad_norm": 3.067713499069214,
+    "learning_rate": 0.00023571428571428569
+  },
+  {
+    "step": 12,
+    "epoch": 0.3380281690140845,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311629824,
+    "loss": 1.4038,
+    "grad_norm": 3.7601993083953857,
+    "learning_rate": 0.0002571428571428571
+  },
+  {
+    "step": 13,
+    "epoch": 0.36619718309859156,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311580672,
+    "loss": 1.3711,
+    "grad_norm": 2.6803042888641357,
+    "learning_rate": 0.00027857142857142854
+  },
+  {
+    "step": 14,
+    "epoch": 0.39436619718309857,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311602176,
+    "loss": 1.3275,
+    "grad_norm": 1.6125613451004028,
+    "learning_rate": 0.0003
+  },
+  {
+    "step": 15,
+    "epoch": 0.4225352112676056,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311579136,
+    "loss": 1.3446,
+    "grad_norm": 1.7154477834701538,
+    "learning_rate": 0.0002999533773001224
+  },
+  {
+    "step": 16,
+    "epoch": 0.4507042253521127,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311583744,
+    "loss": 1.375,
+    "grad_norm": 1.3047140836715698,
+    "learning_rate": 0.0002998135381828383
+  },
+  {
+    "step": 17,
+    "epoch": 0.4788732394366197,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311620608,
+    "loss": 1.3274,
+    "grad_norm": 1.3025691509246826,
+    "learning_rate": 0.00029958056957717696
+  },
+  {
+    "step": 18,
+    "epoch": 0.5070422535211268,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.31163136,
+    "loss": 1.4244,
+    "grad_norm": 3.0251400470733643,
+    "learning_rate": 0.0002992546163048102
+  },
+  {
+    "step": 19,
+    "epoch": 0.5352112676056338,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311574528,
+    "loss": 1.3444,
+    "grad_norm": 2.2817976474761963,
+    "learning_rate": 0.0002988358809900258
+  },
+  {
+    "step": 20,
+    "epoch": 0.5633802816901409,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311645184,
+    "loss": 1.4019,
+    "grad_norm": 2.1219711303710938,
+    "learning_rate": 0.0002983246239337692
+  },
+  {
+    "step": 21,
+    "epoch": 0.5915492957746479,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311643648,
+    "loss": 1.331,
+    "grad_norm": 1.734360694885254,
+    "learning_rate": 0.0002977211629518312
+  },
+  {
+    "step": 22,
+    "epoch": 0.6197183098591549,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.31160064,
+    "loss": 1.3231,
+    "grad_norm": 1.4685757160186768,
+    "learning_rate": 0.00029702587317728153
+  },
+  {
+    "step": 23,
+    "epoch": 0.647887323943662,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311617536,
+    "loss": 1.3441,
+    "grad_norm": 1.3601280450820923,
+    "learning_rate": 0.0002962391868272735
+  },
+  {
+    "step": 24,
+    "epoch": 0.676056338028169,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311574528,
+    "loss": 1.3631,
+    "grad_norm": 1.5260316133499146,
+    "learning_rate": 0.00029536159293436166
+  },
+  {
+    "step": 25,
+    "epoch": 0.704225352112676,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311603712,
+    "loss": 1.3865,
+    "grad_norm": 1.4043406248092651,
+    "learning_rate": 0.00029439363704250176
+  },
+  {
+    "step": 26,
+    "epoch": 0.7323943661971831,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311583744,
+    "loss": 1.4606,
+    "grad_norm": 1.6535191535949707,
+    "learning_rate": 0.00029333592086792107
+  },
+  {
+    "step": 27,
+    "epoch": 0.7605633802816901,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311609856,
+    "loss": 1.3359,
+    "grad_norm": 0.8193792700767517,
+    "learning_rate": 0.0002921891019250697
+  },
+  {
+    "step": 28,
+    "epoch": 0.7887323943661971,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311609856,
+    "loss": 1.3871,
+    "grad_norm": 1.0423256158828735,
+    "learning_rate": 0.0002909538931178862
+  },
+  {
+    "step": 29,
+    "epoch": 0.8169014084507042,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311588352,
+    "loss": 1.2794,
+    "grad_norm": 1.2377030849456787,
+    "learning_rate": 0.00028963106229663063
+  },
+  {
+    "step": 30,
+    "epoch": 0.8450704225352113,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311579136,
+    "loss": 1.3537,
+    "grad_norm": 1.0612547397613525,
+    "learning_rate": 0.00028822143178056114
+  },
+  {
+    "step": 31,
+    "epoch": 0.8732394366197183,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311597568,
+    "loss": 1.3698,
+    "grad_norm": 0.8752574920654297,
+    "learning_rate": 0.00028672587784675096
+  },
+  {
+    "step": 32,
+    "epoch": 0.9014084507042254,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311620608,
+    "loss": 1.3323,
+    "grad_norm": 1.2664260864257812,
+    "learning_rate": 0.0002851453301853628
+  },
+  {
+    "step": 33,
+    "epoch": 0.9295774647887324,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311617536,
+    "loss": 1.375,
+    "grad_norm": 1.1751697063446045,
+    "learning_rate": 0.00028348077132172027
+  },
+  {
+    "step": 34,
+    "epoch": 0.9577464788732394,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311620608,
+    "loss": 1.4226,
+    "grad_norm": 1.6566500663757324,
+    "learning_rate": 0.0002817332360055343
+  },
+  {
+    "step": 35,
+    "epoch": 0.9859154929577465,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311602176,
+    "loss": 1.3242,
+    "grad_norm": 1.2318401336669922,
+    "learning_rate": 0.0002799038105676658
+  },
+  {
+    "step": 36,
+    "epoch": 1.0140845070422535,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395667456,
+    "loss": 2.0355,
+    "grad_norm": 2.13344407081604,
+    "learning_rate": 0.0002779936322448233
+  },
+  {
+    "step": 37,
+    "epoch": 1.0422535211267605,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395672064,
+    "loss": 1.3796,
+    "grad_norm": 1.7476377487182617,
+    "learning_rate": 0.0002760038884726157
+  },
+  {
+    "step": 38,
+    "epoch": 1.0704225352112675,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.39565056,
+    "loss": 1.2405,
+    "grad_norm": 1.125243902206421,
+    "learning_rate": 0.00027393581614739923
+  },
+  {
+    "step": 39,
+    "epoch": 1.0985915492957747,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395639808,
+    "loss": 1.3792,
+    "grad_norm": 2.231471538543701,
+    "learning_rate": 0.0002717907008573785
+  },
+  {
+    "step": 40,
+    "epoch": 1.1267605633802817,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395702784,
+    "loss": 1.3537,
+    "grad_norm": 1.4337620735168457,
+    "learning_rate": 0.0002695698760834384
+  },
+  {
+    "step": 41,
+    "epoch": 1.1549295774647887,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395662848,
+    "loss": 1.3459,
+    "grad_norm": 1.9487059116363525,
+    "learning_rate": 0.00026727472237020447
+  },
+  {
+    "step": 42,
+    "epoch": 1.1830985915492958,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395705856,
+    "loss": 1.3556,
+    "grad_norm": 1.2340912818908691,
+    "learning_rate": 0.00026490666646784665
+  },
+  {
+    "step": 43,
+    "epoch": 1.2112676056338028,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395655168,
+    "loss": 1.3948,
+    "grad_norm": 1.187638521194458,
+    "learning_rate": 0.0002624671804451601
+  },
+  {
+    "step": 44,
+    "epoch": 1.2394366197183098,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.39571968,
+    "loss": 1.3685,
+    "grad_norm": 1.508293628692627,
+    "learning_rate": 0.0002599577807744739
+  },
+  {
+    "step": 45,
+    "epoch": 1.267605633802817,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395687424,
+    "loss": 1.3882,
+    "grad_norm": 0.9527984261512756,
+    "learning_rate": 0.0002573800273889577
+  },
+  {
+    "step": 46,
+    "epoch": 1.295774647887324,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395692032,
+    "loss": 1.3874,
+    "grad_norm": 1.313702940940857,
+    "learning_rate": 0.0002547355227129109
+  },
+  {
+    "step": 47,
+    "epoch": 1.323943661971831,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395638272,
+    "loss": 1.4268,
+    "grad_norm": 2.3709449768066406,
+    "learning_rate": 0.00025202591066563786
+  },
+  {
+    "step": 48,
+    "epoch": 1.352112676056338,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395652096,
+    "loss": 1.3532,
+    "grad_norm": 1.4290915727615356,
+    "learning_rate": 0.0002492528756395289
+  },
+  {
+    "step": 49,
+    "epoch": 1.380281690140845,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395641344,
+    "loss": 1.3,
+    "grad_norm": 0.8550411462783813,
+    "learning_rate": 0.0002464181414529809
+  },
+  {
+    "step": 50,
+    "epoch": 1.408450704225352,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395655168,
+    "loss": 1.324,
+    "grad_norm": 1.0939958095550537,
+    "learning_rate": 0.00024352347027881003
+  },
+  {
+    "step": 51,
+    "epoch": 1.436619718309859,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395707392,
+    "loss": 1.3759,
+    "grad_norm": 1.5617879629135132,
+    "learning_rate": 0.0002405706615488216
+  },
+  {
+    "step": 52,
+    "epoch": 1.4647887323943662,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395655168,
+    "loss": 1.3453,
+    "grad_norm": 1.4145621061325073,
+    "learning_rate": 0.00023756155083521846
+  },
+  {
+    "step": 53,
+    "epoch": 1.4929577464788732,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395724288,
+    "loss": 1.3062,
+    "grad_norm": 0.8992661833763123,
+    "learning_rate": 0.00023449800870954326
+  },
+  {
+    "step": 54,
+    "epoch": 1.5211267605633803,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395692032,
+    "loss": 1.3001,
+    "grad_norm": 1.7878972291946411,
+    "learning_rate": 0.0002313819395798639
+  },
+  {
+    "step": 55,
+    "epoch": 1.5492957746478875,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395701248,
+    "loss": 1.4296,
+    "grad_norm": 2.7013356685638428,
+    "learning_rate": 0.0002282152805069247
+  },
+  {
+    "step": 56,
+    "epoch": 1.5774647887323945,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395676672,
+    "loss": 1.3501,
+    "grad_norm": 1.583625078201294,
+    "learning_rate": 0.000225
+  },
+  {
+    "step": 57,
+    "epoch": 1.6056338028169015,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395710464,
+    "loss": 1.2723,
+    "grad_norm": 0.6726813316345215,
+    "learning_rate": 0.00022173809679319772
+  },
+  {
+    "step": 58,
+    "epoch": 1.6338028169014085,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395692032,
+    "loss": 1.2848,
+    "grad_norm": 0.9176018834114075,
+    "learning_rate": 0.00021843159860297442
+  },
+  {
+    "step": 59,
+    "epoch": 1.6619718309859155,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395678208,
+    "loss": 1.3193,
+    "grad_norm": 1.1201019287109375,
+    "learning_rate": 0.00021508256086763368
+  },
+  {
+    "step": 60,
+    "epoch": 1.6901408450704225,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395716608,
+    "loss": 1.3011,
+    "grad_norm": 1.3688689470291138,
+    "learning_rate": 0.00021169306546959174
+  },
+  {
+    "step": 61,
+    "epoch": 1.7183098591549295,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395649024,
+    "loss": 1.2904,
+    "grad_norm": 1.38422429561615,
+    "learning_rate": 0.0002082652194412042
+  },
+  {
+    "step": 62,
+    "epoch": 1.7464788732394365,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.39569664,
+    "loss": 1.3664,
+    "grad_norm": 1.3409178256988525,
+    "learning_rate": 0.00020480115365495926
+  },
+  {
+    "step": 63,
+    "epoch": 1.7746478873239435,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395645952,
+    "loss": 1.3458,
+    "grad_norm": 1.7384254932403564,
+    "learning_rate": 0.00020130302149885031
+  },
+  {
+    "step": 64,
+    "epoch": 1.8028169014084507,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395695104,
+    "loss": 1.3653,
+    "grad_norm": 1.7220115661621094,
+    "learning_rate": 0.00019777299753775265
+  },
+  {
+    "step": 65,
+    "epoch": 1.8309859154929577,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395693568,
+    "loss": 1.3209,
+    "grad_norm": 1.4237152338027954,
+    "learning_rate": 0.00019421327616163563
+  },
+  {
+    "step": 66,
+    "epoch": 1.8591549295774648,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395712,
+    "loss": 1.2984,
+    "grad_norm": 1.356772780418396,
+    "learning_rate": 0.00019062607022145078
+  },
+  {
+    "step": 67,
+    "epoch": 1.887323943661972,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395653632,
+    "loss": 1.3236,
+    "grad_norm": 1.1635974645614624,
+    "learning_rate": 0.00018701360965354402
+  },
+  {
+    "step": 68,
+    "epoch": 1.915492957746479,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.39566592,
+    "loss": 1.348,
+    "grad_norm": 0.9813856482505798,
+    "learning_rate": 0.00018337814009344714
+  },
+  {
+    "step": 69,
+    "epoch": 1.943661971830986,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395690496,
+    "loss": 1.3142,
+    "grad_norm": 1.3948644399642944,
+    "learning_rate": 0.0001797219214799096
+  },
+  {
+    "step": 70,
+    "epoch": 1.971830985915493,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395667456,
+    "loss": 1.242,
+    "grad_norm": 1.0734961032867432,
+    "learning_rate": 0.00017604722665003956
+  },
+  {
+    "step": 71,
+    "epoch": 2.0,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395503104,
+    "loss": 1.922,
+    "grad_norm": 2.428542375564575,
+    "learning_rate": 0.00017235633992642615
+  },
+  {
+    "step": 72,
+    "epoch": 2.028169014084507,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311612928,
+    "loss": 1.2839,
+    "grad_norm": 1.276829481124878,
+    "learning_rate": 0.00016865155569712278
+  },
+  {
+    "step": 73,
+    "epoch": 2.056338028169014,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311576064,
+    "loss": 1.241,
+    "grad_norm": 1.3086708784103394,
+    "learning_rate": 0.0001649351769893725
+  },
+  {
+    "step": 74,
+    "epoch": 2.084507042253521,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311635968,
+    "loss": 1.2416,
+    "grad_norm": 1.5263572931289673,
+    "learning_rate": 0.00016120951403796364
+  },
+  {
+    "step": 75,
+    "epoch": 2.112676056338028,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311603712,
+    "loss": 1.2897,
+    "grad_norm": 2.197709560394287,
+    "learning_rate": 0.00015747688284910457
+  },
+  {
+    "step": 76,
+    "epoch": 2.140845070422535,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311614464,
+    "loss": 1.1855,
+    "grad_norm": 1.6668955087661743,
+    "learning_rate": 0.00015373960376071093
+  },
+  {
+    "step": 77,
+    "epoch": 2.169014084507042,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311651328,
+    "loss": 1.2452,
+    "grad_norm": 1.952316403388977,
+    "learning_rate": 0.00015
+  },
+  {
+    "step": 78,
+    "epoch": 2.1971830985915495,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311635968,
+    "loss": 1.2823,
+    "grad_norm": 2.414081573486328,
+    "learning_rate": 0.00014626039623928907
+  },
+  {
+    "step": 79,
+    "epoch": 2.2253521126760565,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311586816,
+    "loss": 1.1377,
+    "grad_norm": 2.1895029544830322,
+    "learning_rate": 0.0001425231171508954
+  },
+  {
+    "step": 80,
+    "epoch": 2.2535211267605635,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.31163136,
+    "loss": 1.244,
+    "grad_norm": 1.7564090490341187,
+    "learning_rate": 0.00013879048596203636
+  },
+  {
+    "step": 81,
+    "epoch": 2.2816901408450705,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311617536,
+    "loss": 1.3114,
+    "grad_norm": 2.8045568466186523,
+    "learning_rate": 0.0001350648230106275
+  },
+  {
+    "step": 82,
+    "epoch": 2.3098591549295775,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.31158528,
+    "loss": 1.283,
+    "grad_norm": 2.8252458572387695,
+    "learning_rate": 0.00013134844430287725
+  },
+  {
+    "step": 83,
+    "epoch": 2.3380281690140845,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311635968,
+    "loss": 1.2006,
+    "grad_norm": 2.767235517501831,
+    "learning_rate": 0.0001276436600735738
+  },
+  {
+    "step": 84,
+    "epoch": 2.3661971830985915,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311574528,
+    "loss": 1.2354,
+    "grad_norm": 2.006580352783203,
+    "learning_rate": 0.00012395277334996044
+  },
+  {
+    "step": 85,
+    "epoch": 2.3943661971830985,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311620608,
+    "loss": 1.2189,
+    "grad_norm": 2.486454486846924,
+    "learning_rate": 0.00012027807852009038
+  },
+  {
+    "step": 86,
+    "epoch": 2.4225352112676055,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311574528,
+    "loss": 1.1596,
+    "grad_norm": 1.7841145992279053,
+    "learning_rate": 0.00011662185990655284
+  },
+  {
+    "step": 87,
+    "epoch": 2.4507042253521125,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311605248,
+    "loss": 1.1458,
+    "grad_norm": 2.067601203918457,
+    "learning_rate": 0.00011298639034645593
+  },
+  {
+    "step": 88,
+    "epoch": 2.4788732394366195,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311580672,
+    "loss": 1.1451,
+    "grad_norm": 2.60972261428833,
+    "learning_rate": 0.00010937392977854923
+  },
+  {
+    "step": 89,
+    "epoch": 2.507042253521127,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311634432,
+    "loss": 1.0901,
+    "grad_norm": 1.8266022205352783,
+    "learning_rate": 0.00010578672383836435
+  },
+  {
+    "step": 90,
+    "epoch": 2.535211267605634,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311616,
+    "loss": 1.1689,
+    "grad_norm": 2.385910987854004,
+    "learning_rate": 0.00010222700246224735
+  },
+  {
+    "step": 91,
+    "epoch": 2.563380281690141,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311565312,
+    "loss": 1.1882,
+    "grad_norm": 2.8446080684661865,
+    "learning_rate": 9.869697850114969e-05
+  },
+  {
+    "step": 92,
+    "epoch": 2.591549295774648,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311589888,
+    "loss": 1.2837,
+    "grad_norm": 4.627291202545166,
+    "learning_rate": 9.519884634504074e-05
+  },
+  {
+    "step": 93,
+    "epoch": 2.619718309859155,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.31159296,
+    "loss": 1.1516,
+    "grad_norm": 3.087660551071167,
+    "learning_rate": 9.17347805587958e-05
+  },
+  {
+    "step": 94,
+    "epoch": 2.647887323943662,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.31158528,
+    "loss": 1.1214,
+    "grad_norm": 3.74629282951355,
+    "learning_rate": 8.830693453040829e-05
+  },
+  {
+    "step": 95,
+    "epoch": 2.676056338028169,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.31162368,
+    "loss": 1.1599,
+    "grad_norm": 3.486809730529785,
+    "learning_rate": 8.491743913236628e-05
+  },
+  {
+    "step": 96,
+    "epoch": 2.704225352112676,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311632896,
+    "loss": 1.0801,
+    "grad_norm": 3.3228325843811035,
+    "learning_rate": 8.156840139702554e-05
+  },
+  {
+    "step": 97,
+    "epoch": 2.732394366197183,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311576064,
+    "loss": 1.0717,
+    "grad_norm": 3.0872833728790283,
+    "learning_rate": 7.82619032068023e-05
+  },
+  {
+    "step": 98,
+    "epoch": 2.76056338028169,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311576064,
+    "loss": 1.1351,
+    "grad_norm": 2.749499797821045,
+    "learning_rate": 7.500000000000002e-05
+  },
+  {
+    "step": 99,
+    "epoch": 2.788732394366197,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311572992,
+    "loss": 1.1429,
+    "grad_norm": 3.283785343170166,
+    "learning_rate": 7.17847194930753e-05
+  },
+  {
+    "step": 100,
+    "epoch": 2.816901408450704,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311571456,
+    "loss": 1.1013,
+    "grad_norm": 3.4343783855438232,
+    "learning_rate": 6.86180604201361e-05
+  },
+  {
+    "step": 101,
+    "epoch": 2.845070422535211,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311614464,
+    "loss": 0.9924,
+    "grad_norm": 3.1077940464019775,
+    "learning_rate": 6.550199129045668e-05
+  },
+  {
+    "step": 102,
+    "epoch": 2.873239436619718,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311553024,
+    "loss": 1.196,
+    "grad_norm": 3.4048402309417725,
+    "learning_rate": 6.243844916478155e-05
+  },
+  {
+    "step": 103,
+    "epoch": 2.9014084507042255,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311602176,
+    "loss": 1.0887,
+    "grad_norm": 2.745129346847534,
+    "learning_rate": 5.9429338451178355e-05
+  },
+  {
+    "step": 104,
+    "epoch": 2.9295774647887325,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311665152,
+    "loss": 1.2042,
+    "grad_norm": 4.17969274520874,
+    "learning_rate": 5.6476529721189974e-05
+  },
+  {
+    "step": 105,
+    "epoch": 2.9577464788732395,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311617536,
+    "loss": 1.0371,
+    "grad_norm": 2.640275716781616,
+    "learning_rate": 5.358185854701909e-05
+  },
+  {
+    "step": 106,
+    "epoch": 2.9859154929577465,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.311599104,
+    "loss": 1.1826,
+    "grad_norm": 3.1509358882904053,
+    "learning_rate": 5.074712436047112e-05
+  },
+  {
+    "step": 107,
+    "epoch": 3.0140845070422535,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395690496,
+    "loss": 1.2879,
+    "grad_norm": 5.848189353942871,
+    "learning_rate": 4.7974089334362057e-05
+  },
+  {
+    "step": 108,
+    "epoch": 3.0422535211267605,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395672064,
+    "loss": 0.938,
+    "grad_norm": 2.7139761447906494,
+    "learning_rate": 4.526447728708908e-05
+  },
+  {
+    "step": 109,
+    "epoch": 3.0704225352112675,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395661312,
+    "loss": 0.8768,
+    "grad_norm": 3.0623884201049805,
+    "learning_rate": 4.261997261104223e-05
+  },
+  {
+    "step": 110,
+    "epoch": 3.0985915492957745,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395715072,
+    "loss": 0.8991,
+    "grad_norm": 3.3740732669830322,
+    "learning_rate": 4.004221922552608e-05
+  },
+  {
+    "step": 111,
+    "epoch": 3.1267605633802815,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395675136,
+    "loss": 0.9248,
+    "grad_norm": 3.0837693214416504,
+    "learning_rate": 3.753281955483985e-05
+  },
+  {
+    "step": 112,
+    "epoch": 3.1549295774647885,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395693568,
+    "loss": 0.8681,
+    "grad_norm": 2.892775535583496,
+    "learning_rate": 3.509333353215331e-05
+  },
+  {
+    "step": 113,
+    "epoch": 3.183098591549296,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395756544,
+    "loss": 0.769,
+    "grad_norm": 2.8671891689300537,
+    "learning_rate": 3.2725277629795526e-05
+  },
+  {
+    "step": 114,
+    "epoch": 3.211267605633803,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395684352,
+    "loss": 0.8145,
+    "grad_norm": 3.2121176719665527,
+    "learning_rate": 3.0430123916561672e-05
+  },
+  {
+    "step": 115,
+    "epoch": 3.23943661971831,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395678208,
+    "loss": 0.9297,
+    "grad_norm": 3.5213088989257812,
+    "learning_rate": 2.8209299142621522e-05
+  },
+  {
+    "step": 116,
+    "epoch": 3.267605633802817,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395693568,
+    "loss": 0.8152,
+    "grad_norm": 4.156257629394531,
+    "learning_rate": 2.6064183852600797e-05
+  },
+  {
+    "step": 117,
+    "epoch": 3.295774647887324,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395708928,
+    "loss": 0.7545,
+    "grad_norm": 3.5243821144104004,
+    "learning_rate": 2.3996111527384288e-05
+  },
+  {
+    "step": 118,
+    "epoch": 3.323943661971831,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395699712,
+    "loss": 0.8089,
+    "grad_norm": 4.4013495445251465,
+    "learning_rate": 2.2006367755176655e-05
+  },
+  {
+    "step": 119,
+    "epoch": 3.352112676056338,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395690496,
+    "loss": 0.7094,
+    "grad_norm": 3.5775132179260254,
+    "learning_rate": 2.009618943233419e-05
+  },
+  {
+    "step": 120,
+    "epoch": 3.380281690140845,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395708928,
+    "loss": 0.8269,
+    "grad_norm": 4.176782131195068,
+    "learning_rate": 1.82667639944657e-05
+  },
+  {
+    "step": 121,
+    "epoch": 3.408450704225352,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395707392,
+    "loss": 0.6735,
+    "grad_norm": 4.356370449066162,
+    "learning_rate": 1.6519228678279718e-05
+  },
+  {
+    "step": 122,
+    "epoch": 3.436619718309859,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395664384,
+    "loss": 0.5891,
+    "grad_norm": 3.9058308601379395,
+    "learning_rate": 1.4854669814637143e-05
+  },
+  {
+    "step": 123,
+    "epoch": 3.464788732394366,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.39569664,
+    "loss": 0.674,
+    "grad_norm": 4.811568260192871,
+    "learning_rate": 1.3274122153249028e-05
+  },
+  {
+    "step": 124,
+    "epoch": 3.492957746478873,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.39565056,
+    "loss": 0.696,
+    "grad_norm": 5.005099773406982,
+    "learning_rate": 1.1778568219438839e-05
+  },
+  {
+    "step": 125,
+    "epoch": 3.52112676056338,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395695104,
+    "loss": 0.6493,
+    "grad_norm": 4.956096649169922,
+    "learning_rate": 1.036893770336938e-05
+  },
+  {
+    "step": 126,
+    "epoch": 3.5492957746478875,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395645952,
+    "loss": 0.6994,
+    "grad_norm": 5.525113582611084,
+    "learning_rate": 9.046106882113751e-06
+  },
+  {
+    "step": 127,
+    "epoch": 3.5774647887323945,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.39565824,
+    "loss": 0.6784,
+    "grad_norm": 4.813937187194824,
+    "learning_rate": 7.810898074930243e-06
+  },
+  {
+    "step": 128,
+    "epoch": 3.6056338028169015,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395682816,
+    "loss": 0.6895,
+    "grad_norm": 5.974727630615234,
+    "learning_rate": 6.664079132078881e-06
+  },
+  {
+    "step": 129,
+    "epoch": 3.6338028169014085,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395644416,
+    "loss": 0.6212,
+    "grad_norm": 5.571749210357666,
+    "learning_rate": 5.606362957498195e-06
+  },
+  {
+    "step": 130,
+    "epoch": 3.6619718309859155,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395647488,
+    "loss": 0.7407,
+    "grad_norm": 6.153623580932617,
+    "learning_rate": 4.638407065638322e-06
+  },
+  {
+    "step": 131,
+    "epoch": 3.6901408450704225,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395659776,
+    "loss": 0.6919,
+    "grad_norm": 7.186251163482666,
+    "learning_rate": 3.760813172726457e-06
+  },
+  {
+    "step": 132,
+    "epoch": 3.7183098591549295,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395624448,
+    "loss": 0.7357,
+    "grad_norm": 5.518382549285889,
+    "learning_rate": 2.9741268227184255e-06
+  },
+  {
+    "step": 133,
+    "epoch": 3.7464788732394365,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.39566592,
+    "loss": 0.645,
+    "grad_norm": 4.9238481521606445,
+    "learning_rate": 2.2788370481687965e-06
+  },
+  {
+    "step": 134,
+    "epoch": 3.7746478873239435,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.39568128,
+    "loss": 0.7276,
+    "grad_norm": 5.81230354309082,
+    "learning_rate": 1.6753760662307215e-06
+  },
+  {
+    "step": 135,
+    "epoch": 3.802816901408451,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395645952,
+    "loss": 0.7939,
+    "grad_norm": 7.195316791534424,
+    "learning_rate": 1.1641190099741904e-06
+  },
+  {
+    "step": 136,
+    "epoch": 3.830985915492958,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395653632,
+    "loss": 0.6268,
+    "grad_norm": 5.663719177246094,
+    "learning_rate": 7.453836951897885e-07
+  },
+  {
+    "step": 137,
+    "epoch": 3.859154929577465,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395675136,
+    "loss": 0.6477,
+    "grad_norm": 6.333588600158691,
+    "learning_rate": 4.194304228229806e-07
+  },
+  {
+    "step": 138,
+    "epoch": 3.887323943661972,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395685888,
+    "loss": 0.7271,
+    "grad_norm": 7.243999004364014,
+    "learning_rate": 1.8646181716164831e-07
+  },
+  {
+    "step": 139,
+    "epoch": 3.915492957746479,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395678208,
+    "loss": 0.6908,
+    "grad_norm": 6.188854217529297,
+    "learning_rate": 4.662269987756317e-08
+  },
+  {
+    "step": 140,
+    "epoch": 3.943661971830986,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395712,
+    "loss": 0.6887,
+    "grad_norm": 5.151924133300781,
+    "learning_rate": 0.0
+  },
+  {
+    "step": 140,
+    "epoch": 3.943661971830986,
+    "cpu_mem": 2.557423616,
+    "gpu_mem": 1.395712,
+    "train_runtime": 679.7329,
+    "train_samples_per_second": 13.246,
+    "train_steps_per_second": 0.206,
+    "total_flos": 0.0,
+    "train_loss": 1.2421479872294834
+  }
+]

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-boolq-r2-a2/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha": 4,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "enabled_mlp": true,
+  "enabled_qkv": [
+    "q",
+    "k",
+    "v"
+  ],
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "mixture": false,
+  "modules_to_preserve_errors": null,
+  "modules_to_quantize": null,
+  "modules_to_save": null,
+  "onnx_export": false,
+  "optimization_level": 3,
+  "orthogonal_init": false,
+  "peft_type": "MARS",
+  "quant_n_bits": 4,
+  "r": 2,
+  "revision": null,
+  "seed": 42,
+  "shared_r": 2,
+  "target_modules": [
+    "o_proj",
+    "k_proj",
+    "down_proj",
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "q_proj"
+  ],
+  "task_type": null
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-boolq-r2-a2/eval_results.json

@@ -0,0 +1,4 @@
+{
+    "task": "boolq",
+    "results": 0.6785932721712539
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-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": "mars",
+    "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": 1307064
+  },
+  "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-mars-boolq-r2-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-mars-opt3-q4-new/TinyLlama_v1.1-mars-boolq-r2-a2",
+  "seed": 42,
+  "timestamp": "2025-08-29T13:54:03.881222"
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-boolq-r2-a2/training_logs.json

@@ -0,0 +1,2659 @@
+[
+  {
+    "step": 1,
+    "epoch": 0.006779661016949152,
+    "cpu_mem": 1.692184576,
+    "gpu_mem": 1.039525888,
+    "loss": 8.874,
+    "grad_norm": 291.3755798339844,
+    "learning_rate": 9.999999999999999e-06
+  },
+  {
+    "step": 2,
+    "epoch": 0.013559322033898305,
+    "cpu_mem": 1.774481408,
+    "gpu_mem": 1.067201024,
+    "loss": 8.9593,
+    "grad_norm": 313.7054443359375,
+    "learning_rate": 1.9999999999999998e-05
+  },
+  {
+    "step": 3,
+    "epoch": 0.020338983050847456,
+    "cpu_mem": 1.775071232,
+    "gpu_mem": 1.067119616,
+    "loss": 7.8833,
+    "grad_norm": 301.32061767578125,
+    "learning_rate": 2.9999999999999997e-05
+  },
+  {
+    "step": 4,
+    "epoch": 0.02711864406779661,
+    "cpu_mem": 1.77625088,
+    "gpu_mem": 1.067119616,
+    "loss": 6.035,
+    "grad_norm": 295.67266845703125,
+    "learning_rate": 3.9999999999999996e-05
+  },
+  {
+    "step": 5,
+    "epoch": 0.03389830508474576,
+    "cpu_mem": 1.776840704,
+    "gpu_mem": 1.067055104,
+    "loss": 3.8316,
+    "grad_norm": 244.2612762451172,
+    "learning_rate": 4.9999999999999996e-05
+  },
+  {
+    "step": 6,
+    "epoch": 0.04067796610169491,
+    "cpu_mem": 1.77723392,
+    "gpu_mem": 1.067075072,
+    "loss": 2.2108,
+    "grad_norm": 140.66513061523438,
+    "learning_rate": 5.9999999999999995e-05
+  },
+  {
+    "step": 7,
+    "epoch": 0.04745762711864407,
+    "cpu_mem": 1.77821696,
+    "gpu_mem": 1.067127296,
+    "loss": 1.3536,
+    "grad_norm": 75.02676391601562,
+    "learning_rate": 7e-05
+  },
+  {
+    "step": 8,
+    "epoch": 0.05423728813559322,
+    "cpu_mem": 1.779003392,
+    "gpu_mem": 1.067213312,
+    "loss": 0.8221,
+    "grad_norm": 44.54599380493164,
+    "learning_rate": 7.999999999999999e-05
+  },
+  {
+    "step": 9,
+    "epoch": 0.061016949152542375,
+    "cpu_mem": 1.779396608,
+    "gpu_mem": 1.067121152,
+    "loss": 1.147,
+    "grad_norm": 125.10873413085938,
+    "learning_rate": 8.999999999999999e-05
+  },
+  {
+    "step": 10,
+    "epoch": 0.06779661016949153,
+    "cpu_mem": 1.78018304,
+    "gpu_mem": 1.067021312,
+    "loss": 0.7799,
+    "grad_norm": 66.12985229492188,
+    "learning_rate": 9.999999999999999e-05
+  },
+  {
+    "step": 11,
+    "epoch": 0.07457627118644068,
+    "cpu_mem": 1.780576256,
+    "gpu_mem": 1.06712576,
+    "loss": 1.2022,
+    "grad_norm": 143.1725311279297,
+    "learning_rate": 0.00010999999999999998
+  },
+  {
+    "step": 12,
+    "epoch": 0.08135593220338982,
+    "cpu_mem": 1.780969472,
+    "gpu_mem": 1.067497472,
+    "loss": 1.0554,
+    "grad_norm": 115.85326385498047,
+    "learning_rate": 0.00011999999999999999
+  },
+  {
+    "step": 13,
+    "epoch": 0.08813559322033898,
+    "cpu_mem": 1.781362688,
+    "gpu_mem": 1.067101184,
+    "loss": 0.6992,
+    "grad_norm": 21.397693634033203,
+    "learning_rate": 0.00013
+  },
+  {
+    "step": 14,
+    "epoch": 0.09491525423728814,
+    "cpu_mem": 1.781755904,
+    "gpu_mem": 1.067078144,
+    "loss": 0.6819,
+    "grad_norm": 14.809820175170898,
+    "learning_rate": 0.00014
+  },
+  {
+    "step": 15,
+    "epoch": 0.1016949152542373,
+    "cpu_mem": 1.781952512,
+    "gpu_mem": 1.067016704,
+    "loss": 0.7285,
+    "grad_norm": 27.61170196533203,
+    "learning_rate": 0.00015
+  },
+  {
+    "step": 16,
+    "epoch": 0.10847457627118644,
+    "cpu_mem": 1.782345728,
+    "gpu_mem": 1.067101184,
+    "loss": 0.7103,
+    "grad_norm": 21.049345016479492,
+    "learning_rate": 0.00015999999999999999
+  },
+  {
+    "step": 17,
+    "epoch": 0.1152542372881356,
+    "cpu_mem": 1.782738944,
+    "gpu_mem": 1.06714112,
+    "loss": 0.8616,
+    "grad_norm": 82.16301727294922,
+    "learning_rate": 0.00016999999999999999
+  },
+  {
+    "step": 18,
+    "epoch": 0.12203389830508475,
+    "cpu_mem": 1.782935552,
+    "gpu_mem": 1.067204096,
+    "loss": 0.7385,
+    "grad_norm": 39.1813850402832,
+    "learning_rate": 0.00017999999999999998
+  },
+  {
+    "step": 19,
+    "epoch": 0.1288135593220339,
+    "cpu_mem": 1.783328768,
+    "gpu_mem": 1.06704128,
+    "loss": 0.7023,
+    "grad_norm": 9.508649826049805,
+    "learning_rate": 0.00018999999999999998
+  },
+  {
+    "step": 20,
+    "epoch": 0.13559322033898305,
+    "cpu_mem": 1.7841152,
+    "gpu_mem": 1.067153408,
+    "loss": 0.681,
+    "grad_norm": 37.81815719604492,
+    "learning_rate": 0.00019999999999999998
+  },
+  {
+    "step": 21,
+    "epoch": 0.1423728813559322,
+    "cpu_mem": 1.784311808,
+    "gpu_mem": 1.067311616,
+    "loss": 0.6711,
+    "grad_norm": 31.313589096069336,
+    "learning_rate": 0.00020999999999999998
+  },
+  {
+    "step": 22,
+    "epoch": 0.14915254237288136,
+    "cpu_mem": 1.784393728,
+    "gpu_mem": 1.067204096,
+    "loss": 0.7823,
+    "grad_norm": 41.10369873046875,
+    "learning_rate": 0.00021999999999999995
+  },
+  {
+    "step": 23,
+    "epoch": 0.15593220338983052,
+    "cpu_mem": 1.784983552,
+    "gpu_mem": 1.067176448,
+    "loss": 0.7426,
+    "grad_norm": 43.635948181152344,
+    "learning_rate": 0.00023
+  },
+  {
+    "step": 24,
+    "epoch": 0.16271186440677965,
+    "cpu_mem": 1.78518016,
+    "gpu_mem": 1.06723328,
+    "loss": 0.6711,
+    "grad_norm": 38.893856048583984,
+    "learning_rate": 0.00023999999999999998
+  },
+  {
+    "step": 25,
+    "epoch": 0.1694915254237288,
+    "cpu_mem": 1.78536448,
+    "gpu_mem": 1.06701824,
+    "loss": 0.7459,
+    "grad_norm": 34.04146194458008,
+    "learning_rate": 0.00025
+  },
+  {
+    "step": 26,
+    "epoch": 0.17627118644067796,
+    "cpu_mem": 1.786150912,
+    "gpu_mem": 1.067073536,
+    "loss": 0.7823,
+    "grad_norm": 32.66416549682617,
+    "learning_rate": 0.00026
+  },
+  {
+    "step": 27,
+    "epoch": 0.18305084745762712,
+    "cpu_mem": 1.786343424,
+    "gpu_mem": 1.067365376,
+    "loss": 0.7429,
+    "grad_norm": 37.693138122558594,
+    "learning_rate": 0.00027
+  },
+  {
+    "step": 28,
+    "epoch": 0.18983050847457628,
+    "cpu_mem": 1.786376192,
+    "gpu_mem": 1.067044352,
+    "loss": 0.6878,
+    "grad_norm": 9.707056045532227,
+    "learning_rate": 0.00028
+  },
+  {
+    "step": 29,
+    "epoch": 0.19661016949152543,
+    "cpu_mem": 1.786769408,
+    "gpu_mem": 1.067108864,
+    "loss": 0.6581,
+    "grad_norm": 12.347441673278809,
+    "learning_rate": 0.00029
+  },
+  {
+    "step": 30,
+    "epoch": 0.2033898305084746,
+    "cpu_mem": 1.786966016,
+    "gpu_mem": 1.0671872,
+    "loss": 0.6763,
+    "grad_norm": 14.598573684692383,
+    "learning_rate": 0.0003
+  },
+  {
+    "step": 31,
+    "epoch": 0.21016949152542372,
+    "cpu_mem": 1.787162624,
+    "gpu_mem": 1.066990592,
+    "loss": 0.8017,
+    "grad_norm": 43.43600082397461,
+    "learning_rate": 0.0002999893794250036
+  },
+  {
+    "step": 32,
+    "epoch": 0.21694915254237288,
+    "cpu_mem": 1.787359232,
+    "gpu_mem": 1.067104256,
+    "loss": 0.7734,
+    "grad_norm": 28.66684913635254,
+    "learning_rate": 0.00029995751920396937
+  },
+  {
+    "step": 33,
+    "epoch": 0.22372881355932203,
+    "cpu_mem": 1.78755584,
+    "gpu_mem": 1.067342336,
+    "loss": 0.6755,
+    "grad_norm": 3.923459529876709,
+    "learning_rate": 0.00029990442384854874
+  },
+  {
+    "step": 34,
+    "epoch": 0.2305084745762712,
+    "cpu_mem": 1.787752448,
+    "gpu_mem": 1.067044352,
+    "loss": 0.5818,
+    "grad_norm": 6.199655055999756,
+    "learning_rate": 0.0002998301008774512
+  },
+  {
+    "step": 35,
+    "epoch": 0.23728813559322035,
+    "cpu_mem": 1.788342272,
+    "gpu_mem": 1.067254784,
+    "loss": 0.6999,
+    "grad_norm": 15.666038513183594,
+    "learning_rate": 0.0002997345608153792
+  },
+  {
+    "step": 36,
+    "epoch": 0.2440677966101695,
+    "cpu_mem": 1.78853888,
+    "gpu_mem": 1.067205632,
+    "loss": 0.8204,
+    "grad_norm": 33.818546295166016,
+    "learning_rate": 0.000299617817191538
+  },
+  {
+    "step": 37,
+    "epoch": 0.25084745762711863,
+    "cpu_mem": 1.788735488,
+    "gpu_mem": 1.067016704,
+    "loss": 0.8295,
+    "grad_norm": 37.759639739990234,
+    "learning_rate": 0.0002994798865377198
+  },
+  {
+    "step": 38,
+    "epoch": 0.2576271186440678,
+    "cpu_mem": 1.788932096,
+    "gpu_mem": 1.067264,
+    "loss": 0.7103,
+    "grad_norm": 16.049278259277344,
+    "learning_rate": 0.0002993207883859627
+  },
+  {
+    "step": 39,
+    "epoch": 0.26440677966101694,
+    "cpu_mem": 1.791488,
+    "gpu_mem": 1.067643392,
+    "loss": 0.6793,
+    "grad_norm": 5.071969509124756,
+    "learning_rate": 0.0002991405452657846
+  },
+  {
+    "step": 40,
+    "epoch": 0.2711864406779661,
+    "cpu_mem": 1.791488,
+    "gpu_mem": 1.067213312,
+    "loss": 0.7018,
+    "grad_norm": 16.534059524536133,
+    "learning_rate": 0.00029893918270099324
+  },
+  {
+    "step": 41,
+    "epoch": 0.27796610169491526,
+    "cpu_mem": 1.791684608,
+    "gpu_mem": 1.06744064,
+    "loss": 0.6888,
+    "grad_norm": 5.772024154663086,
+    "learning_rate": 0.00029871672920607153
+  },
+  {
+    "step": 42,
+    "epoch": 0.2847457627118644,
+    "cpu_mem": 1.791881216,
+    "gpu_mem": 1.067337728,
+    "loss": 0.7286,
+    "grad_norm": 24.6931095123291,
+    "learning_rate": 0.0002984732162821399
+  },
+  {
+    "step": 43,
+    "epoch": 0.29152542372881357,
+    "cpu_mem": 1.792077824,
+    "gpu_mem": 1.067159552,
+    "loss": 0.6181,
+    "grad_norm": 10.050416946411133,
+    "learning_rate": 0.0002982086784124952
+  },
+  {
+    "step": 44,
+    "epoch": 0.2983050847457627,
+    "cpu_mem": 1.792274432,
+    "gpu_mem": 1.0673024,
+    "loss": 0.6505,
+    "grad_norm": 15.0296630859375,
+    "learning_rate": 0.00029792315305772796
+  },
+  {
+    "step": 45,
+    "epoch": 0.3050847457627119,
+    "cpu_mem": 1.79247104,
+    "gpu_mem": 1.067082752,
+    "loss": 0.6943,
+    "grad_norm": 10.401119232177734,
+    "learning_rate": 0.0002976166806504174
+  },
+  {
+    "step": 46,
+    "epoch": 0.31186440677966104,
+    "cpu_mem": 1.792667648,
+    "gpu_mem": 1.06732544,
+    "loss": 0.7991,
+    "grad_norm": 26.349973678588867,
+    "learning_rate": 0.00029728930458940595
+  },
+  {
+    "step": 47,
+    "epoch": 0.31864406779661014,
+    "cpu_mem": 1.792864256,
+    "gpu_mem": 1.06704896,
+    "loss": 0.6991,
+    "grad_norm": 18.533720016479492,
+    "learning_rate": 0.00029694107123365385
+  },
+  {
+    "step": 48,
+    "epoch": 0.3254237288135593,
+    "cpu_mem": 1.792864256,
+    "gpu_mem": 1.06712576,
+    "loss": 0.5962,
+    "grad_norm": 3.064704656600952,
+    "learning_rate": 0.00029657202989567393
+  },
+  {
+    "step": 49,
+    "epoch": 0.33220338983050846,
+    "cpu_mem": 1.793060864,
+    "gpu_mem": 1.067142656,
+    "loss": 1.081,
+    "grad_norm": 37.34111022949219,
+    "learning_rate": 0.00029618223283454893
+  },
+  {
+    "step": 50,
+    "epoch": 0.3389830508474576,
+    "cpu_mem": 1.793257472,
+    "gpu_mem": 1.067081216,
+    "loss": 0.74,
+    "grad_norm": 21.937891006469727,
+    "learning_rate": 0.00029577173524853123
+  },
+  {
+    "step": 51,
+    "epoch": 0.34576271186440677,
+    "cpu_mem": 1.793257472,
+    "gpu_mem": 1.067085824,
+    "loss": 0.635,
+    "grad_norm": 22.710704803466797,
+    "learning_rate": 0.0002953405952672261
+  },
+  {
+    "step": 52,
+    "epoch": 0.3525423728813559,
+    "cpu_mem": 1.79345408,
+    "gpu_mem": 1.067165696,
+    "loss": 0.7188,
+    "grad_norm": 9.722867965698242,
+    "learning_rate": 0.0002948888739433602
+  },
+  {
+    "step": 53,
+    "epoch": 0.3593220338983051,
+    "cpu_mem": 1.79345408,
+    "gpu_mem": 1.067188736,
+    "loss": 0.5927,
+    "grad_norm": 7.451268672943115,
+    "learning_rate": 0.0002944166352441363
+  },
+  {
+    "step": 54,
+    "epoch": 0.36610169491525424,
+    "cpu_mem": 1.79345408,
+    "gpu_mem": 1.067116544,
+    "loss": 1.0226,
+    "grad_norm": 33.91159439086914,
+    "learning_rate": 0.0002939239460421746
+  },
+  {
+    "step": 55,
+    "epoch": 0.3728813559322034,
+    "cpu_mem": 1.793650688,
+    "gpu_mem": 1.06738688,
+    "loss": 0.753,
+    "grad_norm": 21.46257972717285,
+    "learning_rate": 0.00029341087610604337
+  },
+  {
+    "step": 56,
+    "epoch": 0.37966101694915255,
+    "cpu_mem": 1.793847296,
+    "gpu_mem": 1.067173376,
+    "loss": 0.712,
+    "grad_norm": 16.427001953125,
+    "learning_rate": 0.00029287749809037904
+  },
+  {
+    "step": 57,
+    "epoch": 0.3864406779661017,
+    "cpu_mem": 1.793847296,
+    "gpu_mem": 1.067167232,
+    "loss": 0.9295,
+    "grad_norm": 48.56226348876953,
+    "learning_rate": 0.0002923238875255979
+  },
+  {
+    "step": 58,
+    "epoch": 0.39322033898305087,
+    "cpu_mem": 1.794043904,
+    "gpu_mem": 1.067062784,
+    "loss": 1.1245,
+    "grad_norm": 55.51771545410156,
+    "learning_rate": 0.00029175012280720024
+  },
+  {
+    "step": 59,
+    "epoch": 0.4,
+    "cpu_mem": 1.794043904,
+    "gpu_mem": 1.06707968,
+    "loss": 0.7755,
+    "grad_norm": 22.440706253051758,
+    "learning_rate": 0.000291156285184669
+  },
+  {
+    "step": 60,
+    "epoch": 0.4067796610169492,
+    "cpu_mem": 1.794043904,
+    "gpu_mem": 1.067173376,
+    "loss": 0.5978,
+    "grad_norm": 5.783385276794434,
+    "learning_rate": 0.00029054245874996426
+  },
+  {
+    "step": 61,
+    "epoch": 0.4135593220338983,
+    "cpu_mem": 1.794240512,
+    "gpu_mem": 1.067184128,
+    "loss": 0.7382,
+    "grad_norm": 24.120975494384766,
+    "learning_rate": 0.0002899087304256151
+  },
+  {
+    "step": 62,
+    "epoch": 0.42033898305084744,
+    "cpu_mem": 1.79443712,
+    "gpu_mem": 1.06717184,
+    "loss": 1.046,
+    "grad_norm": 57.39268112182617,
+    "learning_rate": 0.0002892551899524109
+  },
+  {
+    "step": 63,
+    "epoch": 0.4271186440677966,
+    "cpu_mem": 1.79443712,
+    "gpu_mem": 1.06716416,
+    "loss": 0.6748,
+    "grad_norm": 59.949466705322266,
+    "learning_rate": 0.000288581929876693
+  },
+  {
+    "step": 64,
+    "epoch": 0.43389830508474575,
+    "cpu_mem": 1.79443712,
+    "gpu_mem": 1.067093504,
+    "loss": 0.7373,
+    "grad_norm": 67.06160736083984,
+    "learning_rate": 0.0002878890455372498
+  },
+  {
+    "step": 65,
+    "epoch": 0.4406779661016949,
+    "cpu_mem": 1.79443712,
+    "gpu_mem": 1.067138048,
+    "loss": 0.7317,
+    "grad_norm": 35.02265548706055,
+    "learning_rate": 0.0002871766350518159
+  },
+  {
+    "step": 66,
+    "epoch": 0.44745762711864406,
+    "cpu_mem": 1.794633728,
+    "gpu_mem": 1.067331584,
+    "loss": 0.6341,
+    "grad_norm": 16.198490142822266,
+    "learning_rate": 0.00028644479930317775
+  },
+  {
+    "step": 67,
+    "epoch": 0.4542372881355932,
+    "cpu_mem": 1.794633728,
+    "gpu_mem": 1.06704128,
+    "loss": 0.7393,
+    "grad_norm": 65.17927551269531,
+    "learning_rate": 0.00028569364192488803
+  },
+  {
+    "step": 68,
+    "epoch": 0.4610169491525424,
+    "cpu_mem": 1.794830336,
+    "gpu_mem": 1.067009024,
+    "loss": 0.6964,
+    "grad_norm": 20.7398738861084,
+    "learning_rate": 0.00028492326928659045
+  },
+  {
+    "step": 69,
+    "epoch": 0.46779661016949153,
+    "cpu_mem": 1.795026944,
+    "gpu_mem": 1.067075072,
+    "loss": 0.6231,
+    "grad_norm": 23.524229049682617,
+    "learning_rate": 0.00028413379047895665
+  },
+  {
+    "step": 70,
+    "epoch": 0.4745762711864407,
+    "cpu_mem": 1.795026944,
+    "gpu_mem": 1.067068928,
+    "loss": 0.6883,
+    "grad_norm": 19.696847915649414,
+    "learning_rate": 0.0002833253172982385
+  },
+  {
+    "step": 71,
+    "epoch": 0.48135593220338985,
+    "cpu_mem": 1.795026944,
+    "gpu_mem": 1.067297792,
+    "loss": 0.7668,
+    "grad_norm": 36.587337493896484,
+    "learning_rate": 0.0002824979642304366
+  },
+  {
+    "step": 72,
+    "epoch": 0.488135593220339,
+    "cpu_mem": 1.795223552,
+    "gpu_mem": 1.067290112,
+    "loss": 0.6592,
+    "grad_norm": 19.84144401550293,
+    "learning_rate": 0.0002816518484350883
+  },
+  {
+    "step": 73,
+    "epoch": 0.49491525423728816,
+    "cpu_mem": 1.795223552,
+    "gpu_mem": 1.06725632,
+    "loss": 0.8185,
+    "grad_norm": 26.411832809448242,
+    "learning_rate": 0.0002807870897286772
+  },
+  {
+    "step": 74,
+    "epoch": 0.5016949152542373,
+    "cpu_mem": 1.795223552,
+    "gpu_mem": 1.067116544,
+    "loss": 0.5878,
+    "grad_norm": 11.005779266357422,
+    "learning_rate": 0.0002799038105676658
+  },
+  {
+    "step": 75,
+    "epoch": 0.5084745762711864,
+    "cpu_mem": 1.795223552,
+    "gpu_mem": 1.06704128,
+    "loss": 0.6086,
+    "grad_norm": 16.758214950561523,
+    "learning_rate": 0.000279002136031155
+  },
+  {
+    "step": 76,
+    "epoch": 0.5152542372881356,
+    "cpu_mem": 1.795223552,
+    "gpu_mem": 1.066981376,
+    "loss": 0.6045,
+    "grad_norm": 12.238960266113281,
+    "learning_rate": 0.00027808219380317216
+  },
+  {
+    "step": 77,
+    "epoch": 0.5220338983050847,
+    "cpu_mem": 1.795223552,
+    "gpu_mem": 1.067055104,
+    "loss": 0.5746,
+    "grad_norm": 6.873617172241211,
+    "learning_rate": 0.0002771441141545895
+  },
+  {
+    "step": 78,
+    "epoch": 0.5288135593220339,
+    "cpu_mem": 1.79542016,
+    "gpu_mem": 1.067107328,
+    "loss": 0.9858,
+    "grad_norm": 54.217674255371094,
+    "learning_rate": 0.0002761880299246772
+  },
+  {
+    "step": 79,
+    "epoch": 0.535593220338983,
+    "cpu_mem": 1.795612672,
+    "gpu_mem": 1.067239424,
+    "loss": 0.675,
+    "grad_norm": 15.84262466430664,
+    "learning_rate": 0.000275214076502292
+  },
+  {
+    "step": 80,
+    "epoch": 0.5423728813559322,
+    "cpu_mem": 1.795612672,
+    "gpu_mem": 1.067130368,
+    "loss": 0.6726,
+    "grad_norm": 18.819782257080078,
+    "learning_rate": 0.0002742223918067056
+  },
+  {
+    "step": 81,
+    "epoch": 0.5491525423728814,
+    "cpu_mem": 1.795612672,
+    "gpu_mem": 1.06701056,
+    "loss": 0.6121,
+    "grad_norm": 12.81640911102295,
+    "learning_rate": 0.00027321311626807374
+  },
+  {
+    "step": 82,
+    "epoch": 0.5559322033898305,
+    "cpu_mem": 1.795612672,
+    "gpu_mem": 1.06707968,
+    "loss": 0.8172,
+    "grad_norm": 42.44817352294922,
+    "learning_rate": 0.0002721863928075503
+  },
+  {
+    "step": 83,
+    "epoch": 0.5627118644067797,
+    "cpu_mem": 1.795612672,
+    "gpu_mem": 1.06717952,
+    "loss": 0.7441,
+    "grad_norm": 27.580766677856445,
+    "learning_rate": 0.000271142366817049
+  },
+  {
+    "step": 84,
+    "epoch": 0.5694915254237288,
+    "cpu_mem": 1.79580928,
+    "gpu_mem": 1.067142656,
+    "loss": 0.656,
+    "grad_norm": 16.37428092956543,
+    "learning_rate": 0.00027008118613865406
+  },
+  {
+    "step": 85,
+    "epoch": 0.576271186440678,
+    "cpu_mem": 1.79580928,
+    "gpu_mem": 1.067174912,
+    "loss": 0.6365,
+    "grad_norm": 7.864536285400391,
+    "learning_rate": 0.00026900300104368524
+  },
+  {
+    "step": 86,
+    "epoch": 0.5830508474576271,
+    "cpu_mem": 1.79580928,
+    "gpu_mem": 1.06712576,
+    "loss": 0.6734,
+    "grad_norm": 7.170403480529785,
+    "learning_rate": 0.00026790796421141813
+  },
+  {
+    "step": 87,
+    "epoch": 0.5898305084745763,
+    "cpu_mem": 1.79580928,
+    "gpu_mem": 1.06713344,
+    "loss": 0.613,
+    "grad_norm": 7.2136759757995605,
+    "learning_rate": 0.00026679623070746325
+  },
+  {
+    "step": 88,
+    "epoch": 0.5966101694915255,
+    "cpu_mem": 1.79580928,
+    "gpu_mem": 1.067277824,
+    "loss": 0.7224,
+    "grad_norm": 40.812339782714844,
+    "learning_rate": 0.0002656679579618081
+  },
+  {
+    "step": 89,
+    "epoch": 0.6033898305084746,
+    "cpu_mem": 1.79580928,
+    "gpu_mem": 1.067059712,
+    "loss": 0.6609,
+    "grad_norm": 12.928619384765625,
+    "learning_rate": 0.0002645233057465235
+  },
+  {
+    "step": 90,
+    "epoch": 0.6101694915254238,
+    "cpu_mem": 1.796005888,
+    "gpu_mem": 1.067113472,
+    "loss": 0.6629,
+    "grad_norm": 14.732848167419434,
+    "learning_rate": 0.00026336243615313873
+  },
+  {
+    "step": 91,
+    "epoch": 0.6169491525423729,
+    "cpu_mem": 1.796005888,
+    "gpu_mem": 1.067081216,
+    "loss": 0.6226,
+    "grad_norm": 8.494933128356934,
+    "learning_rate": 0.00026218551356968814
+  },
+  {
+    "step": 92,
+    "epoch": 0.6237288135593221,
+    "cpu_mem": 1.796005888,
+    "gpu_mem": 1.067162624,
+    "loss": 0.693,
+    "grad_norm": 9.185640335083008,
+    "learning_rate": 0.00026099270465743254
+  },
+  {
+    "step": 93,
+    "epoch": 0.6305084745762712,
+    "cpu_mem": 1.796005888,
+    "gpu_mem": 1.066966016,
+    "loss": 0.8204,
+    "grad_norm": 30.638669967651367,
+    "learning_rate": 0.0002597841783272588
+  },
+  {
+    "step": 94,
+    "epoch": 0.6372881355932203,
+    "cpu_mem": 1.796005888,
+    "gpu_mem": 1.06707968,
+    "loss": 0.9576,
+    "grad_norm": 56.41335678100586,
+    "learning_rate": 0.0002585601057157605
+  },
+  {
+    "step": 95,
+    "epoch": 0.6440677966101694,
+    "cpu_mem": 1.796005888,
+    "gpu_mem": 1.067099648,
+    "loss": 0.6816,
+    "grad_norm": 17.54722785949707,
+    "learning_rate": 0.00025732066016100394
+  },
+  {
+    "step": 96,
+    "epoch": 0.6508474576271186,
+    "cpu_mem": 1.796005888,
+    "gpu_mem": 1.067138048,
+    "loss": 0.8321,
+    "grad_norm": 61.89482498168945,
+    "learning_rate": 0.00025606601717798207
+  },
+  {
+    "step": 97,
+    "epoch": 0.6576271186440678,
+    "cpu_mem": 1.796005888,
+    "gpu_mem": 1.067122688,
+    "loss": 0.9806,
+    "grad_norm": 64.5597915649414,
+    "learning_rate": 0.0002547963544337602
+  },
+  {
+    "step": 98,
+    "epoch": 0.6644067796610169,
+    "cpu_mem": 1.796005888,
+    "gpu_mem": 1.067035136,
+    "loss": 0.6897,
+    "grad_norm": 16.576112747192383,
+    "learning_rate": 0.0002535118517223168
+  },
+  {
+    "step": 99,
+    "epoch": 0.6711864406779661,
+    "cpu_mem": 1.796005888,
+    "gpu_mem": 1.066984448,
+    "loss": 0.6967,
+    "grad_norm": 21.121322631835938,
+    "learning_rate": 0.00025221269093908365
+  },
+  {
+    "step": 100,
+    "epoch": 0.6779661016949152,
+    "cpu_mem": 1.796005888,
+    "gpu_mem": 1.067101184,
+    "loss": 0.6994,
+    "grad_norm": 21.069868087768555,
+    "learning_rate": 0.0002508990560551879
+  },
+  {
+    "step": 101,
+    "epoch": 0.6847457627118644,
+    "cpu_mem": 1.796005888,
+    "gpu_mem": 1.06713344,
+    "loss": 1.2423,
+    "grad_norm": 117.06101989746094,
+    "learning_rate": 0.0002495711330914001
+  },
+  {
+    "step": 102,
+    "epoch": 0.6915254237288135,
+    "cpu_mem": 1.796005888,
+    "gpu_mem": 1.067167232,
+    "loss": 0.8385,
+    "grad_norm": 38.71371841430664,
+    "learning_rate": 0.00024822911009179276
+  },
+  {
+    "step": 103,
+    "epoch": 0.6983050847457627,
+    "cpu_mem": 1.796005888,
+    "gpu_mem": 1.06721792,
+    "loss": 0.6489,
+    "grad_norm": 5.8236260414123535,
+    "learning_rate": 0.0002468731770971113
+  },
+  {
+    "step": 104,
+    "epoch": 0.7050847457627119,
+    "cpu_mem": 1.796005888,
+    "gpu_mem": 1.067122688,
+    "loss": 0.7339,
+    "grad_norm": 66.14034271240234,
+    "learning_rate": 0.0002455035261178632
+  },
+  {
+    "step": 105,
+    "epoch": 0.711864406779661,
+    "cpu_mem": 1.796005888,
+    "gpu_mem": 1.067224064,
+    "loss": 0.6994,
+    "grad_norm": 16.291786193847656,
+    "learning_rate": 0.0002441203511071278
+  },
+  {
+    "step": 106,
+    "epoch": 0.7186440677966102,
+    "cpu_mem": 1.796005888,
+    "gpu_mem": 1.067174912,
+    "loss": 0.6202,
+    "grad_norm": 4.222020149230957,
+    "learning_rate": 0.00024272384793309077
+  },
+  {
+    "step": 107,
+    "epoch": 0.7254237288135593,
+    "cpu_mem": 1.796202496,
+    "gpu_mem": 1.067062784,
+    "loss": 0.6355,
+    "grad_norm": 7.902516841888428,
+    "learning_rate": 0.00024131421435130807
+  },
+  {
+    "step": 108,
+    "epoch": 0.7322033898305085,
+    "cpu_mem": 1.796202496,
+    "gpu_mem": 1.067247104,
+    "loss": 0.6792,
+    "grad_norm": 31.659059524536133,
+    "learning_rate": 0.00023989164997670202
+  },
+  {
+    "step": 109,
+    "epoch": 0.7389830508474576,
+    "cpu_mem": 1.79613696,
+    "gpu_mem": 1.067101184,
+    "loss": 0.6705,
+    "grad_norm": 4.247161388397217,
+    "learning_rate": 0.0002384563562552943
+  },
+  {
+    "step": 110,
+    "epoch": 0.7457627118644068,
+    "cpu_mem": 1.79613696,
+    "gpu_mem": 1.067104256,
+    "loss": 0.6483,
+    "grad_norm": 4.0431647300720215,
+    "learning_rate": 0.0002370085364356797
+  },
+  {
+    "step": 111,
+    "epoch": 0.752542372881356,
+    "cpu_mem": 1.79613696,
+    "gpu_mem": 1.067073536,
+    "loss": 0.6156,
+    "grad_norm": 4.198848247528076,
+    "learning_rate": 0.0002355483955402446
+  },
+  {
+    "step": 112,
+    "epoch": 0.7593220338983051,
+    "cpu_mem": 1.79613696,
+    "gpu_mem": 1.067119616,
+    "loss": 0.6729,
+    "grad_norm": 21.031217575073242,
+    "learning_rate": 0.00023407614033613407
+  },
+  {
+    "step": 113,
+    "epoch": 0.7661016949152543,
+    "cpu_mem": 1.796333568,
+    "gpu_mem": 1.0671104,
+    "loss": 0.6631,
+    "grad_norm": 6.252691268920898,
+    "learning_rate": 0.0002325919793059723
+  },
+  {
+    "step": 114,
+    "epoch": 0.7728813559322034,
+    "cpu_mem": 1.796333568,
+    "gpu_mem": 1.067091968,
+    "loss": 0.6223,
+    "grad_norm": 11.559258460998535,
+    "learning_rate": 0.00023109612261833963
+  },
+  {
+    "step": 115,
+    "epoch": 0.7796610169491526,
+    "cpu_mem": 1.79625984,
+    "gpu_mem": 1.067167232,
+    "loss": 0.774,
+    "grad_norm": 35.43349838256836,
+    "learning_rate": 0.0002295887820980112
+  },
+  {
+    "step": 116,
+    "epoch": 0.7864406779661017,
+    "cpu_mem": 1.79625984,
+    "gpu_mem": 1.06708736,
+    "loss": 0.6297,
+    "grad_norm": 10.36580753326416,
+    "learning_rate": 0.0002280701711959608
+  },
+  {
+    "step": 117,
+    "epoch": 0.7932203389830509,
+    "cpu_mem": 1.79625984,
+    "gpu_mem": 1.066978304,
+    "loss": 0.6969,
+    "grad_norm": 17.803829193115234,
+    "learning_rate": 0.00022654050495913495
+  },
+  {
+    "step": 118,
+    "epoch": 0.8,
+    "cpu_mem": 1.796456448,
+    "gpu_mem": 1.067216384,
+    "loss": 0.6875,
+    "grad_norm": 14.91399097442627,
+    "learning_rate": 0.000225
+  },
+  {
+    "step": 119,
+    "epoch": 0.8067796610169492,
+    "cpu_mem": 1.796456448,
+    "gpu_mem": 1.06738688,
+    "loss": 0.6005,
+    "grad_norm": 4.876289367675781,
+    "learning_rate": 0.00022344887446586865
+  },
+  {
+    "step": 120,
+    "epoch": 0.8135593220338984,
+    "cpu_mem": 1.796653056,
+    "gpu_mem": 1.067119616,
+    "loss": 0.6439,
+    "grad_norm": 11.509865760803223,
+    "learning_rate": 0.00022188734800800852
+  },
+  {
+    "step": 121,
+    "epoch": 0.8203389830508474,
+    "cpu_mem": 1.796653056,
+    "gpu_mem": 1.067147264,
+    "loss": 0.6838,
+    "grad_norm": 96.23526763916016,
+    "learning_rate": 0.00022031564175053754
+  },
+  {
+    "step": 122,
+    "epoch": 0.8271186440677966,
+    "cpu_mem": 1.796653056,
+    "gpu_mem": 1.067197952,
+    "loss": 0.5861,
+    "grad_norm": 16.30876350402832,
+    "learning_rate": 0.00021873397825911153
+  },
+  {
+    "step": 123,
+    "epoch": 0.8338983050847457,
+    "cpu_mem": 1.796849664,
+    "gpu_mem": 1.067007488,
+    "loss": 0.6853,
+    "grad_norm": 15.60392951965332,
+    "learning_rate": 0.00021714258150940685
+  },
+  {
+    "step": 124,
+    "epoch": 0.8406779661016949,
+    "cpu_mem": 1.797046272,
+    "gpu_mem": 1.067449856,
+    "loss": 0.6229,
+    "grad_norm": 10.207924842834473,
+    "learning_rate": 0.0002155416768554039
+  },
+  {
+    "step": 125,
+    "epoch": 0.847457627118644,
+    "cpu_mem": 1.797042176,
+    "gpu_mem": 1.067176448,
+    "loss": 0.5491,
+    "grad_norm": 7.521026611328125,
+    "learning_rate": 0.00021393149099747523
+  },
+  {
+    "step": 126,
+    "epoch": 0.8542372881355932,
+    "cpu_mem": 1.797042176,
+    "gpu_mem": 1.067059712,
+    "loss": 0.5683,
+    "grad_norm": 7.79779052734375,
+    "learning_rate": 0.00021231225195028297
+  },
+  {
+    "step": 127,
+    "epoch": 0.8610169491525423,
+    "cpu_mem": 1.797042176,
+    "gpu_mem": 1.067499008,
+    "loss": 0.6645,
+    "grad_norm": 18.44670295715332,
+    "learning_rate": 0.00021068418901049025
+  },
+  {
+    "step": 128,
+    "epoch": 0.8677966101694915,
+    "cpu_mem": 1.797238784,
+    "gpu_mem": 1.067274752,
+    "loss": 0.6625,
+    "grad_norm": 28.01342010498047,
+    "learning_rate": 0.0002090475327242912
+  },
+  {
+    "step": 129,
+    "epoch": 0.8745762711864407,
+    "cpu_mem": 1.797013504,
+    "gpu_mem": 1.067314688,
+    "loss": 0.7477,
+    "grad_norm": 11.30091381072998,
+    "learning_rate": 0.00020740251485476345
+  },
+  {
+    "step": 130,
+    "epoch": 0.8813559322033898,
+    "cpu_mem": 1.797013504,
+    "gpu_mem": 1.067096576,
+    "loss": 0.6573,
+    "grad_norm": 9.1353120803833,
+    "learning_rate": 0.0002057493683490491
+  },
+  {
+    "step": 131,
+    "epoch": 0.888135593220339,
+    "cpu_mem": 1.797013504,
+    "gpu_mem": 1.0672256,
+    "loss": 0.6643,
+    "grad_norm": 9.245180130004883,
+    "learning_rate": 0.00020408832730536746
+  },
+  {
+    "step": 132,
+    "epoch": 0.8949152542372881,
+    "cpu_mem": 1.797013504,
+    "gpu_mem": 1.067307008,
+    "loss": 0.6864,
+    "grad_norm": 20.19770050048828,
+    "learning_rate": 0.00020241962693986476
+  },
+  {
+    "step": 133,
+    "epoch": 0.9016949152542373,
+    "cpu_mem": 1.797210112,
+    "gpu_mem": 1.067090432,
+    "loss": 0.661,
+    "grad_norm": 230.81329345703125,
+    "learning_rate": 0.0002007435035533061
+  },
+  {
+    "step": 134,
+    "epoch": 0.9084745762711864,
+    "cpu_mem": 1.797210112,
+    "gpu_mem": 1.067224064,
+    "loss": 0.578,
+    "grad_norm": 14.924580574035645,
+    "learning_rate": 0.00019906019449761325
+  },
+  {
+    "step": 135,
+    "epoch": 0.9152542372881356,
+    "cpu_mem": 1.797107712,
+    "gpu_mem": 1.067247104,
+    "loss": 0.6823,
+    "grad_norm": 19.97910499572754,
+    "learning_rate": 0.00019736993814225374
+  },
+  {
+    "step": 136,
+    "epoch": 0.9220338983050848,
+    "cpu_mem": 1.797107712,
+    "gpu_mem": 1.067084288,
+    "loss": 0.6087,
+    "grad_norm": 9.597871780395508,
+    "learning_rate": 0.00019567297384048604
+  },
+  {
+    "step": 137,
+    "epoch": 0.9288135593220339,
+    "cpu_mem": 1.797107712,
+    "gpu_mem": 1.06696448,
+    "loss": 0.6557,
+    "grad_norm": 5.5722551345825195,
+    "learning_rate": 0.0001939695418954653
+  },
+  {
+    "step": 138,
+    "epoch": 0.9355932203389831,
+    "cpu_mem": 1.796898816,
+    "gpu_mem": 1.067145728,
+    "loss": 0.655,
+    "grad_norm": 8.457000732421875,
+    "learning_rate": 0.00019225988352621445
+  },
+  {
+    "step": 139,
+    "epoch": 0.9423728813559322,
+    "cpu_mem": 1.79689472,
+    "gpu_mem": 1.067044352,
+    "loss": 0.655,
+    "grad_norm": 6.6021409034729,
+    "learning_rate": 0.00019054424083346592
+  },
+  {
+    "step": 140,
+    "epoch": 0.9491525423728814,
+    "cpu_mem": 1.797050368,
+    "gpu_mem": 1.067096576,
+    "loss": 0.636,
+    "grad_norm": 4.045039176940918,
+    "learning_rate": 0.0001888228567653781
+  },
+  {
+    "step": 141,
+    "epoch": 0.9559322033898305,
+    "cpu_mem": 1.796988928,
+    "gpu_mem": 1.067128832,
+    "loss": 0.6422,
+    "grad_norm": 6.502439975738525,
+    "learning_rate": 0.0001870959750831323
+  },
+  {
+    "step": 142,
+    "epoch": 0.9627118644067797,
+    "cpu_mem": 1.797185536,
+    "gpu_mem": 1.067268608,
+    "loss": 0.5553,
+    "grad_norm": 7.805881023406982,
+    "learning_rate": 0.0001853638403264141
+  },
+  {
+    "step": 143,
+    "epoch": 0.9694915254237289,
+    "cpu_mem": 1.797382144,
+    "gpu_mem": 1.067251712,
+    "loss": 0.6842,
+    "grad_norm": 15.737360000610352,
+    "learning_rate": 0.00018362669777878453
+  },
+  {
+    "step": 144,
+    "epoch": 0.976271186440678,
+    "cpu_mem": 1.797382144,
+    "gpu_mem": 1.067443712,
+    "loss": 0.6948,
+    "grad_norm": 17.525169372558594,
+    "learning_rate": 0.00018188479343294648
+  },
+  {
+    "step": 145,
+    "epoch": 0.9830508474576272,
+    "cpu_mem": 1.797578752,
+    "gpu_mem": 1.067154944,
+    "loss": 0.5722,
+    "grad_norm": 7.157924175262451,
+    "learning_rate": 0.0001801383739559098
+  },
+  {
+    "step": 146,
+    "epoch": 0.9898305084745763,
+    "cpu_mem": 1.797578752,
+    "gpu_mem": 1.067190272,
+    "loss": 0.5911,
+    "grad_norm": 7.200604438781738,
+    "learning_rate": 0.0001783876866540615
+  },
+  {
+    "step": 147,
+    "epoch": 0.9966101694915255,
+    "cpu_mem": 1.797373952,
+    "gpu_mem": 1.067088896,
+    "loss": 0.6567,
+    "grad_norm": 19.34654998779297,
+    "learning_rate": 0.00017663297943814552
+  },
+  {
+    "step": 148,
+    "epoch": 1.0033898305084745,
+    "cpu_mem": 1.79757056,
+    "gpu_mem": 1.072502272,
+    "loss": 0.9631,
+    "grad_norm": 10.374675750732422,
+    "learning_rate": 0.0001748745007881561
+  },
+  {
+    "step": 149,
+    "epoch": 1.0101694915254238,
+    "cpu_mem": 1.79746816,
+    "gpu_mem": 1.07243776,
+    "loss": 0.6399,
+    "grad_norm": 9.10370922088623,
+    "learning_rate": 0.00017311249971815185
+  },
+  {
+    "step": 150,
+    "epoch": 1.0169491525423728,
+    "cpu_mem": 1.797419008,
+    "gpu_mem": 1.072274944,
+    "loss": 0.6068,
+    "grad_norm": 8.218132972717285,
+    "learning_rate": 0.00017134722574099276
+  },
+  {
+    "step": 151,
+    "epoch": 1.023728813559322,
+    "cpu_mem": 1.797615616,
+    "gpu_mem": 1.072347136,
+    "loss": 0.7122,
+    "grad_norm": 16.563411712646484,
+    "learning_rate": 0.00016957892883300775
+  },
+  {
+    "step": 152,
+    "epoch": 1.0305084745762711,
+    "cpu_mem": 1.797615616,
+    "gpu_mem": 1.072382464,
+    "loss": 0.5872,
+    "grad_norm": 7.684173107147217,
+    "learning_rate": 0.00016780785939859576
+  },
+  {
+    "step": 153,
+    "epoch": 1.0372881355932204,
+    "cpu_mem": 1.797615616,
+    "gpu_mem": 1.07240704,
+    "loss": 0.7744,
+    "grad_norm": 34.04887008666992,
+    "learning_rate": 0.00016603426823476693
+  },
+  {
+    "step": 154,
+    "epoch": 1.0440677966101695,
+    "cpu_mem": 1.797615616,
+    "gpu_mem": 1.07236864,
+    "loss": 0.6853,
+    "grad_norm": 25.140567779541016,
+    "learning_rate": 0.00016425840649562736
+  },
+  {
+    "step": 155,
+    "epoch": 1.0508474576271187,
+    "cpu_mem": 1.797615616,
+    "gpu_mem": 1.072589824,
+    "loss": 0.6239,
+    "grad_norm": 11.781804084777832,
+    "learning_rate": 0.00016248052565681436
+  },
+  {
+    "step": 156,
+    "epoch": 1.0576271186440678,
+    "cpu_mem": 1.797615616,
+    "gpu_mem": 1.072497664,
+    "loss": 0.6295,
+    "grad_norm": 8.846559524536133,
+    "learning_rate": 0.00016070087747988482
+  },
+  {
+    "step": 157,
+    "epoch": 1.064406779661017,
+    "cpu_mem": 1.797615616,
+    "gpu_mem": 1.072403968,
+    "loss": 0.7732,
+    "grad_norm": 28.42833137512207,
+    "learning_rate": 0.00015891971397666464
+  },
+  {
+    "step": 158,
+    "epoch": 1.071186440677966,
+    "cpu_mem": 1.797615616,
+    "gpu_mem": 1.07233024,
+    "loss": 0.7523,
+    "grad_norm": 22.629343032836914,
+    "learning_rate": 0.00015713728737356137
+  },
+  {
+    "step": 159,
+    "epoch": 1.0779661016949154,
+    "cpu_mem": 1.797615616,
+    "gpu_mem": 1.072678912,
+    "loss": 0.5702,
+    "grad_norm": 6.326220989227295,
+    "learning_rate": 0.00015535385007584706
+  },
+  {
+    "step": 160,
+    "epoch": 1.0847457627118644,
+    "cpu_mem": 1.797582848,
+    "gpu_mem": 1.072273408,
+    "loss": 0.6681,
+    "grad_norm": 9.057672500610352,
+    "learning_rate": 0.0001535696546319161
+  },
+  {
+    "step": 161,
+    "epoch": 1.0915254237288137,
+    "cpu_mem": 1.797582848,
+    "gpu_mem": 1.072219648,
+    "loss": 0.5916,
+    "grad_norm": 15.630982398986816,
+    "learning_rate": 0.00015178495369752213
+  },
+  {
+    "step": 162,
+    "epoch": 1.0983050847457627,
+    "cpu_mem": 1.797545984,
+    "gpu_mem": 1.072995328,
+    "loss": 0.6475,
+    "grad_norm": 9.621928215026855,
+    "learning_rate": 0.00015
+  },
+  {
+    "step": 163,
+    "epoch": 1.1050847457627118,
+    "cpu_mem": 1.797742592,
+    "gpu_mem": 1.072471552,
+    "loss": 0.6628,
+    "grad_norm": 10.567188262939453,
+    "learning_rate": 0.00014821504630247785
+  },
+  {
+    "step": 164,
+    "epoch": 1.111864406779661,
+    "cpu_mem": 1.797742592,
+    "gpu_mem": 1.072384,
+    "loss": 0.6579,
+    "grad_norm": 7.776843547821045,
+    "learning_rate": 0.00014643034536808387
+  },
+  {
+    "step": 165,
+    "epoch": 1.11864406779661,
+    "cpu_mem": 1.797742592,
+    "gpu_mem": 1.072333312,
+    "loss": 0.6304,
+    "grad_norm": 6.893831253051758,
+    "learning_rate": 0.00014464614992415294
+  },
+  {
+    "step": 166,
+    "epoch": 1.1254237288135593,
+    "cpu_mem": 1.797742592,
+    "gpu_mem": 1.072428544,
+    "loss": 0.653,
+    "grad_norm": 17.217058181762695,
+    "learning_rate": 0.00014286271262643866
+  },
+  {
+    "step": 167,
+    "epoch": 1.1322033898305084,
+    "cpu_mem": 1.797742592,
+    "gpu_mem": 1.0723456,
+    "loss": 0.6726,
+    "grad_norm": 10.91408634185791,
+    "learning_rate": 0.00014108028602333536
+  },
+  {
+    "step": 168,
+    "epoch": 1.1389830508474577,
+    "cpu_mem": 1.797742592,
+    "gpu_mem": 1.072364032,
+    "loss": 0.6239,
+    "grad_norm": 6.429533958435059,
+    "learning_rate": 0.00013929912252011516
+  },
+  {
+    "step": 169,
+    "epoch": 1.1457627118644067,
+    "cpu_mem": 1.797742592,
+    "gpu_mem": 1.072451584,
+    "loss": 0.588,
+    "grad_norm": 6.233065605163574,
+    "learning_rate": 0.00013751947434318564
+  },
+  {
+    "step": 170,
+    "epoch": 1.152542372881356,
+    "cpu_mem": 1.797742592,
+    "gpu_mem": 1.072336384,
+    "loss": 0.612,
+    "grad_norm": 11.369017601013184,
+    "learning_rate": 0.00013574159350437261
+  },
+  {
+    "step": 171,
+    "epoch": 1.159322033898305,
+    "cpu_mem": 1.797742592,
+    "gpu_mem": 1.07239936,
+    "loss": 0.6753,
+    "grad_norm": 9.935527801513672,
+    "learning_rate": 0.0001339657317652331
+  },
+  {
+    "step": 172,
+    "epoch": 1.1661016949152543,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.0723072,
+    "loss": 0.6529,
+    "grad_norm": 6.226833820343018,
+    "learning_rate": 0.00013219214060140424
+  },
+  {
+    "step": 173,
+    "epoch": 1.1728813559322033,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.07260672,
+    "loss": 0.6305,
+    "grad_norm": 13.911097526550293,
+    "learning_rate": 0.00013042107116699228
+  },
+  {
+    "step": 174,
+    "epoch": 1.1796610169491526,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.07233024,
+    "loss": 0.6629,
+    "grad_norm": 10.934565544128418,
+    "learning_rate": 0.00012865277425900724
+  },
+  {
+    "step": 175,
+    "epoch": 1.1864406779661016,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072296448,
+    "loss": 0.6504,
+    "grad_norm": 15.064614295959473,
+    "learning_rate": 0.00012688750028184818
+  },
+  {
+    "step": 176,
+    "epoch": 1.193220338983051,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072434688,
+    "loss": 0.591,
+    "grad_norm": 13.485958099365234,
+    "learning_rate": 0.0001251254992118439
+  },
+  {
+    "step": 177,
+    "epoch": 1.2,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072532992,
+    "loss": 0.5742,
+    "grad_norm": 6.3795294761657715,
+    "learning_rate": 0.00012336702056185453
+  },
+  {
+    "step": 178,
+    "epoch": 1.2067796610169492,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072279552,
+    "loss": 0.6229,
+    "grad_norm": 10.318319320678711,
+    "learning_rate": 0.00012161231334593851
+  },
+  {
+    "step": 179,
+    "epoch": 1.2135593220338983,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072379392,
+    "loss": 0.5779,
+    "grad_norm": 6.194882869720459,
+    "learning_rate": 0.00011986162604409015
+  },
+  {
+    "step": 180,
+    "epoch": 1.2203389830508475,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072351744,
+    "loss": 0.6831,
+    "grad_norm": 13.943729400634766,
+    "learning_rate": 0.00011811520656705348
+  },
+  {
+    "step": 181,
+    "epoch": 1.2271186440677966,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072288768,
+    "loss": 0.607,
+    "grad_norm": 17.86134910583496,
+    "learning_rate": 0.00011637330222121543
+  },
+  {
+    "step": 182,
+    "epoch": 1.2338983050847459,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.07250688,
+    "loss": 0.6728,
+    "grad_norm": 13.313254356384277,
+    "learning_rate": 0.00011463615967358588
+  },
+  {
+    "step": 183,
+    "epoch": 1.240677966101695,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072403968,
+    "loss": 0.7095,
+    "grad_norm": 16.932268142700195,
+    "learning_rate": 0.00011290402491686766
+  },
+  {
+    "step": 184,
+    "epoch": 1.2474576271186442,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072351744,
+    "loss": 0.6871,
+    "grad_norm": 22.442790985107422,
+    "learning_rate": 0.00011117714323462186
+  },
+  {
+    "step": 185,
+    "epoch": 1.2542372881355932,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.07233024,
+    "loss": 0.635,
+    "grad_norm": 17.37511444091797,
+    "learning_rate": 0.00010945575916653407
+  },
+  {
+    "step": 186,
+    "epoch": 1.2610169491525425,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072339456,
+    "loss": 0.5487,
+    "grad_norm": 6.469228267669678,
+    "learning_rate": 0.00010774011647378553
+  },
+  {
+    "step": 187,
+    "epoch": 1.2677966101694915,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072271872,
+    "loss": 0.6375,
+    "grad_norm": 8.593610763549805,
+    "learning_rate": 0.00010603045810453468
+  },
+  {
+    "step": 188,
+    "epoch": 1.2745762711864406,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072434688,
+    "loss": 0.658,
+    "grad_norm": 12.747215270996094,
+    "learning_rate": 0.00010432702615951396
+  },
+  {
+    "step": 189,
+    "epoch": 1.2813559322033898,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072304128,
+    "loss": 0.63,
+    "grad_norm": 8.740394592285156,
+    "learning_rate": 0.00010263006185774627
+  },
+  {
+    "step": 190,
+    "epoch": 1.288135593220339,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072423936,
+    "loss": 0.5315,
+    "grad_norm": 5.967763900756836,
+    "learning_rate": 0.00010093980550238675
+  },
+  {
+    "step": 191,
+    "epoch": 1.2949152542372881,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072242688,
+    "loss": 0.5519,
+    "grad_norm": 7.465148448944092,
+    "learning_rate": 9.925649644669391e-05
+  },
+  {
+    "step": 192,
+    "epoch": 1.3016949152542372,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072374784,
+    "loss": 0.5694,
+    "grad_norm": 17.18702507019043,
+    "learning_rate": 9.758037306013526e-05
+  },
+  {
+    "step": 193,
+    "epoch": 1.3084745762711865,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072348672,
+    "loss": 0.606,
+    "grad_norm": 7.86152982711792,
+    "learning_rate": 9.591167269463255e-05
+  },
+  {
+    "step": 194,
+    "epoch": 1.3152542372881357,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.07231488,
+    "loss": 0.6232,
+    "grad_norm": 6.897183418273926,
+    "learning_rate": 9.425063165095088e-05
+  },
+  {
+    "step": 195,
+    "epoch": 1.3220338983050848,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072419328,
+    "loss": 0.5539,
+    "grad_norm": 8.286925315856934,
+    "learning_rate": 9.259748514523653e-05
+  },
+  {
+    "step": 196,
+    "epoch": 1.3288135593220338,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.07241472,
+    "loss": 0.5208,
+    "grad_norm": 7.31152868270874,
+    "learning_rate": 9.095246727570879e-05
+  },
+  {
+    "step": 197,
+    "epoch": 1.335593220338983,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072273408,
+    "loss": 0.5487,
+    "grad_norm": 8.306645393371582,
+    "learning_rate": 8.931581098950973e-05
+  },
+  {
+    "step": 198,
+    "epoch": 1.3423728813559321,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072465408,
+    "loss": 0.5854,
+    "grad_norm": 8.069097518920898,
+    "learning_rate": 8.768774804971705e-05
+  },
+  {
+    "step": 199,
+    "epoch": 1.3491525423728814,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072316416,
+    "loss": 0.6259,
+    "grad_norm": 16.867874145507812,
+    "learning_rate": 8.606850900252478e-05
+  },
+  {
+    "step": 200,
+    "epoch": 1.3559322033898304,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072419328,
+    "loss": 0.55,
+    "grad_norm": 8.23694896697998,
+    "learning_rate": 8.445832314459608e-05
+  },
+  {
+    "step": 201,
+    "epoch": 1.3627118644067797,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.07262208,
+    "loss": 0.5431,
+    "grad_norm": 12.741939544677734,
+    "learning_rate": 8.285741849059311e-05
+  },
+  {
+    "step": 202,
+    "epoch": 1.3694915254237288,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072423936,
+    "loss": 0.6053,
+    "grad_norm": 9.624649047851562,
+    "learning_rate": 8.126602174088843e-05
+  },
+  {
+    "step": 203,
+    "epoch": 1.376271186440678,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072310272,
+    "loss": 0.4861,
+    "grad_norm": 11.901386260986328,
+    "learning_rate": 7.968435824946242e-05
+  },
+  {
+    "step": 204,
+    "epoch": 1.383050847457627,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072324096,
+    "loss": 0.6024,
+    "grad_norm": 7.298447132110596,
+    "learning_rate": 7.811265199199152e-05
+  },
+  {
+    "step": 205,
+    "epoch": 1.3898305084745763,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.07236864,
+    "loss": 0.6356,
+    "grad_norm": 11.019448280334473,
+    "learning_rate": 7.655112553413135e-05
+  },
+  {
+    "step": 206,
+    "epoch": 1.3966101694915254,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072310272,
+    "loss": 0.5197,
+    "grad_norm": 6.436569690704346,
+    "learning_rate": 7.500000000000002e-05
+  },
+  {
+    "step": 207,
+    "epoch": 1.4033898305084747,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072543744,
+    "loss": 0.5731,
+    "grad_norm": 9.095263481140137,
+    "learning_rate": 7.345949504086507e-05
+  },
+  {
+    "step": 208,
+    "epoch": 1.4101694915254237,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072574464,
+    "loss": 0.5065,
+    "grad_norm": 11.13254165649414,
+    "learning_rate": 7.192982880403917e-05
+  },
+  {
+    "step": 209,
+    "epoch": 1.4169491525423727,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072500736,
+    "loss": 0.6019,
+    "grad_norm": 9.443418502807617,
+    "learning_rate": 7.041121790198881e-05
+  },
+  {
+    "step": 210,
+    "epoch": 1.423728813559322,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072388608,
+    "loss": 0.5776,
+    "grad_norm": 6.862228870391846,
+    "learning_rate": 6.890387738166041e-05
+  },
+  {
+    "step": 211,
+    "epoch": 1.4305084745762713,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.07233792,
+    "loss": 0.547,
+    "grad_norm": 8.054023742675781,
+    "learning_rate": 6.740802069402771e-05
+  },
+  {
+    "step": 212,
+    "epoch": 1.4372881355932203,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.0723072,
+    "loss": 0.5822,
+    "grad_norm": 8.296786308288574,
+    "learning_rate": 6.592385966386588e-05
+  },
+  {
+    "step": 213,
+    "epoch": 1.4440677966101694,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.07233024,
+    "loss": 0.6606,
+    "grad_norm": 8.03130054473877,
+    "learning_rate": 6.445160445975536e-05
+  },
+  {
+    "step": 214,
+    "epoch": 1.4508474576271186,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072413184,
+    "loss": 0.607,
+    "grad_norm": 7.151822090148926,
+    "learning_rate": 6.299146356432029e-05
+  },
+  {
+    "step": 215,
+    "epoch": 1.457627118644068,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072340992,
+    "loss": 0.5881,
+    "grad_norm": 10.393117904663086,
+    "learning_rate": 6.154364374470568e-05
+  },
+  {
+    "step": 216,
+    "epoch": 1.464406779661017,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.07250688,
+    "loss": 0.5719,
+    "grad_norm": 8.044678688049316,
+    "learning_rate": 6.010835002329795e-05
+  },
+  {
+    "step": 217,
+    "epoch": 1.471186440677966,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072348672,
+    "loss": 0.579,
+    "grad_norm": 8.284494400024414,
+    "learning_rate": 5.8685785648691894e-05
+  },
+  {
+    "step": 218,
+    "epoch": 1.4779661016949153,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072325632,
+    "loss": 0.5693,
+    "grad_norm": 9.51559829711914,
+    "learning_rate": 5.72761520669092e-05
+  },
+  {
+    "step": 219,
+    "epoch": 1.4847457627118645,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072451584,
+    "loss": 0.5657,
+    "grad_norm": 9.279644966125488,
+    "learning_rate": 5.587964889287218e-05
+  },
+  {
+    "step": 220,
+    "epoch": 1.4915254237288136,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072485376,
+    "loss": 0.5521,
+    "grad_norm": 7.055351257324219,
+    "learning_rate": 5.449647388213678e-05
+  },
+  {
+    "step": 221,
+    "epoch": 1.4983050847457626,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.07235328,
+    "loss": 0.6077,
+    "grad_norm": 8.214317321777344,
+    "learning_rate": 5.312682290288869e-05
+  },
+  {
+    "step": 222,
+    "epoch": 1.505084745762712,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072489984,
+    "loss": 0.5135,
+    "grad_norm": 6.798290252685547,
+    "learning_rate": 5.1770889908207245e-05
+  },
+  {
+    "step": 223,
+    "epoch": 1.5118644067796612,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072403968,
+    "loss": 0.5714,
+    "grad_norm": 8.6298189163208,
+    "learning_rate": 5.0428866908599864e-05
+  },
+  {
+    "step": 224,
+    "epoch": 1.5186440677966102,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.07236864,
+    "loss": 0.6006,
+    "grad_norm": 7.770857810974121,
+    "learning_rate": 4.9100943944812114e-05
+  },
+  {
+    "step": 225,
+    "epoch": 1.5254237288135593,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072333312,
+    "loss": 0.6025,
+    "grad_norm": 9.466155052185059,
+    "learning_rate": 4.778730906091632e-05
+  },
+  {
+    "step": 226,
+    "epoch": 1.5322033898305085,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072482304,
+    "loss": 0.4778,
+    "grad_norm": 6.76349401473999,
+    "learning_rate": 4.648814827768322e-05
+  },
+  {
+    "step": 227,
+    "epoch": 1.5389830508474578,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072371712,
+    "loss": 0.5687,
+    "grad_norm": 8.24431324005127,
+    "learning_rate": 4.5203645566239816e-05
+  },
+  {
+    "step": 228,
+    "epoch": 1.5457627118644068,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072316416,
+    "loss": 0.6535,
+    "grad_norm": 10.470850944519043,
+    "learning_rate": 4.3933982822017876e-05
+  },
+  {
+    "step": 229,
+    "epoch": 1.5525423728813559,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072258048,
+    "loss": 0.5761,
+    "grad_norm": 8.238707542419434,
+    "learning_rate": 4.267933983899601e-05
+  },
+  {
+    "step": 230,
+    "epoch": 1.559322033898305,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.07231488,
+    "loss": 0.603,
+    "grad_norm": 8.463348388671875,
+    "learning_rate": 4.143989428423947e-05
+  },
+  {
+    "step": 231,
+    "epoch": 1.5661016949152542,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072592896,
+    "loss": 0.5968,
+    "grad_norm": 6.370932579040527,
+    "learning_rate": 4.0215821672741213e-05
+  },
+  {
+    "step": 232,
+    "epoch": 1.5728813559322035,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072316416,
+    "loss": 0.6799,
+    "grad_norm": 10.808438301086426,
+    "learning_rate": 3.900729534256745e-05
+  },
+  {
+    "step": 233,
+    "epoch": 1.5796610169491525,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.07262976,
+    "loss": 0.5498,
+    "grad_norm": 9.736083030700684,
+    "learning_rate": 3.781448643031187e-05
+  },
+  {
+    "step": 234,
+    "epoch": 1.5864406779661016,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072505344,
+    "loss": 0.544,
+    "grad_norm": 10.119654655456543,
+    "learning_rate": 3.663756384686127e-05
+  },
+  {
+    "step": 235,
+    "epoch": 1.5932203389830508,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.07226112,
+    "loss": 0.5947,
+    "grad_norm": 18.05272674560547,
+    "learning_rate": 3.547669425347647e-05
+  },
+  {
+    "step": 236,
+    "epoch": 1.6,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072321024,
+    "loss": 0.5395,
+    "grad_norm": 8.5020170211792,
+    "learning_rate": 3.433204203819185e-05
+  },
+  {
+    "step": 237,
+    "epoch": 1.6067796610169491,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072382464,
+    "loss": 0.5431,
+    "grad_norm": 11.324089050292969,
+    "learning_rate": 3.3203769292536764e-05
+  },
+  {
+    "step": 238,
+    "epoch": 1.6135593220338982,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072384,
+    "loss": 0.5979,
+    "grad_norm": 9.914655685424805,
+    "learning_rate": 3.209203578858191e-05
+  },
+  {
+    "step": 239,
+    "epoch": 1.6203389830508474,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.07263744,
+    "loss": 0.6107,
+    "grad_norm": 10.563992500305176,
+    "learning_rate": 3.099699895631474e-05
+  },
+  {
+    "step": 240,
+    "epoch": 1.6271186440677967,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072287232,
+    "loss": 0.6307,
+    "grad_norm": 9.804160118103027,
+    "learning_rate": 2.9918813861345952e-05
+  },
+  {
+    "step": 241,
+    "epoch": 1.6338983050847458,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.07258368,
+    "loss": 0.5666,
+    "grad_norm": 8.388216972351074,
+    "learning_rate": 2.885763318295102e-05
+  },
+  {
+    "step": 242,
+    "epoch": 1.6406779661016948,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.07244544,
+    "loss": 0.5815,
+    "grad_norm": 9.455634117126465,
+    "learning_rate": 2.781360719244964e-05
+  },
+  {
+    "step": 243,
+    "epoch": 1.647457627118644,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072297984,
+    "loss": 0.7665,
+    "grad_norm": 16.60271644592285,
+    "learning_rate": 2.6786883731926306e-05
+  },
+  {
+    "step": 244,
+    "epoch": 1.6542372881355933,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.07243776,
+    "loss": 0.5376,
+    "grad_norm": 7.040289878845215,
+    "learning_rate": 2.5777608193294396e-05
+  },
+  {
+    "step": 245,
+    "epoch": 1.6610169491525424,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072316416,
+    "loss": 0.588,
+    "grad_norm": 9.260379791259766,
+    "learning_rate": 2.4785923497707956e-05
+  },
+  {
+    "step": 246,
+    "epoch": 1.6677966101694914,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072410112,
+    "loss": 0.5779,
+    "grad_norm": 9.03203010559082,
+    "learning_rate": 2.38119700753228e-05
+  },
+  {
+    "step": 247,
+    "epoch": 1.6745762711864407,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072428544,
+    "loss": 0.5881,
+    "grad_norm": 8.537538528442383,
+    "learning_rate": 2.285588584541047e-05
+  },
+  {
+    "step": 248,
+    "epoch": 1.68135593220339,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072380928,
+    "loss": 0.5312,
+    "grad_norm": 5.940595626831055,
+    "learning_rate": 2.1917806196827792e-05
+  },
+  {
+    "step": 249,
+    "epoch": 1.688135593220339,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072287232,
+    "loss": 0.5151,
+    "grad_norm": 9.854302406311035,
+    "learning_rate": 2.0997863968844914e-05
+  },
+  {
+    "step": 250,
+    "epoch": 1.694915254237288,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072379392,
+    "loss": 0.5526,
+    "grad_norm": 12.617406845092773,
+    "learning_rate": 2.009618943233419e-05
+  },
+  {
+    "step": 251,
+    "epoch": 1.7016949152542373,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.07229184,
+    "loss": 0.5842,
+    "grad_norm": 7.597400665283203,
+    "learning_rate": 1.921291027132278e-05
+  },
+  {
+    "step": 252,
+    "epoch": 1.7084745762711866,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072334848,
+    "loss": 0.5797,
+    "grad_norm": 7.297758102416992,
+    "learning_rate": 1.834815156491165e-05
+  },
+  {
+    "step": 253,
+    "epoch": 1.7152542372881356,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072528384,
+    "loss": 0.5631,
+    "grad_norm": 8.775803565979004,
+    "learning_rate": 1.750203576956341e-05
+  },
+  {
+    "step": 254,
+    "epoch": 1.7220338983050847,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072324096,
+    "loss": 0.6357,
+    "grad_norm": 11.823816299438477,
+    "learning_rate": 1.6674682701761493e-05
+  },
+  {
+    "step": 255,
+    "epoch": 1.7288135593220337,
+    "cpu_mem": 1.7979392,
+    "gpu_mem": 1.072480768,
+    "loss": 0.5455,
+    "grad_norm": 9.827958106994629,
+    "learning_rate": 1.5866209521043304e-05
+  },
+  {
+    "step": 256,
+    "epoch": 1.735593220338983,
+    "cpu_mem": 1.797906432,
+    "gpu_mem": 1.0723072,
+    "loss": 0.5356,
+    "grad_norm": 8.764373779296875,
+    "learning_rate": 1.5076730713409523e-05
+  },
+  {
+    "step": 257,
+    "epoch": 1.7423728813559323,
+    "cpu_mem": 1.79810304,
+    "gpu_mem": 1.072720384,
+    "loss": 0.6121,
+    "grad_norm": 15.294634819030762,
+    "learning_rate": 1.4306358075111923e-05
+  },
+  {
+    "step": 258,
+    "epoch": 1.7491525423728813,
+    "cpu_mem": 1.79810304,
+    "gpu_mem": 1.072379392,
+    "loss": 0.5652,
+    "grad_norm": 12.042337417602539,
+    "learning_rate": 1.3555200696822232e-05
+  },
+  {
+    "step": 259,
+    "epoch": 1.7559322033898304,
+    "cpu_mem": 1.79810304,
+    "gpu_mem": 1.072296448,
+    "loss": 0.6021,
+    "grad_norm": 8.471673011779785,
+    "learning_rate": 1.2823364948184095e-05
+  },
+  {
+    "step": 260,
+    "epoch": 1.7627118644067796,
+    "cpu_mem": 1.79810304,
+    "gpu_mem": 1.072413184,
+    "loss": 0.4417,
+    "grad_norm": 7.8211774826049805,
+    "learning_rate": 1.2110954462750166e-05
+  },
+  {
+    "step": 261,
+    "epoch": 1.769491525423729,
+    "cpu_mem": 1.79810304,
+    "gpu_mem": 1.07236864,
+    "loss": 0.4433,
+    "grad_norm": 7.459161758422852,
+    "learning_rate": 1.1418070123306989e-05
+  },
+  {
+    "step": 262,
+    "epoch": 1.776271186440678,
+    "cpu_mem": 1.79810304,
+    "gpu_mem": 1.072325632,
+    "loss": 0.4973,
+    "grad_norm": 8.854447364807129,
+    "learning_rate": 1.0744810047589115e-05
+  },
+  {
+    "step": 263,
+    "epoch": 1.783050847457627,
+    "cpu_mem": 1.79810304,
+    "gpu_mem": 1.072362496,
+    "loss": 0.5719,
+    "grad_norm": 7.8189520835876465,
+    "learning_rate": 1.0091269574384874e-05
+  },
+  {
+    "step": 264,
+    "epoch": 1.7898305084745763,
+    "cpu_mem": 1.79810304,
+    "gpu_mem": 1.072450048,
+    "loss": 0.5331,
+    "grad_norm": 7.528275489807129,
+    "learning_rate": 9.45754125003576e-06
+  },
+  {
+    "step": 265,
+    "epoch": 1.7966101694915255,
+    "cpu_mem": 1.79810304,
+    "gpu_mem": 1.07236864,
+    "loss": 0.5941,
+    "grad_norm": 10.641229629516602,
+    "learning_rate": 8.843714815330987e-06
+  },
+  {
+    "step": 266,
+    "epoch": 1.8033898305084746,
+    "cpu_mem": 1.79810304,
+    "gpu_mem": 1.07258368,
+    "loss": 0.591,
+    "grad_norm": 8.315288543701172,
+    "learning_rate": 8.249877192799731e-06
+  },
+  {
+    "step": 267,
+    "epoch": 1.8101694915254236,
+    "cpu_mem": 1.79810304,
+    "gpu_mem": 1.07237632,
+    "loss": 0.4775,
+    "grad_norm": 11.404972076416016,
+    "learning_rate": 7.676112474402068e-06
+  },
+  {
+    "step": 268,
+    "epoch": 1.8169491525423729,
+    "cpu_mem": 1.79810304,
+    "gpu_mem": 1.072380928,
+    "loss": 0.5234,
+    "grad_norm": 10.700228691101074,
+    "learning_rate": 7.122501909620926e-06
+  },
+  {
+    "step": 269,
+    "epoch": 1.8237288135593221,
+    "cpu_mem": 1.79810304,
+    "gpu_mem": 1.07239168,
+    "loss": 0.5836,
+    "grad_norm": 13.900726318359375,
+    "learning_rate": 6.5891238939566275e-06
+  },
+  {
+    "step": 270,
+    "epoch": 1.8305084745762712,
+    "cpu_mem": 1.79810304,
+    "gpu_mem": 1.07243008,
+    "loss": 0.6022,
+    "grad_norm": 10.767265319824219,
+    "learning_rate": 6.076053957825411e-06
+  },
+  {
+    "step": 271,
+    "epoch": 1.8372881355932202,
+    "cpu_mem": 1.799479296,
+    "gpu_mem": 1.072482304,
+    "loss": 0.5524,
+    "grad_norm": 10.918460845947266,
+    "learning_rate": 5.583364755863701e-06
+  },
+  {
+    "step": 272,
+    "epoch": 1.8440677966101695,
+    "cpu_mem": 1.801641984,
+    "gpu_mem": 1.072340992,
+    "loss": 0.6307,
+    "grad_norm": 11.49914836883545,
+    "learning_rate": 5.11112605663977e-06
+  },
+  {
+    "step": 273,
+    "epoch": 1.8508474576271188,
+    "cpu_mem": 1.801641984,
+    "gpu_mem": 1.072221184,
+    "loss": 0.5899,
+    "grad_norm": 10.842668533325195,
+    "learning_rate": 4.659404732773908e-06
+  },
+  {
+    "step": 274,
+    "epoch": 1.8576271186440678,
+    "cpu_mem": 1.801641984,
+    "gpu_mem": 1.072448512,
+    "loss": 0.6161,
+    "grad_norm": 10.53109359741211,
+    "learning_rate": 4.228264751468752e-06
+  },
+  {
+    "step": 275,
+    "epoch": 1.8644067796610169,
+    "cpu_mem": 1.801641984,
+    "gpu_mem": 1.072692736,
+    "loss": 0.5369,
+    "grad_norm": 11.559465408325195,
+    "learning_rate": 3.817767165451041e-06
+  },
+  {
+    "step": 276,
+    "epoch": 1.8711864406779661,
+    "cpu_mem": 1.801641984,
+    "gpu_mem": 1.07235328,
+    "loss": 0.5555,
+    "grad_norm": 7.54006290435791,
+    "learning_rate": 3.4279701043260886e-06
+  },
+  {
+    "step": 277,
+    "epoch": 1.8779661016949154,
+    "cpu_mem": 1.801641984,
+    "gpu_mem": 1.07229952,
+    "loss": 0.6145,
+    "grad_norm": 8.808974266052246,
+    "learning_rate": 3.0589287663461472e-06
+  },
+  {
+    "step": 278,
+    "epoch": 1.8847457627118644,
+    "cpu_mem": 1.801641984,
+    "gpu_mem": 1.072462336,
+    "loss": 0.5865,
+    "grad_norm": 8.585607528686523,
+    "learning_rate": 2.710695410593994e-06
+  },
+  {
+    "step": 279,
+    "epoch": 1.8915254237288135,
+    "cpu_mem": 1.802428416,
+    "gpu_mem": 1.072402432,
+    "loss": 0.5963,
+    "grad_norm": 8.654258728027344,
+    "learning_rate": 2.3833193495825853e-06
+  },
+  {
+    "step": 280,
+    "epoch": 1.8983050847457628,
+    "cpu_mem": 1.802428416,
+    "gpu_mem": 1.072382464,
+    "loss": 0.5161,
+    "grad_norm": 6.762901306152344,
+    "learning_rate": 2.076846942272026e-06
+  },
+  {
+    "step": 281,
+    "epoch": 1.905084745762712,
+    "cpu_mem": 1.802428416,
+    "gpu_mem": 1.072317952,
+    "loss": 0.6325,
+    "grad_norm": 9.365017890930176,
+    "learning_rate": 1.791321587504768e-06
+  },
+  {
+    "step": 282,
+    "epoch": 1.911864406779661,
+    "cpu_mem": 1.804529664,
+    "gpu_mem": 1.072746496,
+    "loss": 0.5636,
+    "grad_norm": 11.185689926147461,
+    "learning_rate": 1.5267837178600972e-06
+  },
+  {
+    "step": 283,
+    "epoch": 1.9186440677966101,
+    "cpu_mem": 1.80492288,
+    "gpu_mem": 1.07245312,
+    "loss": 0.6543,
+    "grad_norm": 10.285652160644531,
+    "learning_rate": 1.2832707939284427e-06
+  },
+  {
+    "step": 284,
+    "epoch": 1.9254237288135592,
+    "cpu_mem": 1.80492288,
+    "gpu_mem": 1.072308736,
+    "loss": 0.5674,
+    "grad_norm": 8.522608757019043,
+    "learning_rate": 1.0608172990067553e-06
+  },
+  {
+    "step": 285,
+    "epoch": 1.9322033898305084,
+    "cpu_mem": 1.80492288,
+    "gpu_mem": 1.072362496,
+    "loss": 0.6468,
+    "grad_norm": 11.664175987243652,
+    "learning_rate": 8.594547342153979e-07
+  },
+  {
+    "step": 286,
+    "epoch": 1.9389830508474577,
+    "cpu_mem": 1.8049024,
+    "gpu_mem": 1.072780288,
+    "loss": 0.5655,
+    "grad_norm": 10.680014610290527,
+    "learning_rate": 6.792116140373116e-07
+  },
+  {
+    "step": 287,
+    "epoch": 1.9457627118644067,
+    "cpu_mem": 1.805099008,
+    "gpu_mem": 1.072549888,
+    "loss": 0.5575,
+    "grad_norm": 10.169900894165039,
+    "learning_rate": 5.201134622801473e-07
+  },
+  {
+    "step": 288,
+    "epoch": 1.9525423728813558,
+    "cpu_mem": 1.804886016,
+    "gpu_mem": 1.072334848,
+    "loss": 0.5905,
+    "grad_norm": 9.122550010681152,
+    "learning_rate": 3.821828084619727e-07
+  },
+  {
+    "step": 289,
+    "epoch": 1.959322033898305,
+    "cpu_mem": 1.805082624,
+    "gpu_mem": 1.072419328,
+    "loss": 0.5947,
+    "grad_norm": 10.595670700073242,
+    "learning_rate": 2.654391846207915e-07
+  },
+  {
+    "step": 290,
+    "epoch": 1.9661016949152543,
+    "cpu_mem": 1.808031744,
+    "gpu_mem": 1.072344064,
+    "loss": 0.6375,
+    "grad_norm": 9.97242259979248,
+    "learning_rate": 1.6989912254880556e-07
+  },
+  {
+    "step": 291,
+    "epoch": 1.9728813559322034,
+    "cpu_mem": 1.808818176,
+    "gpu_mem": 1.072379392,
+    "loss": 0.6539,
+    "grad_norm": 9.128472328186035,
+    "learning_rate": 9.557615145123765e-08
+  },
+  {
+    "step": 292,
+    "epoch": 1.9796610169491524,
+    "cpu_mem": 1.813532672,
+    "gpu_mem": 1.072462336,
+    "loss": 0.5639,
+    "grad_norm": 8.636130332946777,
+    "learning_rate": 4.248079603064724e-08
+  },
+  {
+    "step": 293,
+    "epoch": 1.9864406779661017,
+    "cpu_mem": 1.815302144,
+    "gpu_mem": 1.072379392,
+    "loss": 0.6293,
+    "grad_norm": 9.215025901794434,
+    "learning_rate": 1.0620574996372811e-08
+  },
+  {
+    "step": 294,
+    "epoch": 1.993220338983051,
+    "cpu_mem": 1.821396992,
+    "gpu_mem": 1.072405504,
+    "loss": 0.6499,
+    "grad_norm": 9.986368179321289,
+    "learning_rate": 0.0
+  },
+  {
+    "step": 294,
+    "epoch": 1.993220338983051,
+    "cpu_mem": 1.821396992,
+    "gpu_mem": 1.072405504,
+    "train_runtime": 4463.1665,
+    "train_samples_per_second": 4.224,
+    "train_steps_per_second": 0.066,
+    "total_flos": 0.0,
+    "train_loss": 0.7742782291911897
+  }
+]

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-hellaswag-r8-a2/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha": 16,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "enabled_mlp": true,
+  "enabled_qkv": [
+    "q",
+    "k",
+    "v"
+  ],
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "mixture": false,
+  "modules_to_preserve_errors": null,
+  "modules_to_quantize": null,
+  "modules_to_save": null,
+  "onnx_export": false,
+  "optimization_level": 3,
+  "orthogonal_init": false,
+  "peft_type": "MARS",
+  "quant_n_bits": 4,
+  "r": 8,
+  "revision": null,
+  "seed": 42,
+  "shared_r": 8,
+  "target_modules": [
+    "o_proj",
+    "k_proj",
+    "down_proj",
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "q_proj"
+  ],
+  "task_type": null
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-hellaswag-r8-a2/eval_results.json

@@ -0,0 +1,4 @@
+{
+    "task": "hellaswag",
+    "results": 0.813483369846644
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-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": "mars",
+    "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": 5233536
+  },
+  "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-mars-hellaswag-r8-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-mars-opt3-q4-new/TinyLlama_v1.1-mars-hellaswag-r8-a2",
+  "seed": 42,
+  "timestamp": "2025-08-30T01:39:19.750745"
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-logiqa-r2-a2/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha": 4,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "enabled_mlp": true,
+  "enabled_qkv": [
+    "q",
+    "k",
+    "v"
+  ],
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "mixture": false,
+  "modules_to_preserve_errors": null,
+  "modules_to_quantize": null,
+  "modules_to_save": null,
+  "onnx_export": false,
+  "optimization_level": 3,
+  "orthogonal_init": false,
+  "peft_type": "MARS",
+  "quant_n_bits": 4,
+  "r": 2,
+  "revision": null,
+  "seed": 42,
+  "shared_r": 2,
+  "target_modules": [
+    "o_proj",
+    "k_proj",
+    "down_proj",
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "q_proj"
+  ],
+  "task_type": null
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-logiqa-r2-a2/eval_results.json

@@ -0,0 +1,4 @@
+{
+    "task": "logiqa",
+    "results": 0.27101838463127453
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-logiqa-r2-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": "mars",
+    "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": 1307064
+  },
+  "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-mars-logiqa-r2-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-mars-opt3-q4-new/TinyLlama_v1.1-mars-logiqa-r2-a2",
+  "seed": 42,
+  "timestamp": "2025-08-29T15:14:52.117394"
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-logiqa-r8-a2/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha": 16,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "enabled_mlp": true,
+  "enabled_qkv": [
+    "q",
+    "k",
+    "v"
+  ],
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "mixture": false,
+  "modules_to_preserve_errors": null,
+  "modules_to_quantize": null,
+  "modules_to_save": null,
+  "onnx_export": false,
+  "optimization_level": 3,
+  "orthogonal_init": false,
+  "peft_type": "MARS",
+  "quant_n_bits": 4,
+  "r": 8,
+  "revision": null,
+  "seed": 42,
+  "shared_r": 8,
+  "target_modules": [
+    "o_proj",
+    "k_proj",
+    "down_proj",
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "q_proj"
+  ],
+  "task_type": null
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-logiqa-r8-a2/eval_results.json

@@ -0,0 +1,4 @@
+{
+    "task": "logiqa",
+    "results": 0.40921297252633754
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-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": "mars",
+    "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": 5233536
+  },
+  "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-mars-logiqa-r8-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-mars-opt3-q4-new/TinyLlama_v1.1-mars-logiqa-r8-a2",
+  "seed": 42,
+  "timestamp": "2025-08-29T22:21:48.612520"
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-winogrande-r32-a2/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha": 64,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "enabled_mlp": true,
+  "enabled_qkv": [
+    "q",
+    "k",
+    "v"
+  ],
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "mixture": false,
+  "modules_to_preserve_errors": null,
+  "modules_to_quantize": null,
+  "modules_to_save": null,
+  "onnx_export": false,
+  "optimization_level": 3,
+  "orthogonal_init": false,
+  "peft_type": "MARS",
+  "quant_n_bits": 4,
+  "r": 32,
+  "revision": null,
+  "seed": 42,
+  "shared_r": 32,
+  "target_modules": [
+    "o_proj",
+    "k_proj",
+    "down_proj",
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "q_proj"
+  ],
+  "task_type": null
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-winogrande-r32-a2/eval_results.json

@@ -0,0 +1,4 @@
+{
+    "task": "winogrande",
+    "results": 0.516179952644041
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-winogrande-r32-a2/training_configuration.json

@@ -0,0 +1,38 @@
+{
+  "model_id": "TinyLlama/TinyLlama_v1.1",
+  "dataset": {
+    "name": "WINOGRANDE",
+    "dataset_id": "allenai/winogrande",
+    "preprocess_id": "winogrande_train_deepeval"
+  },
+  "peft_config": {
+    "method": "mars",
+    "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": 21018624
+  },
+  "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-mars-winogrande-r32-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-mars-opt3-q4-new/TinyLlama_v1.1-mars-winogrande-r32-a2",
+  "seed": 42,
+  "timestamp": "2025-08-30T08:10:27.000811"
+}

TinyLlama_v1.1-mars-opt0-q4/TinyLlama_v1.1-mars-winogrande-r8-a2/training_configuration.json

@@ -0,0 +1,38 @@
+{
+  "model_id": "TinyLlama/TinyLlama_v1.1",
+  "dataset": {
+    "name": "WINOGRANDE",
+    "dataset_id": "allenai/winogrande",
+    "preprocess_id": "winogrande_train_deepeval"
+  },
+  "peft_config": {
+    "method": "mars",
+    "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": 5233536
+  },
+  "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-mars-winogrande-r8-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-mars-opt3-q4-new/TinyLlama_v1.1-mars-winogrande-r8-a2",
+  "seed": 42,
+  "timestamp": "2025-08-30T01:04:10.381088"
+}

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

@@ -0,0 +1,40 @@
+{
+  "alpha": 4,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "enabled_mlp": true,
+  "enabled_qkv": [
+    "q",
+    "k",
+    "v"
+  ],
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "mixture": false,
+  "modules_to_preserve_errors": null,
+  "modules_to_quantize": null,
+  "modules_to_save": null,
+  "onnx_export": false,
+  "optimization_level": 0,
+  "orthogonal_init": false,
+  "peft_type": "MARS",
+  "quant_n_bits": 8,
+  "r": 2,
+  "revision": null,
+  "seed": 42,
+  "shared_r": 2,
+  "target_modules": [
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "gate_proj",
+    "up_proj",
+    "v_proj",
+    "k_proj"
+  ],
+  "task_type": null,
+  "use_bnb": true
+}

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

@@ -0,0 +1,4 @@
+{
+    "task": "arc_c",
+    "results": 0.6322525597269625
+}

TinyLlama_v1.1-qmars/TinyLlama_v1.1-qmars-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": "qmars",
+    "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": 1307064
+  },
+  "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-qmars-arc_c-r2-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-qmars/TinyLlama_v1.1-qmars-arc_c-r2-a2",
+  "seed": 42,
+  "timestamp": "2025-09-02T08:46:32.040412"
+}

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

@@ -0,0 +1,625 @@
+[
+  {
+    "step": 1,
+    "epoch": 0.05714285714285714,
+    "cpu_mem": 3.020107776,
+    "gpu_mem": 1.058196992,
+    "loss": 4.8349,
+    "grad_norm": 360.26629638671875,
+    "learning_rate": 4.285714285714285e-05
+  },
+  {
+    "step": 2,
+    "epoch": 0.11428571428571428,
+    "cpu_mem": 3.025219584,
+    "gpu_mem": 1.068685312,
+    "loss": 5.1818,
+    "grad_norm": 364.10589599609375,
+    "learning_rate": 8.57142857142857e-05
+  },
+  {
+    "step": 3,
+    "epoch": 0.17142857142857143,
+    "cpu_mem": 3.025416192,
+    "gpu_mem": 1.068716032,
+    "loss": 1.9128,
+    "grad_norm": 82.8514175415039,
+    "learning_rate": 0.00012857142857142855
+  },
+  {
+    "step": 4,
+    "epoch": 0.22857142857142856,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.06868224,
+    "loss": 1.5444,
+    "grad_norm": 44.257389068603516,
+    "learning_rate": 0.0001714285714285714
+  },
+  {
+    "step": 5,
+    "epoch": 0.2857142857142857,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068669952,
+    "loss": 1.4274,
+    "grad_norm": 38.661521911621094,
+    "learning_rate": 0.00021428571428571427
+  },
+  {
+    "step": 6,
+    "epoch": 0.34285714285714286,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068732928,
+    "loss": 1.4406,
+    "grad_norm": 26.17365837097168,
+    "learning_rate": 0.0002571428571428571
+  },
+  {
+    "step": 7,
+    "epoch": 0.4,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068739072,
+    "loss": 1.522,
+    "grad_norm": 31.642549514770508,
+    "learning_rate": 0.0003
+  },
+  {
+    "step": 8,
+    "epoch": 0.45714285714285713,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.0686976,
+    "loss": 1.5263,
+    "grad_norm": 30.574464797973633,
+    "learning_rate": 0.00029980111348272456
+  },
+  {
+    "step": 9,
+    "epoch": 0.5142857142857142,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068692992,
+    "loss": 1.6581,
+    "grad_norm": 56.67964553833008,
+    "learning_rate": 0.00029920498134218835
+  },
+  {
+    "step": 10,
+    "epoch": 0.5714285714285714,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.06868224,
+    "loss": 1.4143,
+    "grad_norm": 13.563467979431152,
+    "learning_rate": 0.0002982131844136615
+  },
+  {
+    "step": 11,
+    "epoch": 0.6285714285714286,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068692992,
+    "loss": 1.372,
+    "grad_norm": 8.824348449707031,
+    "learning_rate": 0.0002968283527643036
+  },
+  {
+    "step": 12,
+    "epoch": 0.6857142857142857,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068717568,
+    "loss": 1.4953,
+    "grad_norm": 19.334678649902344,
+    "learning_rate": 0.000295054158718698
+  },
+  {
+    "step": 13,
+    "epoch": 0.7428571428571429,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068717568,
+    "loss": 1.4141,
+    "grad_norm": 17.438413619995117,
+    "learning_rate": 0.00029289530712050735
+  },
+  {
+    "step": 14,
+    "epoch": 0.8,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068665344,
+    "loss": 1.4638,
+    "grad_norm": 11.587946891784668,
+    "learning_rate": 0.000290357522856074
+  },
+  {
+    "step": 15,
+    "epoch": 0.8571428571428571,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068740608,
+    "loss": 1.4139,
+    "grad_norm": 7.782718181610107,
+    "learning_rate": 0.0002874475356730507
+  },
+  {
+    "step": 16,
+    "epoch": 0.9142857142857143,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068734464,
+    "loss": 1.4766,
+    "grad_norm": 12.382817268371582,
+    "learning_rate": 0.0002841730623343193
+  },
+  {
+    "step": 17,
+    "epoch": 0.9714285714285714,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068739072,
+    "loss": 1.4863,
+    "grad_norm": 10.532707214355469,
+    "learning_rate": 0.00028054278615452326
+  },
+  {
+    "step": 18,
+    "epoch": 1.0285714285714285,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073945088,
+    "loss": 2.0762,
+    "grad_norm": 13.852774620056152,
+    "learning_rate": 0.0002765663339734778
+  },
+  {
+    "step": 19,
+    "epoch": 1.0857142857142856,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073943552,
+    "loss": 1.39,
+    "grad_norm": 9.768491744995117,
+    "learning_rate": 0.00027225425062752165
+  },
+  {
+    "step": 20,
+    "epoch": 1.1428571428571428,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073918976,
+    "loss": 1.3337,
+    "grad_norm": 11.261113166809082,
+    "learning_rate": 0.0002676179709865066
+  },
+  {
+    "step": 21,
+    "epoch": 1.2,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073926656,
+    "loss": 1.3633,
+    "grad_norm": 10.808806419372559,
+    "learning_rate": 0.0002626697896305779
+  },
+  {
+    "step": 22,
+    "epoch": 1.2571428571428571,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.07395584,
+    "loss": 1.3556,
+    "grad_norm": 14.811458587646484,
+    "learning_rate": 0.000257422828247159
+  },
+  {
+    "step": 23,
+    "epoch": 1.3142857142857143,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073985024,
+    "loss": 1.3555,
+    "grad_norm": 17.6370906829834,
+    "learning_rate": 0.00025189100083459397
+  },
+  {
+    "step": 24,
+    "epoch": 1.3714285714285714,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073928192,
+    "loss": 1.3234,
+    "grad_norm": 11.551438331604004,
+    "learning_rate": 0.0002460889768047263
+  },
+  {
+    "step": 25,
+    "epoch": 1.4285714285714286,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073997312,
+    "loss": 1.3358,
+    "grad_norm": 14.671205520629883,
+    "learning_rate": 0.00024003214208225522
+  },
+  {
+    "step": 26,
+    "epoch": 1.4857142857142858,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073954304,
+    "loss": 1.3709,
+    "grad_norm": 16.04226303100586,
+    "learning_rate": 0.00023373655830402968
+  },
+  {
+    "step": 27,
+    "epoch": 1.5428571428571427,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073912832,
+    "loss": 1.391,
+    "grad_norm": 14.387223243713379,
+    "learning_rate": 0.00022721892022647462
+  },
+  {
+    "step": 28,
+    "epoch": 1.6,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073958912,
+    "loss": 1.6496,
+    "grad_norm": 26.753936767578125,
+    "learning_rate": 0.000220496511454098
+  },
+  {
+    "step": 29,
+    "epoch": 1.657142857142857,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073954304,
+    "loss": 1.3764,
+    "grad_norm": 10.248379707336426,
+    "learning_rate": 0.0002135871586064791
+  },
+  {
+    "step": 30,
+    "epoch": 1.7142857142857144,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073943552,
+    "loss": 1.3444,
+    "grad_norm": 8.105730056762695,
+    "learning_rate": 0.00020650918404527775
+  },
+  {
+    "step": 31,
+    "epoch": 1.7714285714285714,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073974272,
+    "loss": 1.3598,
+    "grad_norm": 9.523242950439453,
+    "learning_rate": 0.00019928135728662522
+  },
+  {
+    "step": 32,
+    "epoch": 1.8285714285714287,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073983488,
+    "loss": 1.3474,
+    "grad_norm": 10.56064224243164,
+    "learning_rate": 0.00019192284522774142
+  },
+  {
+    "step": 33,
+    "epoch": 1.8857142857142857,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.07396352,
+    "loss": 1.3986,
+    "grad_norm": 15.865970611572266,
+    "learning_rate": 0.00018445316131976934
+  },
+  {
+    "step": 34,
+    "epoch": 1.9428571428571428,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073942016,
+    "loss": 1.4546,
+    "grad_norm": 21.07072639465332,
+    "learning_rate": 0.00017689211382161034
+  },
+  {
+    "step": 35,
+    "epoch": 2.0,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073829888,
+    "loss": 2.0834,
+    "grad_norm": 17.78594398498535,
+    "learning_rate": 0.00016925975327198266
+  },
+  {
+    "step": 36,
+    "epoch": 2.057142857142857,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068711424,
+    "loss": 1.2406,
+    "grad_norm": 12.084159851074219,
+    "learning_rate": 0.00016157631931899697
+  },
+  {
+    "step": 37,
+    "epoch": 2.1142857142857143,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.06872064,
+    "loss": 1.5605,
+    "grad_norm": 59.46875762939453,
+    "learning_rate": 0.0001538621870482483
+  },
+  {
+    "step": 38,
+    "epoch": 2.1714285714285713,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068691456,
+    "loss": 1.3837,
+    "grad_norm": 37.64811706542969,
+    "learning_rate": 0.00014613781295175172
+  },
+  {
+    "step": 39,
+    "epoch": 2.2285714285714286,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068709888,
+    "loss": 1.3123,
+    "grad_norm": 21.93562889099121,
+    "learning_rate": 0.00013842368068100303
+  },
+  {
+    "step": 40,
+    "epoch": 2.2857142857142856,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068686848,
+    "loss": 1.3359,
+    "grad_norm": 17.83315658569336,
+    "learning_rate": 0.00013074024672801731
+  },
+  {
+    "step": 41,
+    "epoch": 2.342857142857143,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068688384,
+    "loss": 1.2908,
+    "grad_norm": 12.607256889343262,
+    "learning_rate": 0.00012310788617838966
+  },
+  {
+    "step": 42,
+    "epoch": 2.4,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068717568,
+    "loss": 1.3378,
+    "grad_norm": 29.468717575073242,
+    "learning_rate": 0.00011554683868023067
+  },
+  {
+    "step": 43,
+    "epoch": 2.4571428571428573,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068732928,
+    "loss": 1.2618,
+    "grad_norm": 13.891093254089355,
+    "learning_rate": 0.00010807715477225858
+  },
+  {
+    "step": 44,
+    "epoch": 2.5142857142857142,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.06875136,
+    "loss": 1.2953,
+    "grad_norm": 20.18376922607422,
+    "learning_rate": 0.00010071864271337478
+  },
+  {
+    "step": 45,
+    "epoch": 2.571428571428571,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.06870528,
+    "loss": 1.2423,
+    "grad_norm": 12.552496910095215,
+    "learning_rate": 9.34908159547222e-05
+  },
+  {
+    "step": 46,
+    "epoch": 2.6285714285714286,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068699136,
+    "loss": 1.197,
+    "grad_norm": 11.219213485717773,
+    "learning_rate": 8.641284139352091e-05
+  },
+  {
+    "step": 47,
+    "epoch": 2.685714285714286,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068692992,
+    "loss": 1.1792,
+    "grad_norm": 11.913239479064941,
+    "learning_rate": 7.950348854590204e-05
+  },
+  {
+    "step": 48,
+    "epoch": 2.742857142857143,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.0686976,
+    "loss": 1.0754,
+    "grad_norm": 11.84579086303711,
+    "learning_rate": 7.278107977352543e-05
+  },
+  {
+    "step": 49,
+    "epoch": 2.8,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068688384,
+    "loss": 1.1704,
+    "grad_norm": 11.788681030273438,
+    "learning_rate": 6.626344169597031e-05
+  },
+  {
+    "step": 50,
+    "epoch": 2.857142857142857,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068669952,
+    "loss": 1.1673,
+    "grad_norm": 12.49934196472168,
+    "learning_rate": 5.996785791774478e-05
+  },
+  {
+    "step": 51,
+    "epoch": 2.914285714285714,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068694528,
+    "loss": 1.1689,
+    "grad_norm": 20.719552993774414,
+    "learning_rate": 5.391102319527373e-05
+  },
+  {
+    "step": 52,
+    "epoch": 2.9714285714285715,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.068722176,
+    "loss": 1.2007,
+    "grad_norm": 18.217304229736328,
+    "learning_rate": 4.8108999165406026e-05
+  },
+  {
+    "step": 53,
+    "epoch": 3.0285714285714285,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.07394048,
+    "loss": 1.6113,
+    "grad_norm": 27.60318374633789,
+    "learning_rate": 4.257717175284103e-05
+  },
+  {
+    "step": 54,
+    "epoch": 3.085714285714286,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.07390976,
+    "loss": 1.0235,
+    "grad_norm": 18.013338088989258,
+    "learning_rate": 3.733021036942205e-05
+  },
+  {
+    "step": 55,
+    "epoch": 3.142857142857143,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073943552,
+    "loss": 1.009,
+    "grad_norm": 22.722183227539062,
+    "learning_rate": 3.238202901349345e-05
+  },
+  {
+    "step": 56,
+    "epoch": 3.2,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.07401728,
+    "loss": 1.0239,
+    "grad_norm": 21.968795776367188,
+    "learning_rate": 2.774574937247831e-05
+  },
+  {
+    "step": 57,
+    "epoch": 3.257142857142857,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073960448,
+    "loss": 0.8981,
+    "grad_norm": 19.004884719848633,
+    "learning_rate": 2.3433666026522153e-05
+  },
+  {
+    "step": 58,
+    "epoch": 3.314285714285714,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073954304,
+    "loss": 0.8651,
+    "grad_norm": 24.159976959228516,
+    "learning_rate": 1.945721384547671e-05
+  },
+  {
+    "step": 59,
+    "epoch": 3.3714285714285714,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.074004992,
+    "loss": 0.8833,
+    "grad_norm": 21.838836669921875,
+    "learning_rate": 1.5826937665680693e-05
+  },
+  {
+    "step": 60,
+    "epoch": 3.4285714285714284,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073931264,
+    "loss": 0.9072,
+    "grad_norm": 27.530853271484375,
+    "learning_rate": 1.2552464326949302e-05
+  },
+  {
+    "step": 61,
+    "epoch": 3.4857142857142858,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073945088,
+    "loss": 0.9804,
+    "grad_norm": 28.941020965576172,
+    "learning_rate": 9.64247714392597e-06
+  },
+  {
+    "step": 62,
+    "epoch": 3.5428571428571427,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073946624,
+    "loss": 0.9688,
+    "grad_norm": 28.280624389648438,
+    "learning_rate": 7.104692879492624e-06
+  },
+  {
+    "step": 63,
+    "epoch": 3.6,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073935872,
+    "loss": 1.0072,
+    "grad_norm": 28.708356857299805,
+    "learning_rate": 4.945841281301943e-06
+  },
+  {
+    "step": 64,
+    "epoch": 3.657142857142857,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073952768,
+    "loss": 0.9184,
+    "grad_norm": 29.564748764038086,
+    "learning_rate": 3.1716472356963286e-06
+  },
+  {
+    "step": 65,
+    "epoch": 3.7142857142857144,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073974272,
+    "loss": 0.8999,
+    "grad_norm": 42.21194839477539,
+    "learning_rate": 1.7868155863384415e-06
+  },
+  {
+    "step": 66,
+    "epoch": 3.7714285714285714,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073965056,
+    "loss": 0.8607,
+    "grad_norm": 30.265111923217773,
+    "learning_rate": 7.950186578116413e-07
+  },
+  {
+    "step": 67,
+    "epoch": 3.8285714285714287,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073991168,
+    "loss": 0.8626,
+    "grad_norm": 28.273231506347656,
+    "learning_rate": 1.988865172754206e-07
+  },
+  {
+    "step": 68,
+    "epoch": 3.8857142857142857,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073942016,
+    "loss": 0.8573,
+    "grad_norm": 29.98070526123047,
+    "learning_rate": 0.0
+  },
+  {
+    "step": 68,
+    "epoch": 3.8857142857142857,
+    "cpu_mem": 3.0256128,
+    "gpu_mem": 1.073942016,
+    "train_runtime": 372.3545,
+    "train_samples_per_second": 12.021,
+    "train_steps_per_second": 0.183,
+    "total_flos": 0.0,
+    "train_loss": 1.4101540008011986
+  }
+]

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

@@ -0,0 +1,40 @@
+{
+  "alpha": 64,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "enabled_mlp": true,
+  "enabled_qkv": [
+    "q",
+    "k",
+    "v"
+  ],
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "mixture": false,
+  "modules_to_preserve_errors": null,
+  "modules_to_quantize": null,
+  "modules_to_save": null,
+  "onnx_export": false,
+  "optimization_level": 0,
+  "orthogonal_init": false,
+  "peft_type": "MARS",
+  "quant_n_bits": 8,
+  "r": 32,
+  "revision": null,
+  "seed": 42,
+  "shared_r": 32,
+  "target_modules": [
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "gate_proj",
+    "up_proj",
+    "v_proj",
+    "k_proj"
+  ],
+  "task_type": null,
+  "use_bnb": true
+}

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

@@ -0,0 +1,4 @@
+{
+    "task": "arc_c",
+    "results": 0.6109215017064846
+}

TinyLlama_v1.1-qmars/TinyLlama_v1.1-qmars-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": "qmars",
+    "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": 21018624
+  },
+  "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-qmars-arc_c-r32-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-qmars/TinyLlama_v1.1-qmars-arc_c-r32-a2",
+  "seed": 42,
+  "timestamp": "2025-09-02T21:50:02.458769"
+}

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

@@ -0,0 +1,625 @@
+[
+  {
+    "step": 1,
+    "epoch": 0.05714285714285714,
+    "cpu_mem": 2.860781568,
+    "gpu_mem": 1.137022464,
+    "loss": 4.8349,
+    "grad_norm": 75.46897888183594,
+    "learning_rate": 4.285714285714285e-05
+  },
+  {
+    "step": 2,
+    "epoch": 0.11428571428571428,
+    "cpu_mem": 2.865893376,
+    "gpu_mem": 1.305161728,
+    "loss": 5.1818,
+    "grad_norm": 76.17707824707031,
+    "learning_rate": 8.57142857142857e-05
+  },
+  {
+    "step": 3,
+    "epoch": 0.17142857142857143,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305192448,
+    "loss": 2.218,
+    "grad_norm": 29.5435733795166,
+    "learning_rate": 0.00012857142857142855
+  },
+  {
+    "step": 4,
+    "epoch": 0.22857142857142856,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305158656,
+    "loss": 1.5022,
+    "grad_norm": 6.774408340454102,
+    "learning_rate": 0.0001714285714285714
+  },
+  {
+    "step": 5,
+    "epoch": 0.2857142857142857,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305146368,
+    "loss": 1.4632,
+    "grad_norm": 7.44521427154541,
+    "learning_rate": 0.00021428571428571427
+  },
+  {
+    "step": 6,
+    "epoch": 0.34285714285714286,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305209344,
+    "loss": 1.4715,
+    "grad_norm": 7.740568161010742,
+    "learning_rate": 0.0002571428571428571
+  },
+  {
+    "step": 7,
+    "epoch": 0.4,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305215488,
+    "loss": 1.5742,
+    "grad_norm": 8.267945289611816,
+    "learning_rate": 0.0003
+  },
+  {
+    "step": 8,
+    "epoch": 0.45714285714285713,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305174016,
+    "loss": 1.3934,
+    "grad_norm": 3.597435235977173,
+    "learning_rate": 0.00029980111348272456
+  },
+  {
+    "step": 9,
+    "epoch": 0.5142857142857142,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305169408,
+    "loss": 1.3452,
+    "grad_norm": 2.5027928352355957,
+    "learning_rate": 0.00029920498134218835
+  },
+  {
+    "step": 10,
+    "epoch": 0.5714285714285714,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305158656,
+    "loss": 1.4474,
+    "grad_norm": 3.489070415496826,
+    "learning_rate": 0.0002982131844136615
+  },
+  {
+    "step": 11,
+    "epoch": 0.6285714285714286,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305169408,
+    "loss": 1.3633,
+    "grad_norm": 1.7864291667938232,
+    "learning_rate": 0.0002968283527643036
+  },
+  {
+    "step": 12,
+    "epoch": 0.6857142857142857,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305193984,
+    "loss": 1.4912,
+    "grad_norm": 4.657872200012207,
+    "learning_rate": 0.000295054158718698
+  },
+  {
+    "step": 13,
+    "epoch": 0.7428571428571429,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305193984,
+    "loss": 1.4818,
+    "grad_norm": 5.277657985687256,
+    "learning_rate": 0.00029289530712050735
+  },
+  {
+    "step": 14,
+    "epoch": 0.8,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.30514176,
+    "loss": 1.5029,
+    "grad_norm": 2.831676959991455,
+    "learning_rate": 0.000290357522856074
+  },
+  {
+    "step": 15,
+    "epoch": 0.8571428571428571,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305217024,
+    "loss": 1.5351,
+    "grad_norm": 3.477987051010132,
+    "learning_rate": 0.0002874475356730507
+  },
+  {
+    "step": 16,
+    "epoch": 0.9142857142857143,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.30521088,
+    "loss": 1.3817,
+    "grad_norm": 1.5456229448318481,
+    "learning_rate": 0.0002841730623343193
+  },
+  {
+    "step": 17,
+    "epoch": 0.9714285714285714,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305215488,
+    "loss": 1.4314,
+    "grad_norm": 2.3160977363586426,
+    "learning_rate": 0.00028054278615452326
+  },
+  {
+    "step": 18,
+    "epoch": 1.0285714285714285,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389246976,
+    "loss": 2.0422,
+    "grad_norm": 2.573564052581787,
+    "learning_rate": 0.0002765663339734778
+  },
+  {
+    "step": 19,
+    "epoch": 1.0857142857142856,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.38924544,
+    "loss": 1.3818,
+    "grad_norm": 2.031684637069702,
+    "learning_rate": 0.00027225425062752165
+  },
+  {
+    "step": 20,
+    "epoch": 1.1428571428571428,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389220864,
+    "loss": 1.3827,
+    "grad_norm": 2.671165704727173,
+    "learning_rate": 0.0002676179709865066
+  },
+  {
+    "step": 21,
+    "epoch": 1.2,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389228544,
+    "loss": 1.3347,
+    "grad_norm": 1.5832688808441162,
+    "learning_rate": 0.0002626697896305779
+  },
+  {
+    "step": 22,
+    "epoch": 1.2571428571428571,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389257728,
+    "loss": 1.3429,
+    "grad_norm": 2.4159576892852783,
+    "learning_rate": 0.000257422828247159
+  },
+  {
+    "step": 23,
+    "epoch": 1.3142857142857143,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389286912,
+    "loss": 1.29,
+    "grad_norm": 1.643193006515503,
+    "learning_rate": 0.00025189100083459397
+  },
+  {
+    "step": 24,
+    "epoch": 1.3714285714285714,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.38923008,
+    "loss": 1.4058,
+    "grad_norm": 2.3235573768615723,
+    "learning_rate": 0.0002460889768047263
+  },
+  {
+    "step": 25,
+    "epoch": 1.4285714285714286,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.3892992,
+    "loss": 1.3284,
+    "grad_norm": 2.1304330825805664,
+    "learning_rate": 0.00024003214208225522
+  },
+  {
+    "step": 26,
+    "epoch": 1.4857142857142858,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389256192,
+    "loss": 1.3167,
+    "grad_norm": 1.504705786705017,
+    "learning_rate": 0.00023373655830402968
+  },
+  {
+    "step": 27,
+    "epoch": 1.5428571428571427,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.38921472,
+    "loss": 1.3632,
+    "grad_norm": 2.2097530364990234,
+    "learning_rate": 0.00022721892022647462
+  },
+  {
+    "step": 28,
+    "epoch": 1.6,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.3892608,
+    "loss": 1.5934,
+    "grad_norm": 4.348984241485596,
+    "learning_rate": 0.000220496511454098
+  },
+  {
+    "step": 29,
+    "epoch": 1.657142857142857,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389256192,
+    "loss": 1.3625,
+    "grad_norm": 1.59405517578125,
+    "learning_rate": 0.0002135871586064791
+  },
+  {
+    "step": 30,
+    "epoch": 1.7142857142857144,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.38924544,
+    "loss": 1.3605,
+    "grad_norm": 2.354477643966675,
+    "learning_rate": 0.00020650918404527775
+  },
+  {
+    "step": 31,
+    "epoch": 1.7714285714285714,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.38927616,
+    "loss": 1.3167,
+    "grad_norm": 1.2073973417282104,
+    "learning_rate": 0.00019928135728662522
+  },
+  {
+    "step": 32,
+    "epoch": 1.8285714285714287,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389285376,
+    "loss": 1.3462,
+    "grad_norm": 1.477156639099121,
+    "learning_rate": 0.00019192284522774142
+  },
+  {
+    "step": 33,
+    "epoch": 1.8857142857142857,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389265408,
+    "loss": 1.4029,
+    "grad_norm": 2.4271631240844727,
+    "learning_rate": 0.00018445316131976934
+  },
+  {
+    "step": 34,
+    "epoch": 1.9428571428571428,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389243904,
+    "loss": 1.4064,
+    "grad_norm": 1.9992505311965942,
+    "learning_rate": 0.00017689211382161034
+  },
+  {
+    "step": 35,
+    "epoch": 2.0,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389131776,
+    "loss": 2.0767,
+    "grad_norm": 2.3909573554992676,
+    "learning_rate": 0.00016925975327198266
+  },
+  {
+    "step": 36,
+    "epoch": 2.057142857142857,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.30518784,
+    "loss": 1.3289,
+    "grad_norm": 1.7182539701461792,
+    "learning_rate": 0.00016157631931899697
+  },
+  {
+    "step": 37,
+    "epoch": 2.1142857142857143,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305197056,
+    "loss": 1.3174,
+    "grad_norm": 1.4748281240463257,
+    "learning_rate": 0.0001538621870482483
+  },
+  {
+    "step": 38,
+    "epoch": 2.1714285714285713,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305167872,
+    "loss": 1.3186,
+    "grad_norm": 1.1838716268539429,
+    "learning_rate": 0.00014613781295175172
+  },
+  {
+    "step": 39,
+    "epoch": 2.2285714285714286,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305186304,
+    "loss": 1.3067,
+    "grad_norm": 1.746688723564148,
+    "learning_rate": 0.00013842368068100303
+  },
+  {
+    "step": 40,
+    "epoch": 2.2857142857142856,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305163264,
+    "loss": 1.316,
+    "grad_norm": 1.3366076946258545,
+    "learning_rate": 0.00013074024672801731
+  },
+  {
+    "step": 41,
+    "epoch": 2.342857142857143,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.3051648,
+    "loss": 1.3113,
+    "grad_norm": 2.173269510269165,
+    "learning_rate": 0.00012310788617838966
+  },
+  {
+    "step": 42,
+    "epoch": 2.4,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305193984,
+    "loss": 1.287,
+    "grad_norm": 1.9705522060394287,
+    "learning_rate": 0.00011554683868023067
+  },
+  {
+    "step": 43,
+    "epoch": 2.4571428571428573,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305209344,
+    "loss": 1.2743,
+    "grad_norm": 1.440999984741211,
+    "learning_rate": 0.00010807715477225858
+  },
+  {
+    "step": 44,
+    "epoch": 2.5142857142857142,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305227776,
+    "loss": 1.2127,
+    "grad_norm": 1.4414082765579224,
+    "learning_rate": 0.00010071864271337478
+  },
+  {
+    "step": 45,
+    "epoch": 2.571428571428571,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305181696,
+    "loss": 1.2363,
+    "grad_norm": 1.410117745399475,
+    "learning_rate": 9.34908159547222e-05
+  },
+  {
+    "step": 46,
+    "epoch": 2.6285714285714286,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305175552,
+    "loss": 1.1618,
+    "grad_norm": 1.3951239585876465,
+    "learning_rate": 8.641284139352091e-05
+  },
+  {
+    "step": 47,
+    "epoch": 2.685714285714286,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305169408,
+    "loss": 1.1931,
+    "grad_norm": 1.4868206977844238,
+    "learning_rate": 7.950348854590204e-05
+  },
+  {
+    "step": 48,
+    "epoch": 2.742857142857143,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305174016,
+    "loss": 1.1658,
+    "grad_norm": 2.2503466606140137,
+    "learning_rate": 7.278107977352543e-05
+  },
+  {
+    "step": 49,
+    "epoch": 2.8,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.3051648,
+    "loss": 1.1666,
+    "grad_norm": 2.3152573108673096,
+    "learning_rate": 6.626344169597031e-05
+  },
+  {
+    "step": 50,
+    "epoch": 2.857142857142857,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305146368,
+    "loss": 1.1869,
+    "grad_norm": 2.246035575866699,
+    "learning_rate": 5.996785791774478e-05
+  },
+  {
+    "step": 51,
+    "epoch": 2.914285714285714,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305170944,
+    "loss": 1.1132,
+    "grad_norm": 2.2531511783599854,
+    "learning_rate": 5.391102319527373e-05
+  },
+  {
+    "step": 52,
+    "epoch": 2.9714285714285715,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.305198592,
+    "loss": 1.2356,
+    "grad_norm": 2.720115900039673,
+    "learning_rate": 4.8108999165406026e-05
+  },
+  {
+    "step": 53,
+    "epoch": 3.0285714285714285,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389242368,
+    "loss": 1.7441,
+    "grad_norm": 3.720527172088623,
+    "learning_rate": 4.257717175284103e-05
+  },
+  {
+    "step": 54,
+    "epoch": 3.085714285714286,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389211648,
+    "loss": 1.0292,
+    "grad_norm": 2.4771134853363037,
+    "learning_rate": 3.733021036942205e-05
+  },
+  {
+    "step": 55,
+    "epoch": 3.142857142857143,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.38924544,
+    "loss": 1.0343,
+    "grad_norm": 3.095989942550659,
+    "learning_rate": 3.238202901349345e-05
+  },
+  {
+    "step": 56,
+    "epoch": 3.2,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389319168,
+    "loss": 1.0445,
+    "grad_norm": 3.5430476665496826,
+    "learning_rate": 2.774574937247831e-05
+  },
+  {
+    "step": 57,
+    "epoch": 3.257142857142857,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389262336,
+    "loss": 0.9922,
+    "grad_norm": 2.634396553039551,
+    "learning_rate": 2.3433666026522153e-05
+  },
+  {
+    "step": 58,
+    "epoch": 3.314285714285714,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389256192,
+    "loss": 0.9425,
+    "grad_norm": 3.697744131088257,
+    "learning_rate": 1.945721384547671e-05
+  },
+  {
+    "step": 59,
+    "epoch": 3.3714285714285714,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.38930688,
+    "loss": 0.8918,
+    "grad_norm": 3.4737136363983154,
+    "learning_rate": 1.5826937665680693e-05
+  },
+  {
+    "step": 60,
+    "epoch": 3.4285714285714284,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389233152,
+    "loss": 1.091,
+    "grad_norm": 4.901792049407959,
+    "learning_rate": 1.2552464326949302e-05
+  },
+  {
+    "step": 61,
+    "epoch": 3.4857142857142858,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389246976,
+    "loss": 1.0153,
+    "grad_norm": 3.8909311294555664,
+    "learning_rate": 9.64247714392597e-06
+  },
+  {
+    "step": 62,
+    "epoch": 3.5428571428571427,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389248512,
+    "loss": 1.0389,
+    "grad_norm": 4.103649616241455,
+    "learning_rate": 7.104692879492624e-06
+  },
+  {
+    "step": 63,
+    "epoch": 3.6,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.38923776,
+    "loss": 1.0792,
+    "grad_norm": 4.326652526855469,
+    "learning_rate": 4.945841281301943e-06
+  },
+  {
+    "step": 64,
+    "epoch": 3.657142857142857,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389254656,
+    "loss": 1.0662,
+    "grad_norm": 4.315281391143799,
+    "learning_rate": 3.1716472356963286e-06
+  },
+  {
+    "step": 65,
+    "epoch": 3.7142857142857144,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.38927616,
+    "loss": 1.0428,
+    "grad_norm": 4.233317852020264,
+    "learning_rate": 1.7868155863384415e-06
+  },
+  {
+    "step": 66,
+    "epoch": 3.7714285714285714,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389266944,
+    "loss": 0.8833,
+    "grad_norm": 3.347691297531128,
+    "learning_rate": 7.950186578116413e-07
+  },
+  {
+    "step": 67,
+    "epoch": 3.8285714285714287,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389293056,
+    "loss": 0.8826,
+    "grad_norm": 3.1047513484954834,
+    "learning_rate": 1.988865172754206e-07
+  },
+  {
+    "step": 68,
+    "epoch": 3.8857142857142857,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389243904,
+    "loss": 1.0324,
+    "grad_norm": 3.6745681762695312,
+    "learning_rate": 0.0
+  },
+  {
+    "step": 68,
+    "epoch": 3.8857142857142857,
+    "cpu_mem": 2.866089984,
+    "gpu_mem": 1.389243904,
+    "train_runtime": 384.6701,
+    "train_samples_per_second": 11.636,
+    "train_steps_per_second": 0.177,
+    "total_flos": 0.0,
+    "train_loss": 1.420820819104419
+  }
+]

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

@@ -0,0 +1,40 @@
+{
+  "alpha": 16,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "enabled_mlp": true,
+  "enabled_qkv": [
+    "q",
+    "k",
+    "v"
+  ],
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "mixture": false,
+  "modules_to_preserve_errors": null,
+  "modules_to_quantize": null,
+  "modules_to_save": null,
+  "onnx_export": false,
+  "optimization_level": 0,
+  "orthogonal_init": false,
+  "peft_type": "MARS",
+  "quant_n_bits": 8,
+  "r": 8,
+  "revision": null,
+  "seed": 42,
+  "shared_r": 8,
+  "target_modules": [
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "gate_proj",
+    "up_proj",
+    "v_proj",
+    "k_proj"
+  ],
+  "task_type": null,
+  "use_bnb": true
+}

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

@@ -0,0 +1,4 @@
+{
+    "task": "arc_c",
+    "results": 0.6203071672354948
+}

TinyLlama_v1.1-qmars/TinyLlama_v1.1-qmars-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": "qmars",
+    "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": 5233536
+  },
+  "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-qmars-arc_c-r8-a2",
+  "output_dir": "./experiment_results/TinyLlama_v1.1-qmars/TinyLlama_v1.1-qmars-arc_c-r8-a2",
+  "seed": 42,
+  "timestamp": "2025-09-02T15:16:51.501968"
+}

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

@@ -0,0 +1,625 @@
+[
+  {
+    "step": 1,
+    "epoch": 0.05714285714285714,
+    "cpu_mem": 2.883121152,
+    "gpu_mem": 1.073887744,
+    "loss": 4.8349,
+    "grad_norm": 167.41305541992188,
+    "learning_rate": 4.285714285714285e-05
+  },
+  {
+    "step": 2,
+    "epoch": 0.11428571428571428,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115757568,
+    "loss": 5.1818,
+    "grad_norm": 169.5339813232422,
+    "learning_rate": 8.57142857142857e-05
+  },
+  {
+    "step": 3,
+    "epoch": 0.17142857142857143,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115788288,
+    "loss": 2.0398,
+    "grad_norm": 45.44222640991211,
+    "learning_rate": 0.00012857142857142855
+  },
+  {
+    "step": 4,
+    "epoch": 0.22857142857142856,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115754496,
+    "loss": 1.4888,
+    "grad_norm": 13.766551971435547,
+    "learning_rate": 0.0001714285714285714
+  },
+  {
+    "step": 5,
+    "epoch": 0.2857142857142857,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115742208,
+    "loss": 1.4426,
+    "grad_norm": 19.619434356689453,
+    "learning_rate": 0.00021428571428571427
+  },
+  {
+    "step": 6,
+    "epoch": 0.34285714285714286,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115805184,
+    "loss": 1.4259,
+    "grad_norm": 12.592857360839844,
+    "learning_rate": 0.0002571428571428571
+  },
+  {
+    "step": 7,
+    "epoch": 0.4,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115811328,
+    "loss": 1.4734,
+    "grad_norm": 10.968993186950684,
+    "learning_rate": 0.0003
+  },
+  {
+    "step": 8,
+    "epoch": 0.45714285714285713,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115769856,
+    "loss": 1.4618,
+    "grad_norm": 9.009696006774902,
+    "learning_rate": 0.00029980111348272456
+  },
+  {
+    "step": 9,
+    "epoch": 0.5142857142857142,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115765248,
+    "loss": 1.3015,
+    "grad_norm": 4.542705535888672,
+    "learning_rate": 0.00029920498134218835
+  },
+  {
+    "step": 10,
+    "epoch": 0.5714285714285714,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115754496,
+    "loss": 1.5272,
+    "grad_norm": 11.390000343322754,
+    "learning_rate": 0.0002982131844136615
+  },
+  {
+    "step": 11,
+    "epoch": 0.6285714285714286,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115765248,
+    "loss": 1.3546,
+    "grad_norm": 3.7129766941070557,
+    "learning_rate": 0.0002968283527643036
+  },
+  {
+    "step": 12,
+    "epoch": 0.6857142857142857,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115789824,
+    "loss": 1.5265,
+    "grad_norm": 9.75503921508789,
+    "learning_rate": 0.000295054158718698
+  },
+  {
+    "step": 13,
+    "epoch": 0.7428571428571429,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115789824,
+    "loss": 1.3401,
+    "grad_norm": 9.924088478088379,
+    "learning_rate": 0.00029289530712050735
+  },
+  {
+    "step": 14,
+    "epoch": 0.8,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.1157376,
+    "loss": 1.6946,
+    "grad_norm": 11.323637008666992,
+    "learning_rate": 0.000290357522856074
+  },
+  {
+    "step": 15,
+    "epoch": 0.8571428571428571,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115812864,
+    "loss": 1.5725,
+    "grad_norm": 8.125307083129883,
+    "learning_rate": 0.0002874475356730507
+  },
+  {
+    "step": 16,
+    "epoch": 0.9142857142857143,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.11580672,
+    "loss": 1.4913,
+    "grad_norm": 7.795830726623535,
+    "learning_rate": 0.0002841730623343193
+  },
+  {
+    "step": 17,
+    "epoch": 0.9714285714285714,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115811328,
+    "loss": 1.4569,
+    "grad_norm": 5.1771240234375,
+    "learning_rate": 0.00028054278615452326
+  },
+  {
+    "step": 18,
+    "epoch": 1.0285714285714285,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136708096,
+    "loss": 2.1211,
+    "grad_norm": 18.78937339782715,
+    "learning_rate": 0.0002765663339734778
+  },
+  {
+    "step": 19,
+    "epoch": 1.0857142857142856,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.13670656,
+    "loss": 1.4114,
+    "grad_norm": 4.584958553314209,
+    "learning_rate": 0.00027225425062752165
+  },
+  {
+    "step": 20,
+    "epoch": 1.1428571428571428,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136681984,
+    "loss": 1.3739,
+    "grad_norm": 7.459542751312256,
+    "learning_rate": 0.0002676179709865066
+  },
+  {
+    "step": 21,
+    "epoch": 1.2,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136689664,
+    "loss": 1.3671,
+    "grad_norm": 3.8549575805664062,
+    "learning_rate": 0.0002626697896305779
+  },
+  {
+    "step": 22,
+    "epoch": 1.2571428571428571,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136718848,
+    "loss": 1.3602,
+    "grad_norm": 4.934852123260498,
+    "learning_rate": 0.000257422828247159
+  },
+  {
+    "step": 23,
+    "epoch": 1.3142857142857143,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136748032,
+    "loss": 1.3096,
+    "grad_norm": 2.996155261993408,
+    "learning_rate": 0.00025189100083459397
+  },
+  {
+    "step": 24,
+    "epoch": 1.3714285714285714,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.1366912,
+    "loss": 1.3381,
+    "grad_norm": 3.403679370880127,
+    "learning_rate": 0.0002460889768047263
+  },
+  {
+    "step": 25,
+    "epoch": 1.4285714285714286,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.13676032,
+    "loss": 1.3265,
+    "grad_norm": 3.279336452484131,
+    "learning_rate": 0.00024003214208225522
+  },
+  {
+    "step": 26,
+    "epoch": 1.4857142857142858,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136717312,
+    "loss": 1.3729,
+    "grad_norm": 3.639723539352417,
+    "learning_rate": 0.00023373655830402968
+  },
+  {
+    "step": 27,
+    "epoch": 1.5428571428571427,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.13667584,
+    "loss": 1.3882,
+    "grad_norm": 4.425181865692139,
+    "learning_rate": 0.00022721892022647462
+  },
+  {
+    "step": 28,
+    "epoch": 1.6,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.13672192,
+    "loss": 1.5116,
+    "grad_norm": 6.004555702209473,
+    "learning_rate": 0.000220496511454098
+  },
+  {
+    "step": 29,
+    "epoch": 1.657142857142857,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136717312,
+    "loss": 1.3472,
+    "grad_norm": 3.100022792816162,
+    "learning_rate": 0.0002135871586064791
+  },
+  {
+    "step": 30,
+    "epoch": 1.7142857142857144,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.13670656,
+    "loss": 1.3379,
+    "grad_norm": 3.413043260574341,
+    "learning_rate": 0.00020650918404527775
+  },
+  {
+    "step": 31,
+    "epoch": 1.7714285714285714,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.13673728,
+    "loss": 1.3091,
+    "grad_norm": 1.8831239938735962,
+    "learning_rate": 0.00019928135728662522
+  },
+  {
+    "step": 32,
+    "epoch": 1.8285714285714287,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136746496,
+    "loss": 1.3402,
+    "grad_norm": 2.753385543823242,
+    "learning_rate": 0.00019192284522774142
+  },
+  {
+    "step": 33,
+    "epoch": 1.8857142857142857,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136726528,
+    "loss": 1.3955,
+    "grad_norm": 3.8380284309387207,
+    "learning_rate": 0.00018445316131976934
+  },
+  {
+    "step": 34,
+    "epoch": 1.9428571428571428,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136705024,
+    "loss": 1.3676,
+    "grad_norm": 2.57751202583313,
+    "learning_rate": 0.00017689211382161034
+  },
+  {
+    "step": 35,
+    "epoch": 2.0,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136592896,
+    "loss": 2.1042,
+    "grad_norm": 4.824999809265137,
+    "learning_rate": 0.00016925975327198266
+  },
+  {
+    "step": 36,
+    "epoch": 2.057142857142857,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.11578368,
+    "loss": 1.2838,
+    "grad_norm": 3.206289052963257,
+    "learning_rate": 0.00016157631931899697
+  },
+  {
+    "step": 37,
+    "epoch": 2.1142857142857143,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115792896,
+    "loss": 1.3517,
+    "grad_norm": 3.9317543506622314,
+    "learning_rate": 0.0001538621870482483
+  },
+  {
+    "step": 38,
+    "epoch": 2.1714285714285713,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115763712,
+    "loss": 1.3077,
+    "grad_norm": 3.159288167953491,
+    "learning_rate": 0.00014613781295175172
+  },
+  {
+    "step": 39,
+    "epoch": 2.2285714285714286,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115782144,
+    "loss": 1.3157,
+    "grad_norm": 3.4918620586395264,
+    "learning_rate": 0.00013842368068100303
+  },
+  {
+    "step": 40,
+    "epoch": 2.2857142857142856,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115759104,
+    "loss": 1.3421,
+    "grad_norm": 3.825528383255005,
+    "learning_rate": 0.00013074024672801731
+  },
+  {
+    "step": 41,
+    "epoch": 2.342857142857143,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.11576064,
+    "loss": 1.3183,
+    "grad_norm": 2.9959378242492676,
+    "learning_rate": 0.00012310788617838966
+  },
+  {
+    "step": 42,
+    "epoch": 2.4,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115789824,
+    "loss": 1.24,
+    "grad_norm": 4.1695404052734375,
+    "learning_rate": 0.00011554683868023067
+  },
+  {
+    "step": 43,
+    "epoch": 2.4571428571428573,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115805184,
+    "loss": 1.2721,
+    "grad_norm": 4.863770484924316,
+    "learning_rate": 0.00010807715477225858
+  },
+  {
+    "step": 44,
+    "epoch": 2.5142857142857142,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115823616,
+    "loss": 1.2515,
+    "grad_norm": 3.4958484172821045,
+    "learning_rate": 0.00010071864271337478
+  },
+  {
+    "step": 45,
+    "epoch": 2.571428571428571,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115777536,
+    "loss": 1.2826,
+    "grad_norm": 3.9481709003448486,
+    "learning_rate": 9.34908159547222e-05
+  },
+  {
+    "step": 46,
+    "epoch": 2.6285714285714286,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115771392,
+    "loss": 1.2037,
+    "grad_norm": 3.5659868717193604,
+    "learning_rate": 8.641284139352091e-05
+  },
+  {
+    "step": 47,
+    "epoch": 2.685714285714286,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115765248,
+    "loss": 1.1832,
+    "grad_norm": 4.207839488983154,
+    "learning_rate": 7.950348854590204e-05
+  },
+  {
+    "step": 48,
+    "epoch": 2.742857142857143,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115769856,
+    "loss": 1.093,
+    "grad_norm": 3.661602735519409,
+    "learning_rate": 7.278107977352543e-05
+  },
+  {
+    "step": 49,
+    "epoch": 2.8,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.11576064,
+    "loss": 1.1906,
+    "grad_norm": 4.302423000335693,
+    "learning_rate": 6.626344169597031e-05
+  },
+  {
+    "step": 50,
+    "epoch": 2.857142857142857,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115742208,
+    "loss": 1.1374,
+    "grad_norm": 4.260297775268555,
+    "learning_rate": 5.996785791774478e-05
+  },
+  {
+    "step": 51,
+    "epoch": 2.914285714285714,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115766784,
+    "loss": 1.2067,
+    "grad_norm": 6.081331253051758,
+    "learning_rate": 5.391102319527373e-05
+  },
+  {
+    "step": 52,
+    "epoch": 2.9714285714285715,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.115794432,
+    "loss": 1.1881,
+    "grad_norm": 6.451767444610596,
+    "learning_rate": 4.8108999165406026e-05
+  },
+  {
+    "step": 53,
+    "epoch": 3.0285714285714285,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136703488,
+    "loss": 1.7001,
+    "grad_norm": 9.44800090789795,
+    "learning_rate": 4.257717175284103e-05
+  },
+  {
+    "step": 54,
+    "epoch": 3.085714285714286,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136672768,
+    "loss": 1.0727,
+    "grad_norm": 5.197486400604248,
+    "learning_rate": 3.733021036942205e-05
+  },
+  {
+    "step": 55,
+    "epoch": 3.142857142857143,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.13670656,
+    "loss": 1.0685,
+    "grad_norm": 5.644598007202148,
+    "learning_rate": 3.238202901349345e-05
+  },
+  {
+    "step": 56,
+    "epoch": 3.2,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136780288,
+    "loss": 1.0943,
+    "grad_norm": 6.423681259155273,
+    "learning_rate": 2.774574937247831e-05
+  },
+  {
+    "step": 57,
+    "epoch": 3.257142857142857,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136723456,
+    "loss": 0.9465,
+    "grad_norm": 5.702960968017578,
+    "learning_rate": 2.3433666026522153e-05
+  },
+  {
+    "step": 58,
+    "epoch": 3.314285714285714,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136717312,
+    "loss": 0.9811,
+    "grad_norm": 6.766078472137451,
+    "learning_rate": 1.945721384547671e-05
+  },
+  {
+    "step": 59,
+    "epoch": 3.3714285714285714,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136768,
+    "loss": 0.9603,
+    "grad_norm": 6.710747241973877,
+    "learning_rate": 1.5826937665680693e-05
+  },
+  {
+    "step": 60,
+    "epoch": 3.4285714285714284,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136694272,
+    "loss": 1.0782,
+    "grad_norm": 7.888988018035889,
+    "learning_rate": 1.2552464326949302e-05
+  },
+  {
+    "step": 61,
+    "epoch": 3.4857142857142858,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136708096,
+    "loss": 0.9895,
+    "grad_norm": 6.987651348114014,
+    "learning_rate": 9.64247714392597e-06
+  },
+  {
+    "step": 62,
+    "epoch": 3.5428571428571427,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136709632,
+    "loss": 0.9953,
+    "grad_norm": 6.920251846313477,
+    "learning_rate": 7.104692879492624e-06
+  },
+  {
+    "step": 63,
+    "epoch": 3.6,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.13669888,
+    "loss": 1.0593,
+    "grad_norm": 7.611911773681641,
+    "learning_rate": 4.945841281301943e-06
+  },
+  {
+    "step": 64,
+    "epoch": 3.657142857142857,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136715776,
+    "loss": 0.9854,
+    "grad_norm": 8.091558456420898,
+    "learning_rate": 3.1716472356963286e-06
+  },
+  {
+    "step": 65,
+    "epoch": 3.7142857142857144,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.13673728,
+    "loss": 1.0049,
+    "grad_norm": 8.712056159973145,
+    "learning_rate": 1.7868155863384415e-06
+  },
+  {
+    "step": 66,
+    "epoch": 3.7714285714285714,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136728064,
+    "loss": 0.9963,
+    "grad_norm": 7.288658618927002,
+    "learning_rate": 7.950186578116413e-07
+  },
+  {
+    "step": 67,
+    "epoch": 3.8285714285714287,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136754176,
+    "loss": 0.9171,
+    "grad_norm": 6.7615966796875,
+    "learning_rate": 1.988865172754206e-07
+  },
+  {
+    "step": 68,
+    "epoch": 3.8857142857142857,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136705024,
+    "loss": 0.946,
+    "grad_norm": 8.281795501708984,
+    "learning_rate": 0.0
+  },
+  {
+    "step": 68,
+    "epoch": 3.8857142857142857,
+    "cpu_mem": 2.888429568,
+    "gpu_mem": 1.136705024,
+    "train_runtime": 372.291,
+    "train_samples_per_second": 12.023,
+    "train_steps_per_second": 0.183,
+    "total_flos": 0.0,
+    "train_loss": 1.4214746145641102
+  }
+]

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

@@ -0,0 +1,40 @@
+{
+  "alpha": 64,
+  "auto_mapping": null,
+  "base_model_name_or_path": "TinyLlama/TinyLlama_v1.1",
+  "bias": "none",
+  "enabled_mlp": true,
+  "enabled_qkv": [
+    "q",
+    "k",
+    "v"
+  ],
+  "fan_in_fan_out": false,
+  "inference_mode": false,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "mixture": false,
+  "modules_to_preserve_errors": null,
+  "modules_to_quantize": null,
+  "modules_to_save": null,
+  "onnx_export": false,
+  "optimization_level": 0,
+  "orthogonal_init": false,
+  "peft_type": "MARS",
+  "quant_n_bits": 8,
+  "r": 32,
+  "revision": null,
+  "seed": 42,
+  "shared_r": 32,
+  "target_modules": [
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "gate_proj",
+    "up_proj",
+    "v_proj",
+    "k_proj"
+  ],
+  "task_type": null,
+  "use_bnb": true
+}