add training_config.json
Browse files- training_config.json +195 -0
training_config.json
ADDED
|
@@ -0,0 +1,195 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"mode": "fixed",
|
| 3 |
+
"seed": 0,
|
| 4 |
+
"hf_repo": "Likithp/v10_fixed_s0",
|
| 5 |
+
"base_model": "Qwen/Qwen2.5-0.5B",
|
| 6 |
+
"dataset": "data/cs7_fixed_v3",
|
| 7 |
+
"dataset_version": "cs7_v3",
|
| 8 |
+
"trained_at": "2026-06-05T19:05:24.286663+00:00",
|
| 9 |
+
"optimizer": "AdamW",
|
| 10 |
+
"lr": 3e-05,
|
| 11 |
+
"weight_decay": 0.0,
|
| 12 |
+
"batch_size": 8,
|
| 13 |
+
"grad_accum": 8,
|
| 14 |
+
"effective_batch": 64,
|
| 15 |
+
"epochs": 5,
|
| 16 |
+
"warmup_steps": 200,
|
| 17 |
+
"grad_clip": 0.5,
|
| 18 |
+
"lr_schedule": "cosine",
|
| 19 |
+
"checkpoint_criterion": "val_em",
|
| 20 |
+
"max_seq_len": 256,
|
| 21 |
+
"dtype": "torch.bfloat16",
|
| 22 |
+
"eval_method": "teacher_forcing_argmax",
|
| 23 |
+
"alias_groups": {
|
| 24 |
+
"T1": [
|
| 25 |
+
"act",
|
| 26 |
+
"cst",
|
| 27 |
+
"emp",
|
| 28 |
+
"inv",
|
| 29 |
+
"ord",
|
| 30 |
+
"spl",
|
| 31 |
+
"txn"
|
| 32 |
+
],
|
| 33 |
+
"T2": [
|
| 34 |
+
"brc",
|
| 35 |
+
"ctg",
|
| 36 |
+
"dpt",
|
| 37 |
+
"empl",
|
| 38 |
+
"ordr",
|
| 39 |
+
"prd",
|
| 40 |
+
"prj",
|
| 41 |
+
"rgn",
|
| 42 |
+
"shp",
|
| 43 |
+
"tsk",
|
| 44 |
+
"whs"
|
| 45 |
+
]
|
| 46 |
+
},
|
| 47 |
+
"log_every_steps": 50,
|
| 48 |
+
"eval_n": 500,
|
| 49 |
+
"dry_run": false,
|
| 50 |
+
"best_epoch": 5,
|
| 51 |
+
"best_train_loss": null,
|
| 52 |
+
"val_exact_match": 1.0,
|
| 53 |
+
"val_outer_alias_acc": 1.0,
|
| 54 |
+
"val_inner_alias_acc": 1.0,
|
| 55 |
+
"val_t1_inner_alias_acc": 1.0,
|
| 56 |
+
"val_t2_inner_alias_acc": 1.0,
|
| 57 |
+
"val_t1_t2_gap_pp": 0.0,
|
| 58 |
+
"val_inner_by_alias": {
|
| 59 |
+
"act": {
|
| 60 |
+
"correct": 64,
|
| 61 |
+
"total": 64,
|
| 62 |
+
"pct": 100.0,
|
| 63 |
+
"token_group": "T1"
|
| 64 |
+
},
|
| 65 |
+
"brc": {
|
| 66 |
+
"correct": 50,
|
| 67 |
+
"total": 50,
|
| 68 |
+
"pct": 100.0,
|
| 69 |
+
"token_group": "T2"
|
| 70 |
+
},
|
| 71 |
+
"cst": {
|
| 72 |
+
"correct": 45,
|
| 73 |
+
"total": 45,
|
| 74 |
+
"pct": 100.0,
|
| 75 |
+
"token_group": "T1"
|
| 76 |
+
},
|
| 77 |
+
"ctg": {
|
| 78 |
+
"correct": 61,
|
| 79 |
+
"total": 61,
|
| 80 |
+
"pct": 100.0,
|
| 81 |
+
"token_group": "T2"
|
| 82 |
+
},
|
| 83 |
+
"dpt": {
|
| 84 |
+
"correct": 56,
|
| 85 |
+
"total": 56,
|
| 86 |
+
"pct": 100.0,
|
| 87 |
+
"token_group": "T2"
|
| 88 |
+
},
|
| 89 |
+
"emp": {
|
| 90 |
+
"correct": 51,
|
| 91 |
+
"total": 51,
|
| 92 |
+
"pct": 100.0,
|
| 93 |
+
"token_group": "T1"
|
| 94 |
+
},
|
| 95 |
+
"empl": {
|
| 96 |
+
"correct": 59,
|
| 97 |
+
"total": 59,
|
| 98 |
+
"pct": 100.0,
|
| 99 |
+
"token_group": "T2"
|
| 100 |
+
},
|
| 101 |
+
"inv": {
|
| 102 |
+
"correct": 45,
|
| 103 |
+
"total": 45,
|
| 104 |
+
"pct": 100.0,
|
| 105 |
+
"token_group": "T1"
|
| 106 |
+
},
|
| 107 |
+
"ord": {
|
| 108 |
+
"correct": 67,
|
| 109 |
+
"total": 67,
|
| 110 |
+
"pct": 100.0,
|
| 111 |
+
"token_group": "T1"
|
| 112 |
+
},
|
| 113 |
+
"ordr": {
|
| 114 |
+
"correct": 59,
|
| 115 |
+
"total": 59,
|
| 116 |
+
"pct": 100.0,
|
| 117 |
+
"token_group": "T2"
|
| 118 |
+
},
|
| 119 |
+
"prd": {
|
| 120 |
+
"correct": 57,
|
| 121 |
+
"total": 57,
|
| 122 |
+
"pct": 100.0,
|
| 123 |
+
"token_group": "T2"
|
| 124 |
+
},
|
| 125 |
+
"prj": {
|
| 126 |
+
"correct": 52,
|
| 127 |
+
"total": 52,
|
| 128 |
+
"pct": 100.0,
|
| 129 |
+
"token_group": "T2"
|
| 130 |
+
},
|
| 131 |
+
"rgn": {
|
| 132 |
+
"correct": 70,
|
| 133 |
+
"total": 70,
|
| 134 |
+
"pct": 100.0,
|
| 135 |
+
"token_group": "T2"
|
| 136 |
+
},
|
| 137 |
+
"shp": {
|
| 138 |
+
"correct": 60,
|
| 139 |
+
"total": 60,
|
| 140 |
+
"pct": 100.0,
|
| 141 |
+
"token_group": "T2"
|
| 142 |
+
},
|
| 143 |
+
"spl": {
|
| 144 |
+
"correct": 59,
|
| 145 |
+
"total": 59,
|
| 146 |
+
"pct": 100.0,
|
| 147 |
+
"token_group": "T1"
|
| 148 |
+
},
|
| 149 |
+
"tsk": {
|
| 150 |
+
"correct": 44,
|
| 151 |
+
"total": 44,
|
| 152 |
+
"pct": 100.0,
|
| 153 |
+
"token_group": "T2"
|
| 154 |
+
},
|
| 155 |
+
"txn": {
|
| 156 |
+
"correct": 51,
|
| 157 |
+
"total": 51,
|
| 158 |
+
"pct": 100.0,
|
| 159 |
+
"token_group": "T1"
|
| 160 |
+
},
|
| 161 |
+
"whs": {
|
| 162 |
+
"correct": 50,
|
| 163 |
+
"total": 50,
|
| 164 |
+
"pct": 100.0,
|
| 165 |
+
"token_group": "T2"
|
| 166 |
+
}
|
| 167 |
+
},
|
| 168 |
+
"train_log": [
|
| 169 |
+
{
|
| 170 |
+
"epoch": 1,
|
| 171 |
+
"loss": 0.026001,
|
| 172 |
+
"val_em": 1.0
|
| 173 |
+
},
|
| 174 |
+
{
|
| 175 |
+
"epoch": 2,
|
| 176 |
+
"loss": 1.1e-05,
|
| 177 |
+
"val_em": 1.0
|
| 178 |
+
},
|
| 179 |
+
{
|
| 180 |
+
"epoch": 3,
|
| 181 |
+
"loss": 1.1e-05,
|
| 182 |
+
"val_em": 1.0
|
| 183 |
+
},
|
| 184 |
+
{
|
| 185 |
+
"epoch": 4,
|
| 186 |
+
"loss": 1e-05,
|
| 187 |
+
"val_em": 1.0
|
| 188 |
+
},
|
| 189 |
+
{
|
| 190 |
+
"epoch": 5,
|
| 191 |
+
"loss": 1e-05,
|
| 192 |
+
"val_em": 1.0
|
| 193 |
+
}
|
| 194 |
+
]
|
| 195 |
+
}
|