Konark Verma commited on
Add AdaLoRA training report: qwen3-0.6b_B_1k [20260827_013954]
Browse files
reports/v2.0/adalora/reports_training/qwen3-0.6b_B_1k_20260827_013954.json
ADDED
|
@@ -0,0 +1,683 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_key": "qwen3-0.6b",
|
| 3 |
+
"model_id": "Qwen/Qwen3-0.6B",
|
| 4 |
+
"adalora_config": "B",
|
| 5 |
+
"adalora_config_desc": "Standard adaptive — full attention, init 24 → target 8",
|
| 6 |
+
"init_r": 24,
|
| 7 |
+
"target_r": 8,
|
| 8 |
+
"dataset_size": "1k",
|
| 9 |
+
"technique": "AdaLoRA",
|
| 10 |
+
"device": "cuda",
|
| 11 |
+
"dtype": "bfloat16",
|
| 12 |
+
"timestamp": "20260827_013954",
|
| 13 |
+
"prompt_format": "tool_id_v1",
|
| 14 |
+
"tool_id_scheme": "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ",
|
| 15 |
+
"no_tool_id": "-",
|
| 16 |
+
"trainable_params_init": 6883968,
|
| 17 |
+
"total_params": 602934000,
|
| 18 |
+
"trainable_pct_init": 1.1417,
|
| 19 |
+
"steps_per_epoch": 50,
|
| 20 |
+
"total_steps_trained": 500,
|
| 21 |
+
"peak_memory_mb": 14259.4,
|
| 22 |
+
"total_training_time_s": 1372.4,
|
| 23 |
+
"final_train_loss": 0.312586,
|
| 24 |
+
"final_eval_loss": 0.210603,
|
| 25 |
+
"final_val_accuracy": 0.94,
|
| 26 |
+
"final_train_accuracy": 1.0,
|
| 27 |
+
"hf_repo": "kon172verma/intent-classifier-experiments",
|
| 28 |
+
"hf_subfolder": "v2.0/qwen3-0.6b_AdaLoRA_B_1k_20260827-013947",
|
| 29 |
+
"hf_report_path": "reports/v2.0/adalora/reports_training/qwen3-0.6b_B_1k_20260827_013954.json",
|
| 30 |
+
"version": "v2.0",
|
| 31 |
+
"log_history": [
|
| 32 |
+
{
|
| 33 |
+
"loss": 2.747205352783203,
|
| 34 |
+
"grad_norm": 1.6238194704055786,
|
| 35 |
+
"learning_rate": 7.2e-05,
|
| 36 |
+
"entropy": 0.34799390435218813,
|
| 37 |
+
"num_tokens": 59712.0,
|
| 38 |
+
"mean_token_accuracy": 0.7624999940395355,
|
| 39 |
+
"epoch": 0.2,
|
| 40 |
+
"step": 10
|
| 41 |
+
},
|
| 42 |
+
{
|
| 43 |
+
"loss": 2.6690114974975585,
|
| 44 |
+
"grad_norm": 4.532184600830078,
|
| 45 |
+
"learning_rate": 0.000152,
|
| 46 |
+
"entropy": 0.34436134472489355,
|
| 47 |
+
"num_tokens": 117246.0,
|
| 48 |
+
"mean_token_accuracy": 0.7604166686534881,
|
| 49 |
+
"epoch": 0.4,
|
| 50 |
+
"step": 20
|
| 51 |
+
},
|
| 52 |
+
{
|
| 53 |
+
"loss": 2.4476388931274413,
|
| 54 |
+
"grad_norm": 6.346834182739258,
|
| 55 |
+
"learning_rate": 0.00019996500732179695,
|
| 56 |
+
"entropy": 0.333583864569664,
|
| 57 |
+
"num_tokens": 179597.0,
|
| 58 |
+
"mean_token_accuracy": 0.775,
|
| 59 |
+
"epoch": 0.6,
|
| 60 |
+
"step": 30
|
| 61 |
+
},
|
| 62 |
+
{
|
| 63 |
+
"loss": 1.3019063949584961,
|
| 64 |
+
"grad_norm": 2.3714170455932617,
|
| 65 |
+
"learning_rate": 0.0001995716208873644,
|
| 66 |
+
"entropy": 0.31575099341571333,
|
| 67 |
+
"num_tokens": 241259.0,
|
| 68 |
+
"mean_token_accuracy": 0.8437500059604645,
|
| 69 |
+
"epoch": 0.8,
|
| 70 |
+
"step": 40
|
| 71 |
+
},
|
| 72 |
+
{
|
| 73 |
+
"loss": 0.8112195014953614,
|
| 74 |
+
"grad_norm": 0.8492180705070496,
|
| 75 |
+
"learning_rate": 0.00019874283308955057,
|
| 76 |
+
"entropy": 0.22197748329490424,
|
| 77 |
+
"num_tokens": 304665.0,
|
| 78 |
+
"mean_token_accuracy": 0.9062500029802323,
|
| 79 |
+
"epoch": 1.0,
|
| 80 |
+
"step": 50
|
| 81 |
+
},
|
| 82 |
+
{
|
| 83 |
+
"eval_loss": 0.5988062024116516,
|
| 84 |
+
"eval_model_preparation_time": 0.0136,
|
| 85 |
+
"eval_runtime": 4.4761,
|
| 86 |
+
"eval_samples_per_second": 22.341,
|
| 87 |
+
"eval_steps_per_second": 2.904,
|
| 88 |
+
"eval_entropy": 0.18227221702153867,
|
| 89 |
+
"eval_num_tokens": 304665.0,
|
| 90 |
+
"eval_mean_token_accuracy": 0.9391025534042945,
|
| 91 |
+
"epoch": 1.0,
|
| 92 |
+
"step": 50,
|
| 93 |
+
"train_accuracy": 0.84,
|
| 94 |
+
"eval_accuracy": 0.79
|
| 95 |
+
},
|
| 96 |
+
{
|
| 97 |
+
"loss": 0.5958127021789551,
|
| 98 |
+
"grad_norm": 0.8858842253684998,
|
| 99 |
+
"learning_rate": 0.0001974822680065206,
|
| 100 |
+
"entropy": 0.16555267684161662,
|
| 101 |
+
"num_tokens": 365289.0,
|
| 102 |
+
"mean_token_accuracy": 0.9395833283662796,
|
| 103 |
+
"epoch": 1.2,
|
| 104 |
+
"step": 60
|
| 105 |
+
},
|
| 106 |
+
{
|
| 107 |
+
"loss": 0.4516918182373047,
|
| 108 |
+
"grad_norm": 0.9685459136962891,
|
| 109 |
+
"learning_rate": 0.0001957954377686475,
|
| 110 |
+
"entropy": 0.10552733233198523,
|
| 111 |
+
"num_tokens": 425915.0,
|
| 112 |
+
"mean_token_accuracy": 0.9562499940395355,
|
| 113 |
+
"epoch": 1.4,
|
| 114 |
+
"step": 70
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"loss": 0.4386417865753174,
|
| 118 |
+
"grad_norm": 0.4635484218597412,
|
| 119 |
+
"learning_rate": 0.00019368971845536845,
|
| 120 |
+
"entropy": 0.1300073081627488,
|
| 121 |
+
"num_tokens": 487364.0,
|
| 122 |
+
"mean_token_accuracy": 0.9562499970197678,
|
| 123 |
+
"epoch": 1.6,
|
| 124 |
+
"step": 80
|
| 125 |
+
},
|
| 126 |
+
{
|
| 127 |
+
"loss": 0.33032095432281494,
|
| 128 |
+
"grad_norm": 0.6365861892700195,
|
| 129 |
+
"learning_rate": 0.00019117431784146645,
|
| 130 |
+
"entropy": 0.11662073051556945,
|
| 131 |
+
"num_tokens": 548197.0,
|
| 132 |
+
"mean_token_accuracy": 0.9708333224058151,
|
| 133 |
+
"epoch": 1.8,
|
| 134 |
+
"step": 90
|
| 135 |
+
},
|
| 136 |
+
{
|
| 137 |
+
"loss": 0.34021382331848143,
|
| 138 |
+
"grad_norm": 0.35966211557388306,
|
| 139 |
+
"learning_rate": 0.0001882602351338137,
|
| 140 |
+
"entropy": 0.12453801995143295,
|
| 141 |
+
"num_tokens": 609330.0,
|
| 142 |
+
"mean_token_accuracy": 0.9625,
|
| 143 |
+
"epoch": 2.0,
|
| 144 |
+
"step": 100
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"eval_loss": 0.29777461290359497,
|
| 148 |
+
"eval_model_preparation_time": 0.0136,
|
| 149 |
+
"eval_runtime": 4.4727,
|
| 150 |
+
"eval_samples_per_second": 22.358,
|
| 151 |
+
"eval_steps_per_second": 2.906,
|
| 152 |
+
"eval_entropy": 0.09827234381093429,
|
| 153 |
+
"eval_num_tokens": 609330.0,
|
| 154 |
+
"eval_mean_token_accuracy": 0.9679487118354211,
|
| 155 |
+
"epoch": 2.0,
|
| 156 |
+
"step": 100,
|
| 157 |
+
"train_accuracy": 0.92,
|
| 158 |
+
"eval_accuracy": 0.89
|
| 159 |
+
},
|
| 160 |
+
{
|
| 161 |
+
"loss": 0.25004522800445556,
|
| 162 |
+
"grad_norm": 0.41243985295295715,
|
| 163 |
+
"learning_rate": 0.0001849602128746387,
|
| 164 |
+
"entropy": 0.09664164837449789,
|
| 165 |
+
"num_tokens": 670414.0,
|
| 166 |
+
"mean_token_accuracy": 0.9770833283662796,
|
| 167 |
+
"epoch": 2.2,
|
| 168 |
+
"step": 110
|
| 169 |
+
},
|
| 170 |
+
{
|
| 171 |
+
"loss": 0.22724947929382325,
|
| 172 |
+
"grad_norm": 0.42656299471855164,
|
| 173 |
+
"learning_rate": 0.00018128868122163123,
|
| 174 |
+
"entropy": 0.09195249238982797,
|
| 175 |
+
"num_tokens": 733398.0,
|
| 176 |
+
"mean_token_accuracy": 0.9833333313465118,
|
| 177 |
+
"epoch": 2.4,
|
| 178 |
+
"step": 120
|
| 179 |
+
},
|
| 180 |
+
{
|
| 181 |
+
"loss": 0.2152564287185669,
|
| 182 |
+
"grad_norm": 0.8918176889419556,
|
| 183 |
+
"learning_rate": 0.00017726169484853438,
|
| 184 |
+
"entropy": 0.07094540973193944,
|
| 185 |
+
"num_tokens": 792131.0,
|
| 186 |
+
"mean_token_accuracy": 0.9833333253860473,
|
| 187 |
+
"epoch": 2.6,
|
| 188 |
+
"step": 130
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"loss": 0.1605777382850647,
|
| 192 |
+
"grad_norm": 0.1348947435617447,
|
| 193 |
+
"learning_rate": 0.00017289686274214118,
|
| 194 |
+
"entropy": 0.05082558988360688,
|
| 195 |
+
"num_tokens": 853792.0,
|
| 196 |
+
"mean_token_accuracy": 0.9916666656732559,
|
| 197 |
+
"epoch": 2.8,
|
| 198 |
+
"step": 140
|
| 199 |
+
},
|
| 200 |
+
{
|
| 201 |
+
"loss": 0.18155455589294434,
|
| 202 |
+
"grad_norm": 0.45422282814979553,
|
| 203 |
+
"learning_rate": 0.00016821327120267567,
|
| 204 |
+
"entropy": 0.04635021737776697,
|
| 205 |
+
"num_tokens": 913995.0,
|
| 206 |
+
"mean_token_accuracy": 0.9812499910593033,
|
| 207 |
+
"epoch": 3.0,
|
| 208 |
+
"step": 150
|
| 209 |
+
},
|
| 210 |
+
{
|
| 211 |
+
"eval_loss": 0.21707607805728912,
|
| 212 |
+
"eval_model_preparation_time": 0.0136,
|
| 213 |
+
"eval_runtime": 4.4873,
|
| 214 |
+
"eval_samples_per_second": 22.285,
|
| 215 |
+
"eval_steps_per_second": 2.897,
|
| 216 |
+
"eval_entropy": 0.04368861481690636,
|
| 217 |
+
"eval_num_tokens": 913995.0,
|
| 218 |
+
"eval_mean_token_accuracy": 0.977564091865833,
|
| 219 |
+
"epoch": 3.0,
|
| 220 |
+
"step": 150,
|
| 221 |
+
"train_accuracy": 0.96,
|
| 222 |
+
"eval_accuracy": 0.93
|
| 223 |
+
},
|
| 224 |
+
{
|
| 225 |
+
"loss": 0.1182679533958435,
|
| 226 |
+
"grad_norm": 0.036157894879579544,
|
| 227 |
+
"learning_rate": 0.00016323140038425842,
|
| 228 |
+
"entropy": 0.02517663492471911,
|
| 229 |
+
"num_tokens": 970156.0,
|
| 230 |
+
"mean_token_accuracy": 0.9937499970197677,
|
| 231 |
+
"epoch": 3.2,
|
| 232 |
+
"step": 160
|
| 233 |
+
},
|
| 234 |
+
{
|
| 235 |
+
"loss": 0.1523075819015503,
|
| 236 |
+
"grad_norm": 1.0210810899734497,
|
| 237 |
+
"learning_rate": 0.00015797303474040332,
|
| 238 |
+
"entropy": 0.056804691744036975,
|
| 239 |
+
"num_tokens": 1035564.0,
|
| 240 |
+
"mean_token_accuracy": 0.9854166626930236,
|
| 241 |
+
"epoch": 3.4,
|
| 242 |
+
"step": 170
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"loss": 0.12568256855010987,
|
| 246 |
+
"grad_norm": 0.06671973317861557,
|
| 247 |
+
"learning_rate": 0.00015246116776614538,
|
| 248 |
+
"entropy": 0.047318724950309844,
|
| 249 |
+
"num_tokens": 1100370.0,
|
| 250 |
+
"mean_token_accuracy": 0.9895833313465119,
|
| 251 |
+
"epoch": 3.6,
|
| 252 |
+
"step": 180
|
| 253 |
+
},
|
| 254 |
+
{
|
| 255 |
+
"loss": 0.10956447124481201,
|
| 256 |
+
"grad_norm": 0.1402207762002945,
|
| 257 |
+
"learning_rate": 0.00014671990145333696,
|
| 258 |
+
"entropy": 0.03290305999107659,
|
| 259 |
+
"num_tokens": 1159423.0,
|
| 260 |
+
"mean_token_accuracy": 0.9937499970197677,
|
| 261 |
+
"epoch": 3.8,
|
| 262 |
+
"step": 190
|
| 263 |
+
},
|
| 264 |
+
{
|
| 265 |
+
"loss": 0.10298393964767456,
|
| 266 |
+
"grad_norm": 0.03720708563923836,
|
| 267 |
+
"learning_rate": 0.00014077434089877037,
|
| 268 |
+
"entropy": 0.025664863898418842,
|
| 269 |
+
"num_tokens": 1218660.0,
|
| 270 |
+
"mean_token_accuracy": 0.9916666626930237,
|
| 271 |
+
"epoch": 4.0,
|
| 272 |
+
"step": 200
|
| 273 |
+
},
|
| 274 |
+
{
|
| 275 |
+
"eval_loss": 0.20146024227142334,
|
| 276 |
+
"eval_model_preparation_time": 0.0136,
|
| 277 |
+
"eval_runtime": 4.5009,
|
| 278 |
+
"eval_samples_per_second": 22.218,
|
| 279 |
+
"eval_steps_per_second": 2.888,
|
| 280 |
+
"eval_entropy": 0.029143200632041462,
|
| 281 |
+
"eval_num_tokens": 1218660.0,
|
| 282 |
+
"eval_mean_token_accuracy": 0.9807692215992854,
|
| 283 |
+
"epoch": 4.0,
|
| 284 |
+
"step": 200,
|
| 285 |
+
"train_accuracy": 1.0,
|
| 286 |
+
"eval_accuracy": 0.94
|
| 287 |
+
},
|
| 288 |
+
{
|
| 289 |
+
"loss": 0.10776989459991455,
|
| 290 |
+
"grad_norm": 0.4595985412597656,
|
| 291 |
+
"learning_rate": 0.00013465048452597682,
|
| 292 |
+
"entropy": 0.03145701786270365,
|
| 293 |
+
"num_tokens": 1283139.0,
|
| 294 |
+
"mean_token_accuracy": 0.9916666626930237,
|
| 295 |
+
"epoch": 4.2,
|
| 296 |
+
"step": 210
|
| 297 |
+
},
|
| 298 |
+
{
|
| 299 |
+
"loss": 0.08369647264480591,
|
| 300 |
+
"grad_norm": 0.030895451083779335,
|
| 301 |
+
"learning_rate": 0.0001283751104007355,
|
| 302 |
+
"entropy": 0.02227218730549794,
|
| 303 |
+
"num_tokens": 1343512.0,
|
| 304 |
+
"mean_token_accuracy": 0.9958333313465119,
|
| 305 |
+
"epoch": 4.4,
|
| 306 |
+
"step": 220
|
| 307 |
+
},
|
| 308 |
+
{
|
| 309 |
+
"loss": 0.09575333595275878,
|
| 310 |
+
"grad_norm": 0.09805360436439514,
|
| 311 |
+
"learning_rate": 0.00012197565913740531,
|
| 312 |
+
"entropy": 0.026834382093511522,
|
| 313 |
+
"num_tokens": 1403279.0,
|
| 314 |
+
"mean_token_accuracy": 0.9958333313465119,
|
| 315 |
+
"epoch": 4.6,
|
| 316 |
+
"step": 230
|
| 317 |
+
},
|
| 318 |
+
{
|
| 319 |
+
"loss": 0.07492621541023255,
|
| 320 |
+
"grad_norm": 0.03002256713807583,
|
| 321 |
+
"learning_rate": 0.00011548011390810017,
|
| 322 |
+
"entropy": 0.02416663138428703,
|
| 323 |
+
"num_tokens": 1463858.0,
|
| 324 |
+
"mean_token_accuracy": 0.9958333313465119,
|
| 325 |
+
"epoch": 4.8,
|
| 326 |
+
"step": 240
|
| 327 |
+
},
|
| 328 |
+
{
|
| 329 |
+
"loss": 0.07101120948791503,
|
| 330 |
+
"grad_norm": 0.03387472406029701,
|
| 331 |
+
"learning_rate": 0.00010891687807939707,
|
| 332 |
+
"entropy": 0.023247016553068534,
|
| 333 |
+
"num_tokens": 1523325.0,
|
| 334 |
+
"mean_token_accuracy": 0.9979166656732559,
|
| 335 |
+
"epoch": 5.0,
|
| 336 |
+
"step": 250
|
| 337 |
+
},
|
| 338 |
+
{
|
| 339 |
+
"eval_loss": 0.20226804912090302,
|
| 340 |
+
"eval_model_preparation_time": 0.0136,
|
| 341 |
+
"eval_runtime": 4.4774,
|
| 342 |
+
"eval_samples_per_second": 22.334,
|
| 343 |
+
"eval_steps_per_second": 2.903,
|
| 344 |
+
"eval_entropy": 0.023026770491224643,
|
| 345 |
+
"eval_num_tokens": 1523325.0,
|
| 346 |
+
"eval_mean_token_accuracy": 0.9807692215992854,
|
| 347 |
+
"epoch": 5.0,
|
| 348 |
+
"step": 250,
|
| 349 |
+
"train_accuracy": 1.0,
|
| 350 |
+
"eval_accuracy": 0.94
|
| 351 |
+
},
|
| 352 |
+
{
|
| 353 |
+
"loss": 0.07109203338623046,
|
| 354 |
+
"grad_norm": 0.371147096157074,
|
| 355 |
+
"learning_rate": 0.00010231465101164139,
|
| 356 |
+
"entropy": 0.018356483816751278,
|
| 357 |
+
"num_tokens": 1584286.0,
|
| 358 |
+
"mean_token_accuracy": 0.9958333313465119,
|
| 359 |
+
"epoch": 5.2,
|
| 360 |
+
"step": 260
|
| 361 |
+
},
|
| 362 |
+
{
|
| 363 |
+
"loss": 0.07887292504310608,
|
| 364 |
+
"grad_norm": 0.02899470180273056,
|
| 365 |
+
"learning_rate": 9.570230256394596e-05,
|
| 366 |
+
"entropy": 0.023915311659220607,
|
| 367 |
+
"num_tokens": 1645197.0,
|
| 368 |
+
"mean_token_accuracy": 0.9937499970197677,
|
| 369 |
+
"epoch": 5.4,
|
| 370 |
+
"step": 270
|
| 371 |
+
},
|
| 372 |
+
{
|
| 373 |
+
"loss": 0.06103082299232483,
|
| 374 |
+
"grad_norm": 0.03286702558398247,
|
| 375 |
+
"learning_rate": 8.910874685364275e-05,
|
| 376 |
+
"entropy": 0.014954480109736323,
|
| 377 |
+
"num_tokens": 1703569.0,
|
| 378 |
+
"mean_token_accuracy": 1.0,
|
| 379 |
+
"epoch": 5.6,
|
| 380 |
+
"step": 280
|
| 381 |
+
},
|
| 382 |
+
{
|
| 383 |
+
"loss": 0.08200910687446594,
|
| 384 |
+
"grad_norm": 0.1984379142522812,
|
| 385 |
+
"learning_rate": 8.256281582220485e-05,
|
| 386 |
+
"entropy": 0.019160138722509146,
|
| 387 |
+
"num_tokens": 1765363.0,
|
| 388 |
+
"mean_token_accuracy": 0.9937499970197677,
|
| 389 |
+
"epoch": 5.8,
|
| 390 |
+
"step": 290
|
| 391 |
+
},
|
| 392 |
+
{
|
| 393 |
+
"loss": 0.06011219024658203,
|
| 394 |
+
"grad_norm": 0.029375849291682243,
|
| 395 |
+
"learning_rate": 7.6093133160502e-05,
|
| 396 |
+
"entropy": 0.012518353835912422,
|
| 397 |
+
"num_tokens": 1827990.0,
|
| 398 |
+
"mean_token_accuracy": 0.9979166656732559,
|
| 399 |
+
"epoch": 6.0,
|
| 400 |
+
"step": 300
|
| 401 |
+
},
|
| 402 |
+
{
|
| 403 |
+
"eval_loss": 0.20414863526821136,
|
| 404 |
+
"eval_model_preparation_time": 0.0136,
|
| 405 |
+
"eval_runtime": 4.48,
|
| 406 |
+
"eval_samples_per_second": 22.322,
|
| 407 |
+
"eval_steps_per_second": 2.902,
|
| 408 |
+
"eval_entropy": 0.017313140560872853,
|
| 409 |
+
"eval_num_tokens": 1827990.0,
|
| 410 |
+
"eval_mean_token_accuracy": 0.9807692215992854,
|
| 411 |
+
"epoch": 6.0,
|
| 412 |
+
"step": 300,
|
| 413 |
+
"train_accuracy": 1.0,
|
| 414 |
+
"eval_accuracy": 0.93
|
| 415 |
+
},
|
| 416 |
+
{
|
| 417 |
+
"loss": 0.06169512271881104,
|
| 418 |
+
"grad_norm": 0.8248687386512756,
|
| 419 |
+
"learning_rate": 6.972798914468369e-05,
|
| 420 |
+
"entropy": 0.01321726841852069,
|
| 421 |
+
"num_tokens": 1890301.0,
|
| 422 |
+
"mean_token_accuracy": 0.9979166656732559,
|
| 423 |
+
"epoch": 6.2,
|
| 424 |
+
"step": 310
|
| 425 |
+
},
|
| 426 |
+
{
|
| 427 |
+
"loss": 0.06761946082115174,
|
| 428 |
+
"grad_norm": 0.03681720420718193,
|
| 429 |
+
"learning_rate": 6.349521692999945e-05,
|
| 430 |
+
"entropy": 0.02435837863013148,
|
| 431 |
+
"num_tokens": 1951131.0,
|
| 432 |
+
"mean_token_accuracy": 0.9958333313465119,
|
| 433 |
+
"epoch": 6.4,
|
| 434 |
+
"step": 320
|
| 435 |
+
},
|
| 436 |
+
{
|
| 437 |
+
"loss": 0.054594749212265016,
|
| 438 |
+
"grad_norm": 0.09282302856445312,
|
| 439 |
+
"learning_rate": 5.7422070843492734e-05,
|
| 440 |
+
"entropy": 0.009648823738098144,
|
| 441 |
+
"num_tokens": 2011404.0,
|
| 442 |
+
"mean_token_accuracy": 1.0,
|
| 443 |
+
"epoch": 6.6,
|
| 444 |
+
"step": 330
|
| 445 |
+
},
|
| 446 |
+
{
|
| 447 |
+
"loss": 0.05391166806221008,
|
| 448 |
+
"grad_norm": 0.028602655977010727,
|
| 449 |
+
"learning_rate": 5.1535107207763534e-05,
|
| 450 |
+
"entropy": 0.010603643802460283,
|
| 451 |
+
"num_tokens": 2071811.0,
|
| 452 |
+
"mean_token_accuracy": 1.0,
|
| 453 |
+
"epoch": 6.8,
|
| 454 |
+
"step": 340
|
| 455 |
+
},
|
| 456 |
+
{
|
| 457 |
+
"loss": 0.05777776837348938,
|
| 458 |
+
"grad_norm": 0.16235725581645966,
|
| 459 |
+
"learning_rate": 4.58600682169262e-05,
|
| 460 |
+
"entropy": 0.012113475316436962,
|
| 461 |
+
"num_tokens": 2132655.0,
|
| 462 |
+
"mean_token_accuracy": 0.9958333313465119,
|
| 463 |
+
"epoch": 7.0,
|
| 464 |
+
"step": 350
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"eval_loss": 0.20904219150543213,
|
| 468 |
+
"eval_model_preparation_time": 0.0136,
|
| 469 |
+
"eval_runtime": 4.4965,
|
| 470 |
+
"eval_samples_per_second": 22.239,
|
| 471 |
+
"eval_steps_per_second": 2.891,
|
| 472 |
+
"eval_entropy": 0.013050362100609792,
|
| 473 |
+
"eval_num_tokens": 2132655.0,
|
| 474 |
+
"eval_mean_token_accuracy": 0.9807692215992854,
|
| 475 |
+
"epoch": 7.0,
|
| 476 |
+
"step": 350,
|
| 477 |
+
"train_accuracy": 1.0,
|
| 478 |
+
"eval_accuracy": 0.93
|
| 479 |
+
},
|
| 480 |
+
{
|
| 481 |
+
"loss": 0.05278337597846985,
|
| 482 |
+
"grad_norm": 0.432327538728714,
|
| 483 |
+
"learning_rate": 4.0421769372544736e-05,
|
| 484 |
+
"entropy": 0.010453603573841974,
|
| 485 |
+
"num_tokens": 2193232.0,
|
| 486 |
+
"mean_token_accuracy": 1.0,
|
| 487 |
+
"epoch": 7.2,
|
| 488 |
+
"step": 360
|
| 489 |
+
},
|
| 490 |
+
{
|
| 491 |
+
"loss": 0.05139507651329041,
|
| 492 |
+
"grad_norm": 0.143241748213768,
|
| 493 |
+
"learning_rate": 3.5243990971758125e-05,
|
| 494 |
+
"entropy": 0.00859435934398789,
|
| 495 |
+
"num_tokens": 2254308.0,
|
| 496 |
+
"mean_token_accuracy": 1.0,
|
| 497 |
+
"epoch": 7.4,
|
| 498 |
+
"step": 370
|
| 499 |
+
},
|
| 500 |
+
{
|
| 501 |
+
"loss": 0.05181134939193725,
|
| 502 |
+
"grad_norm": 0.3068942427635193,
|
| 503 |
+
"learning_rate": 3.034937412209178e-05,
|
| 504 |
+
"entropy": 0.010343420712160878,
|
| 505 |
+
"num_tokens": 2314575.0,
|
| 506 |
+
"mean_token_accuracy": 1.0,
|
| 507 |
+
"epoch": 7.6,
|
| 508 |
+
"step": 380
|
| 509 |
+
},
|
| 510 |
+
{
|
| 511 |
+
"loss": 0.05138177275657654,
|
| 512 |
+
"grad_norm": 0.026515545323491096,
|
| 513 |
+
"learning_rate": 2.5759321737655017e-05,
|
| 514 |
+
"entropy": 0.007268380204914137,
|
| 515 |
+
"num_tokens": 2375510.0,
|
| 516 |
+
"mean_token_accuracy": 0.9979166656732559,
|
| 517 |
+
"epoch": 7.8,
|
| 518 |
+
"step": 390
|
| 519 |
+
},
|
| 520 |
+
{
|
| 521 |
+
"loss": 0.05907024741172791,
|
| 522 |
+
"grad_norm": 0.6199167966842651,
|
| 523 |
+
"learning_rate": 2.149390494964323e-05,
|
| 524 |
+
"entropy": 0.017255466303322465,
|
| 525 |
+
"num_tokens": 2437320.0,
|
| 526 |
+
"mean_token_accuracy": 0.9958333313465119,
|
| 527 |
+
"epoch": 8.0,
|
| 528 |
+
"step": 400
|
| 529 |
+
},
|
| 530 |
+
{
|
| 531 |
+
"eval_loss": 0.20484547317028046,
|
| 532 |
+
"eval_model_preparation_time": 0.0136,
|
| 533 |
+
"eval_runtime": 4.489,
|
| 534 |
+
"eval_samples_per_second": 22.277,
|
| 535 |
+
"eval_steps_per_second": 2.896,
|
| 536 |
+
"eval_entropy": 0.01104860689688044,
|
| 537 |
+
"eval_num_tokens": 2437320.0,
|
| 538 |
+
"eval_mean_token_accuracy": 0.977564091865833,
|
| 539 |
+
"epoch": 8.0,
|
| 540 |
+
"step": 400,
|
| 541 |
+
"train_accuracy": 1.0,
|
| 542 |
+
"eval_accuracy": 0.94
|
| 543 |
+
},
|
| 544 |
+
{
|
| 545 |
+
"loss": 0.051655936241149905,
|
| 546 |
+
"grad_norm": 0.20560742914676666,
|
| 547 |
+
"learning_rate": 1.7571775340388276e-05,
|
| 548 |
+
"entropy": 0.01528367143473588,
|
| 549 |
+
"num_tokens": 2496004.0,
|
| 550 |
+
"mean_token_accuracy": 1.0,
|
| 551 |
+
"epoch": 8.2,
|
| 552 |
+
"step": 410
|
| 553 |
+
},
|
| 554 |
+
{
|
| 555 |
+
"loss": 0.048995989561080935,
|
| 556 |
+
"grad_norm": 0.3655153810977936,
|
| 557 |
+
"learning_rate": 1.4010083384734308e-05,
|
| 558 |
+
"entropy": 0.005615859891986474,
|
| 559 |
+
"num_tokens": 2556337.0,
|
| 560 |
+
"mean_token_accuracy": 1.0,
|
| 561 |
+
"epoch": 8.4,
|
| 562 |
+
"step": 420
|
| 563 |
+
},
|
| 564 |
+
{
|
| 565 |
+
"loss": 0.055114632844924925,
|
| 566 |
+
"grad_norm": 0.026448752731084824,
|
| 567 |
+
"learning_rate": 1.0824403455375288e-05,
|
| 568 |
+
"entropy": 0.012873932387446985,
|
| 569 |
+
"num_tokens": 2620160.0,
|
| 570 |
+
"mean_token_accuracy": 0.9979166656732559,
|
| 571 |
+
"epoch": 8.6,
|
| 572 |
+
"step": 430
|
| 573 |
+
},
|
| 574 |
+
{
|
| 575 |
+
"loss": 0.048224377632141116,
|
| 576 |
+
"grad_norm": 0.038070619106292725,
|
| 577 |
+
"learning_rate": 8.02866572008566e-06,
|
| 578 |
+
"entropy": 0.005977723849355243,
|
| 579 |
+
"num_tokens": 2681883.0,
|
| 580 |
+
"mean_token_accuracy": 1.0,
|
| 581 |
+
"epoch": 8.8,
|
| 582 |
+
"step": 440
|
| 583 |
+
},
|
| 584 |
+
{
|
| 585 |
+
"loss": 0.04953903555870056,
|
| 586 |
+
"grad_norm": 0.06778038293123245,
|
| 587 |
+
"learning_rate": 5.63509522864123e-06,
|
| 588 |
+
"entropy": 0.007351557198853697,
|
| 589 |
+
"num_tokens": 2741985.0,
|
| 590 |
+
"mean_token_accuracy": 1.0,
|
| 591 |
+
"epoch": 9.0,
|
| 592 |
+
"step": 450
|
| 593 |
+
},
|
| 594 |
+
{
|
| 595 |
+
"eval_loss": 0.20688867568969727,
|
| 596 |
+
"eval_model_preparation_time": 0.0136,
|
| 597 |
+
"eval_runtime": 4.4812,
|
| 598 |
+
"eval_samples_per_second": 22.315,
|
| 599 |
+
"eval_steps_per_second": 2.901,
|
| 600 |
+
"eval_entropy": 0.009611496970487328,
|
| 601 |
+
"eval_num_tokens": 2741985.0,
|
| 602 |
+
"eval_mean_token_accuracy": 0.9807692215992854,
|
| 603 |
+
"epoch": 9.0,
|
| 604 |
+
"step": 450,
|
| 605 |
+
"train_accuracy": 1.0,
|
| 606 |
+
"eval_accuracy": 0.94
|
| 607 |
+
},
|
| 608 |
+
{
|
| 609 |
+
"loss": 0.050711172819137576,
|
| 610 |
+
"grad_norm": 0.028333287686109543,
|
| 611 |
+
"learning_rate": 3.6541584557868604e-06,
|
| 612 |
+
"entropy": 0.010634195803140756,
|
| 613 |
+
"num_tokens": 2801726.0,
|
| 614 |
+
"mean_token_accuracy": 0.9979166656732559,
|
| 615 |
+
"epoch": 9.2,
|
| 616 |
+
"step": 460
|
| 617 |
+
},
|
| 618 |
+
{
|
| 619 |
+
"loss": 0.05352324843406677,
|
| 620 |
+
"grad_norm": 0.03653257340192795,
|
| 621 |
+
"learning_rate": 2.0945175340055357e-06,
|
| 622 |
+
"entropy": 0.01226643608242739,
|
| 623 |
+
"num_tokens": 2862743.0,
|
| 624 |
+
"mean_token_accuracy": 0.9979166656732559,
|
| 625 |
+
"epoch": 9.4,
|
| 626 |
+
"step": 470
|
| 627 |
+
},
|
| 628 |
+
{
|
| 629 |
+
"loss": 0.04824453592300415,
|
| 630 |
+
"grad_norm": 0.030703146010637283,
|
| 631 |
+
"learning_rate": 9.62992376217009e-07,
|
| 632 |
+
"entropy": 0.005805992777459324,
|
| 633 |
+
"num_tokens": 2923294.0,
|
| 634 |
+
"mean_token_accuracy": 1.0,
|
| 635 |
+
"epoch": 9.6,
|
| 636 |
+
"step": 480
|
| 637 |
+
},
|
| 638 |
+
{
|
| 639 |
+
"loss": 0.04900467991828918,
|
| 640 |
+
"grad_norm": 0.026623884215950966,
|
| 641 |
+
"learning_rate": 2.645308540337843e-07,
|
| 642 |
+
"entropy": 0.009043255326105282,
|
| 643 |
+
"num_tokens": 2984888.0,
|
| 644 |
+
"mean_token_accuracy": 1.0,
|
| 645 |
+
"epoch": 9.8,
|
| 646 |
+
"step": 490
|
| 647 |
+
},
|
| 648 |
+
{
|
| 649 |
+
"loss": 0.048820114135742186,
|
| 650 |
+
"grad_norm": 0.04335764795541763,
|
| 651 |
+
"learning_rate": 2.187161977540431e-09,
|
| 652 |
+
"entropy": 0.008386473459540866,
|
| 653 |
+
"num_tokens": 3046650.0,
|
| 654 |
+
"mean_token_accuracy": 1.0,
|
| 655 |
+
"epoch": 10.0,
|
| 656 |
+
"step": 500
|
| 657 |
+
},
|
| 658 |
+
{
|
| 659 |
+
"eval_loss": 0.21060311794281006,
|
| 660 |
+
"eval_model_preparation_time": 0.0136,
|
| 661 |
+
"eval_runtime": 4.486,
|
| 662 |
+
"eval_samples_per_second": 22.292,
|
| 663 |
+
"eval_steps_per_second": 2.898,
|
| 664 |
+
"eval_entropy": 0.010091413219925016,
|
| 665 |
+
"eval_num_tokens": 3046650.0,
|
| 666 |
+
"eval_mean_token_accuracy": 0.9807692215992854,
|
| 667 |
+
"epoch": 10.0,
|
| 668 |
+
"step": 500,
|
| 669 |
+
"train_accuracy": 1.0,
|
| 670 |
+
"eval_accuracy": 0.94
|
| 671 |
+
},
|
| 672 |
+
{
|
| 673 |
+
"train_runtime": 1372.013,
|
| 674 |
+
"train_samples_per_second": 5.831,
|
| 675 |
+
"train_steps_per_second": 0.364,
|
| 676 |
+
"total_flos": 1.3501885325681664e+16,
|
| 677 |
+
"train_loss": 0.31258602380752565,
|
| 678 |
+
"epoch": 10.0,
|
| 679 |
+
"step": 500
|
| 680 |
+
}
|
| 681 |
+
],
|
| 682 |
+
"lora_config": "B"
|
| 683 |
+
}
|