Upload folder using huggingface_hub
Browse files- p5-mae-15m-d131/exports/step_00040000/fsd50k_lora_1ep_s0.json +284 -0
- p5-mae-15m-d131/exports/step_00040000/fsd50k_lora_1ep_s1.json +284 -0
- p5-mae-15m-d131/exports/step_00040000/fsd50k_lora_1ep_s2.json +284 -0
- p5-mae-15m-d131/exports/step_00040000/lora_1ep_s0.json +260 -0
- p5-mae-15m-d131/exports/step_00040000/lora_1ep_s1.json +260 -0
- p5-mae-15m-d131/exports/step_00040000/lora_1ep_s2.json +260 -0
p5-mae-15m-d131/exports/step_00040000/fsd50k_lora_1ep_s0.json
ADDED
|
@@ -0,0 +1,284 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"export": "/workspace/runs/p5-mae-15m-d131/exports/step_00040000",
|
| 3 |
+
"fingerprint": "f1b23242ec8f5601",
|
| 4 |
+
"campaign_id": "phase5",
|
| 5 |
+
"trial_id": "p5-mae-15m-d131",
|
| 6 |
+
"target": "fsd50k",
|
| 7 |
+
"num_classes": 200,
|
| 8 |
+
"lora_layers_wrapped": 32,
|
| 9 |
+
"params": {
|
| 10 |
+
"trainable": 876488,
|
| 11 |
+
"total": 15158984,
|
| 12 |
+
"fraction": 0.05781970612278501
|
| 13 |
+
},
|
| 14 |
+
"steps_per_epoch": 989,
|
| 15 |
+
"total_steps": 989,
|
| 16 |
+
"train_clips": 36796,
|
| 17 |
+
"train_segments": 47487,
|
| 18 |
+
"val_clips": 10231,
|
| 19 |
+
"val_segments": 15513,
|
| 20 |
+
"train_split": "train",
|
| 21 |
+
"eval_split": "test",
|
| 22 |
+
"config": {
|
| 23 |
+
"rank": 16,
|
| 24 |
+
"alpha": 32.0,
|
| 25 |
+
"targets": [
|
| 26 |
+
"qkv",
|
| 27 |
+
"proj",
|
| 28 |
+
"fc1",
|
| 29 |
+
"fc2"
|
| 30 |
+
],
|
| 31 |
+
"train_norms": true,
|
| 32 |
+
"epochs": 1.0,
|
| 33 |
+
"batch_size": 48,
|
| 34 |
+
"lr": 0.003,
|
| 35 |
+
"min_lr": 1e-05,
|
| 36 |
+
"warmup_frac": 0.1,
|
| 37 |
+
"betas": [
|
| 38 |
+
0.9,
|
| 39 |
+
0.95
|
| 40 |
+
],
|
| 41 |
+
"weight_decay": 0.05,
|
| 42 |
+
"clip_norm": 4.0,
|
| 43 |
+
"mixup": 0.0,
|
| 44 |
+
"mixup_prob": 0.9,
|
| 45 |
+
"freq_mask": 0,
|
| 46 |
+
"time_mask": 0,
|
| 47 |
+
"roll_aug": false,
|
| 48 |
+
"mask_ratio": 0.0,
|
| 49 |
+
"num_classes": 200,
|
| 50 |
+
"prior_bias_init": true,
|
| 51 |
+
"num_workers": 8,
|
| 52 |
+
"seed": 0,
|
| 53 |
+
"train_probe_clips": 4096,
|
| 54 |
+
"train_eval_every": 25,
|
| 55 |
+
"val_evals": 5,
|
| 56 |
+
"eval_batch_size": 128
|
| 57 |
+
},
|
| 58 |
+
"gpu": "NVIDIA GeForce RTX 4090",
|
| 59 |
+
"final_val_map": 0.3861398249492049,
|
| 60 |
+
"lora/map": 0.3861398249492049,
|
| 61 |
+
"fsd50k/map": 0.3861398249492049,
|
| 62 |
+
"train_seconds": 23.93,
|
| 63 |
+
"total_seconds": 32.93,
|
| 64 |
+
"history": [
|
| 65 |
+
{
|
| 66 |
+
"step": 0,
|
| 67 |
+
"epoch": 0.0,
|
| 68 |
+
"loss": null,
|
| 69 |
+
"val_map": 0.022475250469287857,
|
| 70 |
+
"lr": 0.0,
|
| 71 |
+
"train_s": 0.0
|
| 72 |
+
},
|
| 73 |
+
{
|
| 74 |
+
"step": 989,
|
| 75 |
+
"epoch": 1.0,
|
| 76 |
+
"loss": 0.023323401123058668,
|
| 77 |
+
"val_map": 0.3861398249492049,
|
| 78 |
+
"lr": 0.0,
|
| 79 |
+
"train_s": 23.92951434443239
|
| 80 |
+
}
|
| 81 |
+
],
|
| 82 |
+
"per_class_ap": [
|
| 83 |
+
0.25680994987487793,
|
| 84 |
+
0.0971757099032402,
|
| 85 |
+
0.5362106561660767,
|
| 86 |
+
0.3658115267753601,
|
| 87 |
+
0.44647452235221863,
|
| 88 |
+
0.6417517066001892,
|
| 89 |
+
0.8556394577026367,
|
| 90 |
+
0.5427014827728271,
|
| 91 |
+
0.5823351144790649,
|
| 92 |
+
0.4800824820995331,
|
| 93 |
+
0.2308141589164734,
|
| 94 |
+
0.37687188386917114,
|
| 95 |
+
0.20867614448070526,
|
| 96 |
+
0.3311745226383209,
|
| 97 |
+
0.5144516229629517,
|
| 98 |
+
0.4741208553314209,
|
| 99 |
+
0.03795246779918671,
|
| 100 |
+
0.24348898231983185,
|
| 101 |
+
0.39076682925224304,
|
| 102 |
+
0.25247520208358765,
|
| 103 |
+
0.5005587339401245,
|
| 104 |
+
0.40964391827583313,
|
| 105 |
+
0.843904435634613,
|
| 106 |
+
0.08464940637350082,
|
| 107 |
+
0.23559901118278503,
|
| 108 |
+
0.45880788564682007,
|
| 109 |
+
0.36421483755111694,
|
| 110 |
+
0.3015795350074768,
|
| 111 |
+
0.7792471051216125,
|
| 112 |
+
0.33254653215408325,
|
| 113 |
+
0.675590455532074,
|
| 114 |
+
0.5671966671943665,
|
| 115 |
+
0.40456876158714294,
|
| 116 |
+
0.21114613115787506,
|
| 117 |
+
0.29556918144226074,
|
| 118 |
+
0.1927097886800766,
|
| 119 |
+
0.4181755483150482,
|
| 120 |
+
0.1843390166759491,
|
| 121 |
+
0.2905430197715759,
|
| 122 |
+
0.2378394603729248,
|
| 123 |
+
0.5548097491264343,
|
| 124 |
+
0.26775556802749634,
|
| 125 |
+
0.593946099281311,
|
| 126 |
+
0.179182767868042,
|
| 127 |
+
0.46465232968330383,
|
| 128 |
+
0.1051713153719902,
|
| 129 |
+
0.22874987125396729,
|
| 130 |
+
0.2803884744644165,
|
| 131 |
+
0.7194206118583679,
|
| 132 |
+
0.42561736702919006,
|
| 133 |
+
0.3977198004722595,
|
| 134 |
+
0.5646023154258728,
|
| 135 |
+
0.5775550007820129,
|
| 136 |
+
0.15762433409690857,
|
| 137 |
+
0.15458200871944427,
|
| 138 |
+
0.025624345988035202,
|
| 139 |
+
0.258685827255249,
|
| 140 |
+
0.6808355450630188,
|
| 141 |
+
0.13041117787361145,
|
| 142 |
+
0.5131982564926147,
|
| 143 |
+
0.6042957305908203,
|
| 144 |
+
0.6262543201446533,
|
| 145 |
+
0.428829550743103,
|
| 146 |
+
0.1685415655374527,
|
| 147 |
+
0.49451425671577454,
|
| 148 |
+
0.1548914909362793,
|
| 149 |
+
0.2791929841041565,
|
| 150 |
+
0.6667209267616272,
|
| 151 |
+
0.5285017490386963,
|
| 152 |
+
0.556400716304779,
|
| 153 |
+
0.47169405221939087,
|
| 154 |
+
0.20595340430736542,
|
| 155 |
+
0.6765598058700562,
|
| 156 |
+
0.3558964729309082,
|
| 157 |
+
0.3038884103298187,
|
| 158 |
+
0.23192954063415527,
|
| 159 |
+
0.45080792903900146,
|
| 160 |
+
0.3988146483898163,
|
| 161 |
+
0.30952373147010803,
|
| 162 |
+
0.43542686104774475,
|
| 163 |
+
0.2441040277481079,
|
| 164 |
+
0.33495813608169556,
|
| 165 |
+
0.28437814116477966,
|
| 166 |
+
0.36643004417419434,
|
| 167 |
+
0.17093537747859955,
|
| 168 |
+
0.34717854857444763,
|
| 169 |
+
0.40842437744140625,
|
| 170 |
+
0.6450591683387756,
|
| 171 |
+
0.2003302276134491,
|
| 172 |
+
0.11607600003480911,
|
| 173 |
+
0.7350584268569946,
|
| 174 |
+
0.18227629363536835,
|
| 175 |
+
0.5240997672080994,
|
| 176 |
+
0.26156994700431824,
|
| 177 |
+
0.19356487691402435,
|
| 178 |
+
0.3144744634628296,
|
| 179 |
+
0.3741360902786255,
|
| 180 |
+
0.12342321872711182,
|
| 181 |
+
0.37950053811073303,
|
| 182 |
+
0.07869795709848404,
|
| 183 |
+
0.5722806453704834,
|
| 184 |
+
0.7118571996688843,
|
| 185 |
+
0.22066664695739746,
|
| 186 |
+
0.40885046124458313,
|
| 187 |
+
0.24804867804050446,
|
| 188 |
+
0.4083211123943329,
|
| 189 |
+
0.6324588656425476,
|
| 190 |
+
0.5247567296028137,
|
| 191 |
+
0.5555833578109741,
|
| 192 |
+
0.23104174435138702,
|
| 193 |
+
0.20087459683418274,
|
| 194 |
+
0.3294066786766052,
|
| 195 |
+
0.6395897269248962,
|
| 196 |
+
0.06852158159017563,
|
| 197 |
+
0.15888890624046326,
|
| 198 |
+
0.33468568325042725,
|
| 199 |
+
0.7610511779785156,
|
| 200 |
+
0.3734697997570038,
|
| 201 |
+
0.4830629527568817,
|
| 202 |
+
0.21645960211753845,
|
| 203 |
+
0.8664469718933105,
|
| 204 |
+
0.8751571774482727,
|
| 205 |
+
0.3659977316856384,
|
| 206 |
+
0.1444617509841919,
|
| 207 |
+
0.2023295760154724,
|
| 208 |
+
0.7678764462471008,
|
| 209 |
+
0.2653846740722656,
|
| 210 |
+
0.7235323190689087,
|
| 211 |
+
0.3336145579814911,
|
| 212 |
+
0.1560317426919937,
|
| 213 |
+
0.18912406265735626,
|
| 214 |
+
0.7838781476020813,
|
| 215 |
+
0.4739767014980316,
|
| 216 |
+
0.44287699460983276,
|
| 217 |
+
0.6644648313522339,
|
| 218 |
+
0.09214100241661072,
|
| 219 |
+
0.10296466946601868,
|
| 220 |
+
0.05554381385445595,
|
| 221 |
+
0.49781203269958496,
|
| 222 |
+
0.5511995553970337,
|
| 223 |
+
0.21895147860050201,
|
| 224 |
+
0.3198224604129791,
|
| 225 |
+
0.23268595337867737,
|
| 226 |
+
0.31809550523757935,
|
| 227 |
+
0.700894832611084,
|
| 228 |
+
0.582547128200531,
|
| 229 |
+
0.05046294257044792,
|
| 230 |
+
0.5764703154563904,
|
| 231 |
+
0.2232765108346939,
|
| 232 |
+
0.2781132459640503,
|
| 233 |
+
0.4950554072856903,
|
| 234 |
+
0.5020754337310791,
|
| 235 |
+
0.2411850541830063,
|
| 236 |
+
0.3825940489768982,
|
| 237 |
+
0.4553874433040619,
|
| 238 |
+
0.30101123452186584,
|
| 239 |
+
0.4890395998954773,
|
| 240 |
+
0.6620763540267944,
|
| 241 |
+
0.4749201536178589,
|
| 242 |
+
0.09006647020578384,
|
| 243 |
+
0.4413588345050812,
|
| 244 |
+
0.1366984248161316,
|
| 245 |
+
0.6273430585861206,
|
| 246 |
+
0.2543559968471527,
|
| 247 |
+
0.3683604598045349,
|
| 248 |
+
0.28665289282798767,
|
| 249 |
+
0.3957931101322174,
|
| 250 |
+
0.06397455185651779,
|
| 251 |
+
0.37491247057914734,
|
| 252 |
+
0.4218943417072296,
|
| 253 |
+
0.09157717227935791,
|
| 254 |
+
0.8752515316009521,
|
| 255 |
+
0.8830752968788147,
|
| 256 |
+
0.06605520844459534,
|
| 257 |
+
0.5509445667266846,
|
| 258 |
+
0.7815967798233032,
|
| 259 |
+
0.19196942448616028,
|
| 260 |
+
0.2595559060573578,
|
| 261 |
+
0.207984060049057,
|
| 262 |
+
0.3126911222934723,
|
| 263 |
+
0.04826131835579872,
|
| 264 |
+
0.4662254750728607,
|
| 265 |
+
0.26997798681259155,
|
| 266 |
+
0.5731738805770874,
|
| 267 |
+
0.6281453967094421,
|
| 268 |
+
0.40777406096458435,
|
| 269 |
+
0.2834843695163727,
|
| 270 |
+
0.6742706298828125,
|
| 271 |
+
0.5727631449699402,
|
| 272 |
+
0.4162925183773041,
|
| 273 |
+
0.40914252400398254,
|
| 274 |
+
0.4553481936454773,
|
| 275 |
+
0.5688613057136536,
|
| 276 |
+
0.1454440951347351,
|
| 277 |
+
0.16976605355739594,
|
| 278 |
+
0.2992176115512848,
|
| 279 |
+
0.0875486359000206,
|
| 280 |
+
0.4684307873249054,
|
| 281 |
+
0.14272472262382507,
|
| 282 |
+
0.7515833377838135
|
| 283 |
+
]
|
| 284 |
+
}
|
p5-mae-15m-d131/exports/step_00040000/fsd50k_lora_1ep_s1.json
ADDED
|
@@ -0,0 +1,284 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"export": "/workspace/runs/p5-mae-15m-d131/exports/step_00040000",
|
| 3 |
+
"fingerprint": "f1b23242ec8f5601",
|
| 4 |
+
"campaign_id": "phase5",
|
| 5 |
+
"trial_id": "p5-mae-15m-d131",
|
| 6 |
+
"target": "fsd50k",
|
| 7 |
+
"num_classes": 200,
|
| 8 |
+
"lora_layers_wrapped": 32,
|
| 9 |
+
"params": {
|
| 10 |
+
"trainable": 876488,
|
| 11 |
+
"total": 15158984,
|
| 12 |
+
"fraction": 0.05781970612278501
|
| 13 |
+
},
|
| 14 |
+
"steps_per_epoch": 989,
|
| 15 |
+
"total_steps": 989,
|
| 16 |
+
"train_clips": 36796,
|
| 17 |
+
"train_segments": 47487,
|
| 18 |
+
"val_clips": 10231,
|
| 19 |
+
"val_segments": 15513,
|
| 20 |
+
"train_split": "train",
|
| 21 |
+
"eval_split": "test",
|
| 22 |
+
"config": {
|
| 23 |
+
"rank": 16,
|
| 24 |
+
"alpha": 32.0,
|
| 25 |
+
"targets": [
|
| 26 |
+
"qkv",
|
| 27 |
+
"proj",
|
| 28 |
+
"fc1",
|
| 29 |
+
"fc2"
|
| 30 |
+
],
|
| 31 |
+
"train_norms": true,
|
| 32 |
+
"epochs": 1.0,
|
| 33 |
+
"batch_size": 48,
|
| 34 |
+
"lr": 0.003,
|
| 35 |
+
"min_lr": 1e-05,
|
| 36 |
+
"warmup_frac": 0.1,
|
| 37 |
+
"betas": [
|
| 38 |
+
0.9,
|
| 39 |
+
0.95
|
| 40 |
+
],
|
| 41 |
+
"weight_decay": 0.05,
|
| 42 |
+
"clip_norm": 4.0,
|
| 43 |
+
"mixup": 0.0,
|
| 44 |
+
"mixup_prob": 0.9,
|
| 45 |
+
"freq_mask": 0,
|
| 46 |
+
"time_mask": 0,
|
| 47 |
+
"roll_aug": false,
|
| 48 |
+
"mask_ratio": 0.0,
|
| 49 |
+
"num_classes": 200,
|
| 50 |
+
"prior_bias_init": true,
|
| 51 |
+
"num_workers": 8,
|
| 52 |
+
"seed": 1,
|
| 53 |
+
"train_probe_clips": 4096,
|
| 54 |
+
"train_eval_every": 25,
|
| 55 |
+
"val_evals": 5,
|
| 56 |
+
"eval_batch_size": 128
|
| 57 |
+
},
|
| 58 |
+
"gpu": "NVIDIA GeForce RTX 4090",
|
| 59 |
+
"final_val_map": 0.3894634534791112,
|
| 60 |
+
"lora/map": 0.3894634534791112,
|
| 61 |
+
"fsd50k/map": 0.3894634534791112,
|
| 62 |
+
"train_seconds": 24.52,
|
| 63 |
+
"total_seconds": 33.72,
|
| 64 |
+
"history": [
|
| 65 |
+
{
|
| 66 |
+
"step": 0,
|
| 67 |
+
"epoch": 0.0,
|
| 68 |
+
"loss": null,
|
| 69 |
+
"val_map": 0.02233186560915783,
|
| 70 |
+
"lr": 0.0,
|
| 71 |
+
"train_s": 0.0
|
| 72 |
+
},
|
| 73 |
+
{
|
| 74 |
+
"step": 989,
|
| 75 |
+
"epoch": 1.0,
|
| 76 |
+
"loss": 0.023306202863165697,
|
| 77 |
+
"val_map": 0.3894634534791112,
|
| 78 |
+
"lr": 0.0,
|
| 79 |
+
"train_s": 24.52063795097638
|
| 80 |
+
}
|
| 81 |
+
],
|
| 82 |
+
"per_class_ap": [
|
| 83 |
+
0.25124335289001465,
|
| 84 |
+
0.08645329624414444,
|
| 85 |
+
0.5218135714530945,
|
| 86 |
+
0.33888813853263855,
|
| 87 |
+
0.45825478434562683,
|
| 88 |
+
0.6484904289245605,
|
| 89 |
+
0.8570923805236816,
|
| 90 |
+
0.5760295391082764,
|
| 91 |
+
0.5548572540283203,
|
| 92 |
+
0.5174098014831543,
|
| 93 |
+
0.3032286465167999,
|
| 94 |
+
0.37309861183166504,
|
| 95 |
+
0.18065328896045685,
|
| 96 |
+
0.40888404846191406,
|
| 97 |
+
0.5086308717727661,
|
| 98 |
+
0.47925153374671936,
|
| 99 |
+
0.03148803859949112,
|
| 100 |
+
0.22111372649669647,
|
| 101 |
+
0.42936453223228455,
|
| 102 |
+
0.21164962649345398,
|
| 103 |
+
0.5610662698745728,
|
| 104 |
+
0.41700980067253113,
|
| 105 |
+
0.8682270050048828,
|
| 106 |
+
0.09543923288583755,
|
| 107 |
+
0.22518271207809448,
|
| 108 |
+
0.3615342974662781,
|
| 109 |
+
0.35994207859039307,
|
| 110 |
+
0.2850973308086395,
|
| 111 |
+
0.7826710343360901,
|
| 112 |
+
0.3211223781108856,
|
| 113 |
+
0.6485413908958435,
|
| 114 |
+
0.5725330710411072,
|
| 115 |
+
0.4448225796222687,
|
| 116 |
+
0.21007999777793884,
|
| 117 |
+
0.30484744906425476,
|
| 118 |
+
0.24326151609420776,
|
| 119 |
+
0.4368332624435425,
|
| 120 |
+
0.15384992957115173,
|
| 121 |
+
0.3691961169242859,
|
| 122 |
+
0.27461960911750793,
|
| 123 |
+
0.5652273297309875,
|
| 124 |
+
0.24650897085666656,
|
| 125 |
+
0.5872750282287598,
|
| 126 |
+
0.20665410161018372,
|
| 127 |
+
0.5168331861495972,
|
| 128 |
+
0.1884612739086151,
|
| 129 |
+
0.2277853637933731,
|
| 130 |
+
0.2785894572734833,
|
| 131 |
+
0.6667753458023071,
|
| 132 |
+
0.42470577359199524,
|
| 133 |
+
0.36088332533836365,
|
| 134 |
+
0.5308070778846741,
|
| 135 |
+
0.5982879996299744,
|
| 136 |
+
0.22174839675426483,
|
| 137 |
+
0.15696300566196442,
|
| 138 |
+
0.026723049581050873,
|
| 139 |
+
0.24646000564098358,
|
| 140 |
+
0.6731954216957092,
|
| 141 |
+
0.16795702278614044,
|
| 142 |
+
0.5477712154388428,
|
| 143 |
+
0.6309290528297424,
|
| 144 |
+
0.6191468238830566,
|
| 145 |
+
0.41883385181427,
|
| 146 |
+
0.21575164794921875,
|
| 147 |
+
0.44841673970222473,
|
| 148 |
+
0.1122739315032959,
|
| 149 |
+
0.2732701301574707,
|
| 150 |
+
0.6605531573295593,
|
| 151 |
+
0.48651444911956787,
|
| 152 |
+
0.5469843149185181,
|
| 153 |
+
0.46922075748443604,
|
| 154 |
+
0.21774804592132568,
|
| 155 |
+
0.6726953387260437,
|
| 156 |
+
0.29251646995544434,
|
| 157 |
+
0.21061842143535614,
|
| 158 |
+
0.2524806261062622,
|
| 159 |
+
0.4613712430000305,
|
| 160 |
+
0.45288118720054626,
|
| 161 |
+
0.2925284504890442,
|
| 162 |
+
0.42180702090263367,
|
| 163 |
+
0.24759428203105927,
|
| 164 |
+
0.2710499167442322,
|
| 165 |
+
0.26965636014938354,
|
| 166 |
+
0.24934248626232147,
|
| 167 |
+
0.300160676240921,
|
| 168 |
+
0.33580127358436584,
|
| 169 |
+
0.40274277329444885,
|
| 170 |
+
0.6238117814064026,
|
| 171 |
+
0.1949479579925537,
|
| 172 |
+
0.141066312789917,
|
| 173 |
+
0.7468685507774353,
|
| 174 |
+
0.18368427455425262,
|
| 175 |
+
0.5220368504524231,
|
| 176 |
+
0.2300780713558197,
|
| 177 |
+
0.29494765400886536,
|
| 178 |
+
0.3271646797657013,
|
| 179 |
+
0.4796645939350128,
|
| 180 |
+
0.15092073380947113,
|
| 181 |
+
0.361266165971756,
|
| 182 |
+
0.07721953094005585,
|
| 183 |
+
0.5835121870040894,
|
| 184 |
+
0.7144067287445068,
|
| 185 |
+
0.23142492771148682,
|
| 186 |
+
0.41977280378341675,
|
| 187 |
+
0.2395070195198059,
|
| 188 |
+
0.40643757581710815,
|
| 189 |
+
0.6507799625396729,
|
| 190 |
+
0.5272164940834045,
|
| 191 |
+
0.5516348481178284,
|
| 192 |
+
0.23040689527988434,
|
| 193 |
+
0.24978643655776978,
|
| 194 |
+
0.3420216143131256,
|
| 195 |
+
0.6595427989959717,
|
| 196 |
+
0.14628127217292786,
|
| 197 |
+
0.1640399992465973,
|
| 198 |
+
0.30657655000686646,
|
| 199 |
+
0.737471342086792,
|
| 200 |
+
0.3572810888290405,
|
| 201 |
+
0.4862827956676483,
|
| 202 |
+
0.2718505263328552,
|
| 203 |
+
0.8673800826072693,
|
| 204 |
+
0.8760181665420532,
|
| 205 |
+
0.3511139750480652,
|
| 206 |
+
0.08020088821649551,
|
| 207 |
+
0.2604938745498657,
|
| 208 |
+
0.7805554866790771,
|
| 209 |
+
0.29627904295921326,
|
| 210 |
+
0.732421338558197,
|
| 211 |
+
0.33805879950523376,
|
| 212 |
+
0.11262942105531693,
|
| 213 |
+
0.17679576575756073,
|
| 214 |
+
0.7910831570625305,
|
| 215 |
+
0.46220043301582336,
|
| 216 |
+
0.4623101055622101,
|
| 217 |
+
0.5867577791213989,
|
| 218 |
+
0.10483336448669434,
|
| 219 |
+
0.08489292860031128,
|
| 220 |
+
0.052226435393095016,
|
| 221 |
+
0.5658934116363525,
|
| 222 |
+
0.5644920468330383,
|
| 223 |
+
0.24914605915546417,
|
| 224 |
+
0.3091452121734619,
|
| 225 |
+
0.1973947286605835,
|
| 226 |
+
0.24941010773181915,
|
| 227 |
+
0.7322738170623779,
|
| 228 |
+
0.5719164609909058,
|
| 229 |
+
0.06644441187381744,
|
| 230 |
+
0.5652766227722168,
|
| 231 |
+
0.19232939183712006,
|
| 232 |
+
0.2662885785102844,
|
| 233 |
+
0.47477275133132935,
|
| 234 |
+
0.5212418437004089,
|
| 235 |
+
0.31342822313308716,
|
| 236 |
+
0.3610386550426483,
|
| 237 |
+
0.45686882734298706,
|
| 238 |
+
0.347013384103775,
|
| 239 |
+
0.5021817684173584,
|
| 240 |
+
0.6943045258522034,
|
| 241 |
+
0.4723062217235565,
|
| 242 |
+
0.08302810788154602,
|
| 243 |
+
0.4034344255924225,
|
| 244 |
+
0.14741250872612,
|
| 245 |
+
0.662470817565918,
|
| 246 |
+
0.3009205162525177,
|
| 247 |
+
0.35011225938796997,
|
| 248 |
+
0.36533042788505554,
|
| 249 |
+
0.39091169834136963,
|
| 250 |
+
0.06013338640332222,
|
| 251 |
+
0.33804383873939514,
|
| 252 |
+
0.4603709280490875,
|
| 253 |
+
0.09653609991073608,
|
| 254 |
+
0.8574508428573608,
|
| 255 |
+
0.8702770471572876,
|
| 256 |
+
0.04180517420172691,
|
| 257 |
+
0.5206948518753052,
|
| 258 |
+
0.7349156737327576,
|
| 259 |
+
0.2370339184999466,
|
| 260 |
+
0.24688783288002014,
|
| 261 |
+
0.21704867482185364,
|
| 262 |
+
0.30657586455345154,
|
| 263 |
+
0.044147197157144547,
|
| 264 |
+
0.5488054752349854,
|
| 265 |
+
0.2124474048614502,
|
| 266 |
+
0.6194069981575012,
|
| 267 |
+
0.6320005655288696,
|
| 268 |
+
0.32934895157814026,
|
| 269 |
+
0.27156105637550354,
|
| 270 |
+
0.658352255821228,
|
| 271 |
+
0.5760706663131714,
|
| 272 |
+
0.36622902750968933,
|
| 273 |
+
0.45221924781799316,
|
| 274 |
+
0.4480142295360565,
|
| 275 |
+
0.5681304931640625,
|
| 276 |
+
0.13226424157619476,
|
| 277 |
+
0.2094758003950119,
|
| 278 |
+
0.3780907988548279,
|
| 279 |
+
0.1137862429022789,
|
| 280 |
+
0.472388356924057,
|
| 281 |
+
0.14697101712226868,
|
| 282 |
+
0.7559823393821716
|
| 283 |
+
]
|
| 284 |
+
}
|
p5-mae-15m-d131/exports/step_00040000/fsd50k_lora_1ep_s2.json
ADDED
|
@@ -0,0 +1,284 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"export": "/workspace/runs/p5-mae-15m-d131/exports/step_00040000",
|
| 3 |
+
"fingerprint": "f1b23242ec8f5601",
|
| 4 |
+
"campaign_id": "phase5",
|
| 5 |
+
"trial_id": "p5-mae-15m-d131",
|
| 6 |
+
"target": "fsd50k",
|
| 7 |
+
"num_classes": 200,
|
| 8 |
+
"lora_layers_wrapped": 32,
|
| 9 |
+
"params": {
|
| 10 |
+
"trainable": 876488,
|
| 11 |
+
"total": 15158984,
|
| 12 |
+
"fraction": 0.05781970612278501
|
| 13 |
+
},
|
| 14 |
+
"steps_per_epoch": 989,
|
| 15 |
+
"total_steps": 989,
|
| 16 |
+
"train_clips": 36796,
|
| 17 |
+
"train_segments": 47487,
|
| 18 |
+
"val_clips": 10231,
|
| 19 |
+
"val_segments": 15513,
|
| 20 |
+
"train_split": "train",
|
| 21 |
+
"eval_split": "test",
|
| 22 |
+
"config": {
|
| 23 |
+
"rank": 16,
|
| 24 |
+
"alpha": 32.0,
|
| 25 |
+
"targets": [
|
| 26 |
+
"qkv",
|
| 27 |
+
"proj",
|
| 28 |
+
"fc1",
|
| 29 |
+
"fc2"
|
| 30 |
+
],
|
| 31 |
+
"train_norms": true,
|
| 32 |
+
"epochs": 1.0,
|
| 33 |
+
"batch_size": 48,
|
| 34 |
+
"lr": 0.003,
|
| 35 |
+
"min_lr": 1e-05,
|
| 36 |
+
"warmup_frac": 0.1,
|
| 37 |
+
"betas": [
|
| 38 |
+
0.9,
|
| 39 |
+
0.95
|
| 40 |
+
],
|
| 41 |
+
"weight_decay": 0.05,
|
| 42 |
+
"clip_norm": 4.0,
|
| 43 |
+
"mixup": 0.0,
|
| 44 |
+
"mixup_prob": 0.9,
|
| 45 |
+
"freq_mask": 0,
|
| 46 |
+
"time_mask": 0,
|
| 47 |
+
"roll_aug": false,
|
| 48 |
+
"mask_ratio": 0.0,
|
| 49 |
+
"num_classes": 200,
|
| 50 |
+
"prior_bias_init": true,
|
| 51 |
+
"num_workers": 8,
|
| 52 |
+
"seed": 2,
|
| 53 |
+
"train_probe_clips": 4096,
|
| 54 |
+
"train_eval_every": 25,
|
| 55 |
+
"val_evals": 5,
|
| 56 |
+
"eval_batch_size": 128
|
| 57 |
+
},
|
| 58 |
+
"gpu": "NVIDIA GeForce RTX 4090",
|
| 59 |
+
"final_val_map": 0.38830385616980495,
|
| 60 |
+
"lora/map": 0.38830385616980495,
|
| 61 |
+
"fsd50k/map": 0.38830385616980495,
|
| 62 |
+
"train_seconds": 24.11,
|
| 63 |
+
"total_seconds": 33.07,
|
| 64 |
+
"history": [
|
| 65 |
+
{
|
| 66 |
+
"step": 0,
|
| 67 |
+
"epoch": 0.0,
|
| 68 |
+
"loss": null,
|
| 69 |
+
"val_map": 0.02168603423051536,
|
| 70 |
+
"lr": 0.0,
|
| 71 |
+
"train_s": 0.0
|
| 72 |
+
},
|
| 73 |
+
{
|
| 74 |
+
"step": 989,
|
| 75 |
+
"epoch": 1.0,
|
| 76 |
+
"loss": 0.023550587525131506,
|
| 77 |
+
"val_map": 0.38830385616980495,
|
| 78 |
+
"lr": 0.0,
|
| 79 |
+
"train_s": 24.108741654083133
|
| 80 |
+
}
|
| 81 |
+
],
|
| 82 |
+
"per_class_ap": [
|
| 83 |
+
0.25202739238739014,
|
| 84 |
+
0.17466796934604645,
|
| 85 |
+
0.550471305847168,
|
| 86 |
+
0.3467211425304413,
|
| 87 |
+
0.46111077070236206,
|
| 88 |
+
0.6292680501937866,
|
| 89 |
+
0.836385190486908,
|
| 90 |
+
0.5585620403289795,
|
| 91 |
+
0.6023959517478943,
|
| 92 |
+
0.4808495342731476,
|
| 93 |
+
0.30715906620025635,
|
| 94 |
+
0.38769200444221497,
|
| 95 |
+
0.17189836502075195,
|
| 96 |
+
0.44367069005966187,
|
| 97 |
+
0.5085980296134949,
|
| 98 |
+
0.4894905984401703,
|
| 99 |
+
0.04551725089550018,
|
| 100 |
+
0.2465062290430069,
|
| 101 |
+
0.4562309980392456,
|
| 102 |
+
0.23627068102359772,
|
| 103 |
+
0.5166566371917725,
|
| 104 |
+
0.3959124684333801,
|
| 105 |
+
0.8586187958717346,
|
| 106 |
+
0.1258496791124344,
|
| 107 |
+
0.23539744317531586,
|
| 108 |
+
0.4143998622894287,
|
| 109 |
+
0.38313716650009155,
|
| 110 |
+
0.3068198561668396,
|
| 111 |
+
0.7654201984405518,
|
| 112 |
+
0.32470962405204773,
|
| 113 |
+
0.6954589486122131,
|
| 114 |
+
0.5664967894554138,
|
| 115 |
+
0.4611605107784271,
|
| 116 |
+
0.16606390476226807,
|
| 117 |
+
0.3631546199321747,
|
| 118 |
+
0.22389447689056396,
|
| 119 |
+
0.42359355092048645,
|
| 120 |
+
0.20872466266155243,
|
| 121 |
+
0.38807213306427,
|
| 122 |
+
0.2503674328327179,
|
| 123 |
+
0.556196928024292,
|
| 124 |
+
0.17695221304893494,
|
| 125 |
+
0.6156777143478394,
|
| 126 |
+
0.17224587500095367,
|
| 127 |
+
0.47825887799263,
|
| 128 |
+
0.11054529994726181,
|
| 129 |
+
0.24135926365852356,
|
| 130 |
+
0.31213563680648804,
|
| 131 |
+
0.7658893465995789,
|
| 132 |
+
0.4143317937850952,
|
| 133 |
+
0.301190048456192,
|
| 134 |
+
0.5334627032279968,
|
| 135 |
+
0.5606200695037842,
|
| 136 |
+
0.11239024996757507,
|
| 137 |
+
0.21148650348186493,
|
| 138 |
+
0.028980528935790062,
|
| 139 |
+
0.21910794079303741,
|
| 140 |
+
0.6841407418251038,
|
| 141 |
+
0.12257856130599976,
|
| 142 |
+
0.5127536654472351,
|
| 143 |
+
0.5778084397315979,
|
| 144 |
+
0.6313066482543945,
|
| 145 |
+
0.4427897334098816,
|
| 146 |
+
0.13408151268959045,
|
| 147 |
+
0.49202585220336914,
|
| 148 |
+
0.12676146626472473,
|
| 149 |
+
0.224668949842453,
|
| 150 |
+
0.6716817617416382,
|
| 151 |
+
0.512208104133606,
|
| 152 |
+
0.5182145237922668,
|
| 153 |
+
0.4500483572483063,
|
| 154 |
+
0.2135809361934662,
|
| 155 |
+
0.6912334561347961,
|
| 156 |
+
0.2839055061340332,
|
| 157 |
+
0.27587825059890747,
|
| 158 |
+
0.22548051178455353,
|
| 159 |
+
0.4464087188243866,
|
| 160 |
+
0.4005950391292572,
|
| 161 |
+
0.27345168590545654,
|
| 162 |
+
0.44397807121276855,
|
| 163 |
+
0.22746531665325165,
|
| 164 |
+
0.33753225207328796,
|
| 165 |
+
0.31707438826560974,
|
| 166 |
+
0.3374231159687042,
|
| 167 |
+
0.3154838979244232,
|
| 168 |
+
0.25352752208709717,
|
| 169 |
+
0.40072962641716003,
|
| 170 |
+
0.602300763130188,
|
| 171 |
+
0.2361026406288147,
|
| 172 |
+
0.10750501602888107,
|
| 173 |
+
0.7333280444145203,
|
| 174 |
+
0.09270332753658295,
|
| 175 |
+
0.545792281627655,
|
| 176 |
+
0.2203679233789444,
|
| 177 |
+
0.3435760736465454,
|
| 178 |
+
0.3619891107082367,
|
| 179 |
+
0.49304014444351196,
|
| 180 |
+
0.13291648030281067,
|
| 181 |
+
0.37791475653648376,
|
| 182 |
+
0.08095495402812958,
|
| 183 |
+
0.572324275970459,
|
| 184 |
+
0.7063193321228027,
|
| 185 |
+
0.22110992670059204,
|
| 186 |
+
0.4068193733692169,
|
| 187 |
+
0.2578192353248596,
|
| 188 |
+
0.34777042269706726,
|
| 189 |
+
0.6539056301116943,
|
| 190 |
+
0.5259605050086975,
|
| 191 |
+
0.5407907366752625,
|
| 192 |
+
0.2726813852787018,
|
| 193 |
+
0.19224920868873596,
|
| 194 |
+
0.3352855145931244,
|
| 195 |
+
0.6166723966598511,
|
| 196 |
+
0.10281747579574585,
|
| 197 |
+
0.18129698932170868,
|
| 198 |
+
0.327802836894989,
|
| 199 |
+
0.7138245701789856,
|
| 200 |
+
0.4231870174407959,
|
| 201 |
+
0.5122500658035278,
|
| 202 |
+
0.2537550926208496,
|
| 203 |
+
0.870862603187561,
|
| 204 |
+
0.8798378705978394,
|
| 205 |
+
0.35843905806541443,
|
| 206 |
+
0.08868375420570374,
|
| 207 |
+
0.2407846450805664,
|
| 208 |
+
0.7749274969100952,
|
| 209 |
+
0.3230975866317749,
|
| 210 |
+
0.7219465970993042,
|
| 211 |
+
0.34434226155281067,
|
| 212 |
+
0.12430351972579956,
|
| 213 |
+
0.17348188161849976,
|
| 214 |
+
0.7568104267120361,
|
| 215 |
+
0.5163673162460327,
|
| 216 |
+
0.469889760017395,
|
| 217 |
+
0.637536346912384,
|
| 218 |
+
0.09205342829227448,
|
| 219 |
+
0.08725779503583908,
|
| 220 |
+
0.06164243072271347,
|
| 221 |
+
0.585742712020874,
|
| 222 |
+
0.543795108795166,
|
| 223 |
+
0.19628506898880005,
|
| 224 |
+
0.2852400243282318,
|
| 225 |
+
0.22878031432628632,
|
| 226 |
+
0.40922439098358154,
|
| 227 |
+
0.7573243975639343,
|
| 228 |
+
0.5462950468063354,
|
| 229 |
+
0.04603724926710129,
|
| 230 |
+
0.5612765550613403,
|
| 231 |
+
0.1706104278564453,
|
| 232 |
+
0.31587842106819153,
|
| 233 |
+
0.48777928948402405,
|
| 234 |
+
0.517867922782898,
|
| 235 |
+
0.3368295133113861,
|
| 236 |
+
0.27618375420570374,
|
| 237 |
+
0.4475966989994049,
|
| 238 |
+
0.3382795751094818,
|
| 239 |
+
0.5056167244911194,
|
| 240 |
+
0.6517770290374756,
|
| 241 |
+
0.4646458625793457,
|
| 242 |
+
0.07185138761997223,
|
| 243 |
+
0.39396822452545166,
|
| 244 |
+
0.1799331158399582,
|
| 245 |
+
0.6643195748329163,
|
| 246 |
+
0.19953563809394836,
|
| 247 |
+
0.30569928884506226,
|
| 248 |
+
0.3187559247016907,
|
| 249 |
+
0.30288389325141907,
|
| 250 |
+
0.06331872195005417,
|
| 251 |
+
0.35645297169685364,
|
| 252 |
+
0.44346386194229126,
|
| 253 |
+
0.10125245153903961,
|
| 254 |
+
0.8378357887268066,
|
| 255 |
+
0.8423147201538086,
|
| 256 |
+
0.03854668512940407,
|
| 257 |
+
0.583817183971405,
|
| 258 |
+
0.7742074131965637,
|
| 259 |
+
0.23101595044136047,
|
| 260 |
+
0.2448880821466446,
|
| 261 |
+
0.19767674803733826,
|
| 262 |
+
0.31661510467529297,
|
| 263 |
+
0.053762663155794144,
|
| 264 |
+
0.4668136239051819,
|
| 265 |
+
0.18756578862667084,
|
| 266 |
+
0.5936748385429382,
|
| 267 |
+
0.6236546039581299,
|
| 268 |
+
0.43396440148353577,
|
| 269 |
+
0.265097439289093,
|
| 270 |
+
0.6655389070510864,
|
| 271 |
+
0.5516760349273682,
|
| 272 |
+
0.39213114976882935,
|
| 273 |
+
0.4633840322494507,
|
| 274 |
+
0.47202637791633606,
|
| 275 |
+
0.5643339157104492,
|
| 276 |
+
0.12518832087516785,
|
| 277 |
+
0.1820685714483261,
|
| 278 |
+
0.3878677785396576,
|
| 279 |
+
0.1094706803560257,
|
| 280 |
+
0.468102365732193,
|
| 281 |
+
0.13371150195598602,
|
| 282 |
+
0.7488715052604675
|
| 283 |
+
]
|
| 284 |
+
}
|
p5-mae-15m-d131/exports/step_00040000/lora_1ep_s0.json
ADDED
|
@@ -0,0 +1,260 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"meta": {
|
| 3 |
+
"export": "/workspace/runs/p5-mae-15m-d131/exports/step_00040000",
|
| 4 |
+
"fingerprint": "f1b23242ec8f5601",
|
| 5 |
+
"campaign_id": "phase5",
|
| 6 |
+
"trial_id": "p5-mae-15m-d131",
|
| 7 |
+
"lora_layers_wrapped": 32,
|
| 8 |
+
"params": {
|
| 9 |
+
"trainable": 1002383,
|
| 10 |
+
"total": 15284879,
|
| 11 |
+
"fraction": 0.0655800415561026
|
| 12 |
+
},
|
| 13 |
+
"steps_per_epoch": 428,
|
| 14 |
+
"total_steps": 428,
|
| 15 |
+
"train_clips": 20550,
|
| 16 |
+
"val_clips": 18886,
|
| 17 |
+
"config": {
|
| 18 |
+
"rank": 16,
|
| 19 |
+
"alpha": 32.0,
|
| 20 |
+
"targets": [
|
| 21 |
+
"qkv",
|
| 22 |
+
"proj",
|
| 23 |
+
"fc1",
|
| 24 |
+
"fc2"
|
| 25 |
+
],
|
| 26 |
+
"train_norms": true,
|
| 27 |
+
"epochs": 1.0,
|
| 28 |
+
"batch_size": 48,
|
| 29 |
+
"lr": 0.003,
|
| 30 |
+
"min_lr": 1e-05,
|
| 31 |
+
"warmup_frac": 0.1,
|
| 32 |
+
"betas": [
|
| 33 |
+
0.9,
|
| 34 |
+
0.95
|
| 35 |
+
],
|
| 36 |
+
"weight_decay": 0.05,
|
| 37 |
+
"clip_norm": 4.0,
|
| 38 |
+
"mixup": 0.0,
|
| 39 |
+
"mixup_prob": 0.9,
|
| 40 |
+
"freq_mask": 0,
|
| 41 |
+
"time_mask": 0,
|
| 42 |
+
"roll_aug": false,
|
| 43 |
+
"mask_ratio": 0.0,
|
| 44 |
+
"num_classes": 527,
|
| 45 |
+
"prior_bias_init": true,
|
| 46 |
+
"num_workers": 8,
|
| 47 |
+
"seed": 0,
|
| 48 |
+
"train_probe_clips": 4096,
|
| 49 |
+
"train_eval_every": 25,
|
| 50 |
+
"val_evals": 5,
|
| 51 |
+
"eval_batch_size": 128
|
| 52 |
+
},
|
| 53 |
+
"gpu": "NVIDIA GeForce RTX 4090"
|
| 54 |
+
},
|
| 55 |
+
"history": [
|
| 56 |
+
{
|
| 57 |
+
"step": 0,
|
| 58 |
+
"epoch": 0.0,
|
| 59 |
+
"loss": null,
|
| 60 |
+
"train_map": 0.006735008502417939,
|
| 61 |
+
"val_map": 0.005770288722808747,
|
| 62 |
+
"lr": 0.0,
|
| 63 |
+
"train_s": 0.0
|
| 64 |
+
},
|
| 65 |
+
{
|
| 66 |
+
"step": 25,
|
| 67 |
+
"epoch": 0.0584,
|
| 68 |
+
"loss": 0.02353,
|
| 69 |
+
"train_map": 0.029018835189789392,
|
| 70 |
+
"val_map": null,
|
| 71 |
+
"lr": 0.0017857142857142857,
|
| 72 |
+
"train_s": 0.7
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"step": 50,
|
| 76 |
+
"epoch": 0.1168,
|
| 77 |
+
"loss": 0.02293,
|
| 78 |
+
"train_map": 0.056034751141450914,
|
| 79 |
+
"val_map": null,
|
| 80 |
+
"lr": 0.002996832162669313,
|
| 81 |
+
"train_s": 1.2
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"step": 75,
|
| 85 |
+
"epoch": 0.1752,
|
| 86 |
+
"loss": 0.02253,
|
| 87 |
+
"train_map": 0.07593185758037914,
|
| 88 |
+
"val_map": null,
|
| 89 |
+
"lr": 0.0029464015850084926,
|
| 90 |
+
"train_s": 1.8
|
| 91 |
+
},
|
| 92 |
+
{
|
| 93 |
+
"step": 100,
|
| 94 |
+
"epoch": 0.2336,
|
| 95 |
+
"loss": 0.02169,
|
| 96 |
+
"train_map": 0.08663212330976547,
|
| 97 |
+
"val_map": null,
|
| 98 |
+
"lr": 0.0028365019165406584,
|
| 99 |
+
"train_s": 2.4
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"step": 107,
|
| 103 |
+
"epoch": 0.25,
|
| 104 |
+
"loss": 0.02167,
|
| 105 |
+
"train_map": 0.09157179803428447,
|
| 106 |
+
"val_map": 0.06147507686553209,
|
| 107 |
+
"lr": 0.002795633084492174,
|
| 108 |
+
"train_s": 2.6
|
| 109 |
+
},
|
| 110 |
+
{
|
| 111 |
+
"step": 125,
|
| 112 |
+
"epoch": 0.2921,
|
| 113 |
+
"loss": 0.0208,
|
| 114 |
+
"train_map": 0.09879499070427647,
|
| 115 |
+
"val_map": null,
|
| 116 |
+
"lr": 0.0026716673782849554,
|
| 117 |
+
"train_s": 3.0
|
| 118 |
+
},
|
| 119 |
+
{
|
| 120 |
+
"step": 150,
|
| 121 |
+
"epoch": 0.3505,
|
| 122 |
+
"loss": 0.02029,
|
| 123 |
+
"train_map": 0.1138976350500897,
|
| 124 |
+
"val_map": null,
|
| 125 |
+
"lr": 0.002458698684369641,
|
| 126 |
+
"train_s": 3.6
|
| 127 |
+
},
|
| 128 |
+
{
|
| 129 |
+
"step": 175,
|
| 130 |
+
"epoch": 0.4089,
|
| 131 |
+
"loss": 0.02015,
|
| 132 |
+
"train_map": 0.13228270413921314,
|
| 133 |
+
"val_map": null,
|
| 134 |
+
"lr": 0.0022063824593754194,
|
| 135 |
+
"train_s": 4.2
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"step": 200,
|
| 139 |
+
"epoch": 0.4673,
|
| 140 |
+
"loss": 0.01992,
|
| 141 |
+
"train_map": 0.14935516663058263,
|
| 142 |
+
"val_map": null,
|
| 143 |
+
"lr": 0.0019251287213377403,
|
| 144 |
+
"train_s": 4.8
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"step": 214,
|
| 148 |
+
"epoch": 0.5,
|
| 149 |
+
"loss": 0.01924,
|
| 150 |
+
"train_map": 0.1619921555049873,
|
| 151 |
+
"val_map": 0.09994390278619907,
|
| 152 |
+
"lr": 0.0017592766878384548,
|
| 153 |
+
"train_s": 5.2
|
| 154 |
+
},
|
| 155 |
+
{
|
| 156 |
+
"step": 225,
|
| 157 |
+
"epoch": 0.5257,
|
| 158 |
+
"loss": 0.01947,
|
| 159 |
+
"train_map": 0.15973982295159259,
|
| 160 |
+
"val_map": null,
|
| 161 |
+
"lr": 0.0016265413870700269,
|
| 162 |
+
"train_s": 5.4
|
| 163 |
+
},
|
| 164 |
+
{
|
| 165 |
+
"step": 250,
|
| 166 |
+
"epoch": 0.5841,
|
| 167 |
+
"loss": 0.01934,
|
| 168 |
+
"train_map": 0.17656947687034985,
|
| 169 |
+
"val_map": null,
|
| 170 |
+
"lr": 0.0013229395198229876,
|
| 171 |
+
"train_s": 6.0
|
| 172 |
+
},
|
| 173 |
+
{
|
| 174 |
+
"step": 275,
|
| 175 |
+
"epoch": 0.6425,
|
| 176 |
+
"loss": 0.01919,
|
| 177 |
+
"train_map": 0.19340735260906153,
|
| 178 |
+
"val_map": null,
|
| 179 |
+
"lr": 0.0010268490715591258,
|
| 180 |
+
"train_s": 6.6
|
| 181 |
+
},
|
| 182 |
+
{
|
| 183 |
+
"step": 300,
|
| 184 |
+
"epoch": 0.7009,
|
| 185 |
+
"loss": 0.01876,
|
| 186 |
+
"train_map": 0.20726251004500565,
|
| 187 |
+
"val_map": null,
|
| 188 |
+
"lr": 0.000750486089449494,
|
| 189 |
+
"train_s": 7.1
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"step": 321,
|
| 193 |
+
"epoch": 0.75,
|
| 194 |
+
"loss": 0.01806,
|
| 195 |
+
"train_map": 0.21370753363495873,
|
| 196 |
+
"val_map": 0.130612306568093,
|
| 197 |
+
"lr": 0.000541962785848462,
|
| 198 |
+
"train_s": 7.6
|
| 199 |
+
},
|
| 200 |
+
{
|
| 201 |
+
"step": 325,
|
| 202 |
+
"epoch": 0.7593,
|
| 203 |
+
"loss": 0.01848,
|
| 204 |
+
"train_map": 0.21679889505103486,
|
| 205 |
+
"val_map": null,
|
| 206 |
+
"lr": 0.0005052527083673369,
|
| 207 |
+
"train_s": 7.7
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"step": 350,
|
| 211 |
+
"epoch": 0.8178,
|
| 212 |
+
"loss": 0.01791,
|
| 213 |
+
"train_map": 0.2245621782050696,
|
| 214 |
+
"val_map": null,
|
| 215 |
+
"lr": 0.00030126672363088163,
|
| 216 |
+
"train_s": 8.3
|
| 217 |
+
},
|
| 218 |
+
{
|
| 219 |
+
"step": 375,
|
| 220 |
+
"epoch": 0.8762,
|
| 221 |
+
"loss": 0.0174,
|
| 222 |
+
"train_map": 0.23291879061934675,
|
| 223 |
+
"val_map": null,
|
| 224 |
+
"lr": 0.00014694415279957929,
|
| 225 |
+
"train_s": 8.9
|
| 226 |
+
},
|
| 227 |
+
{
|
| 228 |
+
"step": 400,
|
| 229 |
+
"epoch": 0.9346,
|
| 230 |
+
"loss": 0.01707,
|
| 231 |
+
"train_map": 0.23780500212157643,
|
| 232 |
+
"val_map": null,
|
| 233 |
+
"lr": 4.865200911511649e-05,
|
| 234 |
+
"train_s": 9.4
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"step": 425,
|
| 238 |
+
"epoch": 0.993,
|
| 239 |
+
"loss": 0.01771,
|
| 240 |
+
"train_map": 0.23784616059734992,
|
| 241 |
+
"val_map": null,
|
| 242 |
+
"lr": 1.0445612398966596e-05,
|
| 243 |
+
"train_s": 10.0
|
| 244 |
+
},
|
| 245 |
+
{
|
| 246 |
+
"step": 428,
|
| 247 |
+
"epoch": 1.0,
|
| 248 |
+
"loss": 0.01777,
|
| 249 |
+
"train_map": 0.23835410517189773,
|
| 250 |
+
"val_map": 0.14589168229962096,
|
| 251 |
+
"lr": 1e-05,
|
| 252 |
+
"train_s": 10.1
|
| 253 |
+
}
|
| 254 |
+
],
|
| 255 |
+
"final_train_map": 0.23835410517189773,
|
| 256 |
+
"final_val_map": 0.14589168229962096,
|
| 257 |
+
"lora/map": 0.14589168229962096,
|
| 258 |
+
"train_seconds": 10.1,
|
| 259 |
+
"wall_seconds": 61.5
|
| 260 |
+
}
|
p5-mae-15m-d131/exports/step_00040000/lora_1ep_s1.json
ADDED
|
@@ -0,0 +1,260 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"meta": {
|
| 3 |
+
"export": "/workspace/runs/p5-mae-15m-d131/exports/step_00040000",
|
| 4 |
+
"fingerprint": "f1b23242ec8f5601",
|
| 5 |
+
"campaign_id": "phase5",
|
| 6 |
+
"trial_id": "p5-mae-15m-d131",
|
| 7 |
+
"lora_layers_wrapped": 32,
|
| 8 |
+
"params": {
|
| 9 |
+
"trainable": 1002383,
|
| 10 |
+
"total": 15284879,
|
| 11 |
+
"fraction": 0.0655800415561026
|
| 12 |
+
},
|
| 13 |
+
"steps_per_epoch": 428,
|
| 14 |
+
"total_steps": 428,
|
| 15 |
+
"train_clips": 20550,
|
| 16 |
+
"val_clips": 18886,
|
| 17 |
+
"config": {
|
| 18 |
+
"rank": 16,
|
| 19 |
+
"alpha": 32.0,
|
| 20 |
+
"targets": [
|
| 21 |
+
"qkv",
|
| 22 |
+
"proj",
|
| 23 |
+
"fc1",
|
| 24 |
+
"fc2"
|
| 25 |
+
],
|
| 26 |
+
"train_norms": true,
|
| 27 |
+
"epochs": 1.0,
|
| 28 |
+
"batch_size": 48,
|
| 29 |
+
"lr": 0.003,
|
| 30 |
+
"min_lr": 1e-05,
|
| 31 |
+
"warmup_frac": 0.1,
|
| 32 |
+
"betas": [
|
| 33 |
+
0.9,
|
| 34 |
+
0.95
|
| 35 |
+
],
|
| 36 |
+
"weight_decay": 0.05,
|
| 37 |
+
"clip_norm": 4.0,
|
| 38 |
+
"mixup": 0.0,
|
| 39 |
+
"mixup_prob": 0.9,
|
| 40 |
+
"freq_mask": 0,
|
| 41 |
+
"time_mask": 0,
|
| 42 |
+
"roll_aug": false,
|
| 43 |
+
"mask_ratio": 0.0,
|
| 44 |
+
"num_classes": 527,
|
| 45 |
+
"prior_bias_init": true,
|
| 46 |
+
"num_workers": 8,
|
| 47 |
+
"seed": 1,
|
| 48 |
+
"train_probe_clips": 4096,
|
| 49 |
+
"train_eval_every": 25,
|
| 50 |
+
"val_evals": 5,
|
| 51 |
+
"eval_batch_size": 128
|
| 52 |
+
},
|
| 53 |
+
"gpu": "NVIDIA GeForce RTX 4090"
|
| 54 |
+
},
|
| 55 |
+
"history": [
|
| 56 |
+
{
|
| 57 |
+
"step": 0,
|
| 58 |
+
"epoch": 0.0,
|
| 59 |
+
"loss": null,
|
| 60 |
+
"train_map": 0.006481850774034476,
|
| 61 |
+
"val_map": 0.0058123897699253945,
|
| 62 |
+
"lr": 0.0,
|
| 63 |
+
"train_s": 0.0
|
| 64 |
+
},
|
| 65 |
+
{
|
| 66 |
+
"step": 25,
|
| 67 |
+
"epoch": 0.0584,
|
| 68 |
+
"loss": 0.0245,
|
| 69 |
+
"train_map": 0.036172089105086715,
|
| 70 |
+
"val_map": null,
|
| 71 |
+
"lr": 0.0017857142857142857,
|
| 72 |
+
"train_s": 0.6
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"step": 50,
|
| 76 |
+
"epoch": 0.1168,
|
| 77 |
+
"loss": 0.02289,
|
| 78 |
+
"train_map": 0.05671455155182418,
|
| 79 |
+
"val_map": null,
|
| 80 |
+
"lr": 0.002996832162669313,
|
| 81 |
+
"train_s": 1.3
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"step": 75,
|
| 85 |
+
"epoch": 0.1752,
|
| 86 |
+
"loss": 0.02189,
|
| 87 |
+
"train_map": 0.07450216201481284,
|
| 88 |
+
"val_map": null,
|
| 89 |
+
"lr": 0.0029464015850084926,
|
| 90 |
+
"train_s": 1.9
|
| 91 |
+
},
|
| 92 |
+
{
|
| 93 |
+
"step": 100,
|
| 94 |
+
"epoch": 0.2336,
|
| 95 |
+
"loss": 0.02157,
|
| 96 |
+
"train_map": 0.0791278298011145,
|
| 97 |
+
"val_map": null,
|
| 98 |
+
"lr": 0.0028365019165406584,
|
| 99 |
+
"train_s": 2.5
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"step": 107,
|
| 103 |
+
"epoch": 0.25,
|
| 104 |
+
"loss": 0.02185,
|
| 105 |
+
"train_map": 0.09432549659358037,
|
| 106 |
+
"val_map": 0.06076292439842812,
|
| 107 |
+
"lr": 0.002795633084492174,
|
| 108 |
+
"train_s": 2.6
|
| 109 |
+
},
|
| 110 |
+
{
|
| 111 |
+
"step": 125,
|
| 112 |
+
"epoch": 0.2921,
|
| 113 |
+
"loss": 0.02063,
|
| 114 |
+
"train_map": 0.09717436261546658,
|
| 115 |
+
"val_map": null,
|
| 116 |
+
"lr": 0.0026716673782849554,
|
| 117 |
+
"train_s": 3.1
|
| 118 |
+
},
|
| 119 |
+
{
|
| 120 |
+
"step": 150,
|
| 121 |
+
"epoch": 0.3505,
|
| 122 |
+
"loss": 0.02075,
|
| 123 |
+
"train_map": 0.11098967896919323,
|
| 124 |
+
"val_map": null,
|
| 125 |
+
"lr": 0.002458698684369641,
|
| 126 |
+
"train_s": 3.7
|
| 127 |
+
},
|
| 128 |
+
{
|
| 129 |
+
"step": 175,
|
| 130 |
+
"epoch": 0.4089,
|
| 131 |
+
"loss": 0.01994,
|
| 132 |
+
"train_map": 0.12432891728428937,
|
| 133 |
+
"val_map": null,
|
| 134 |
+
"lr": 0.0022063824593754194,
|
| 135 |
+
"train_s": 4.3
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"step": 200,
|
| 139 |
+
"epoch": 0.4673,
|
| 140 |
+
"loss": 0.02033,
|
| 141 |
+
"train_map": 0.13331043523554276,
|
| 142 |
+
"val_map": null,
|
| 143 |
+
"lr": 0.0019251287213377403,
|
| 144 |
+
"train_s": 4.8
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"step": 214,
|
| 148 |
+
"epoch": 0.5,
|
| 149 |
+
"loss": 0.01953,
|
| 150 |
+
"train_map": 0.15225617017816095,
|
| 151 |
+
"val_map": 0.0933940169804392,
|
| 152 |
+
"lr": 0.0017592766878384548,
|
| 153 |
+
"train_s": 5.2
|
| 154 |
+
},
|
| 155 |
+
{
|
| 156 |
+
"step": 225,
|
| 157 |
+
"epoch": 0.5257,
|
| 158 |
+
"loss": 0.01958,
|
| 159 |
+
"train_map": 0.1573216292823345,
|
| 160 |
+
"val_map": null,
|
| 161 |
+
"lr": 0.0016265413870700269,
|
| 162 |
+
"train_s": 5.4
|
| 163 |
+
},
|
| 164 |
+
{
|
| 165 |
+
"step": 250,
|
| 166 |
+
"epoch": 0.5841,
|
| 167 |
+
"loss": 0.01918,
|
| 168 |
+
"train_map": 0.16743108225311285,
|
| 169 |
+
"val_map": null,
|
| 170 |
+
"lr": 0.0013229395198229876,
|
| 171 |
+
"train_s": 6.0
|
| 172 |
+
},
|
| 173 |
+
{
|
| 174 |
+
"step": 275,
|
| 175 |
+
"epoch": 0.6425,
|
| 176 |
+
"loss": 0.019,
|
| 177 |
+
"train_map": 0.1885812535062008,
|
| 178 |
+
"val_map": null,
|
| 179 |
+
"lr": 0.0010268490715591258,
|
| 180 |
+
"train_s": 6.6
|
| 181 |
+
},
|
| 182 |
+
{
|
| 183 |
+
"step": 300,
|
| 184 |
+
"epoch": 0.7009,
|
| 185 |
+
"loss": 0.01845,
|
| 186 |
+
"train_map": 0.20302789795620674,
|
| 187 |
+
"val_map": null,
|
| 188 |
+
"lr": 0.000750486089449494,
|
| 189 |
+
"train_s": 7.2
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"step": 321,
|
| 193 |
+
"epoch": 0.75,
|
| 194 |
+
"loss": 0.01804,
|
| 195 |
+
"train_map": 0.21604634840147433,
|
| 196 |
+
"val_map": 0.1323467909024438,
|
| 197 |
+
"lr": 0.000541962785848462,
|
| 198 |
+
"train_s": 7.7
|
| 199 |
+
},
|
| 200 |
+
{
|
| 201 |
+
"step": 325,
|
| 202 |
+
"epoch": 0.7593,
|
| 203 |
+
"loss": 0.0183,
|
| 204 |
+
"train_map": 0.21663562419066168,
|
| 205 |
+
"val_map": null,
|
| 206 |
+
"lr": 0.0005052527083673369,
|
| 207 |
+
"train_s": 7.8
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"step": 350,
|
| 211 |
+
"epoch": 0.8178,
|
| 212 |
+
"loss": 0.01762,
|
| 213 |
+
"train_map": 0.23043413541811664,
|
| 214 |
+
"val_map": null,
|
| 215 |
+
"lr": 0.00030126672363088163,
|
| 216 |
+
"train_s": 8.4
|
| 217 |
+
},
|
| 218 |
+
{
|
| 219 |
+
"step": 375,
|
| 220 |
+
"epoch": 0.8762,
|
| 221 |
+
"loss": 0.01781,
|
| 222 |
+
"train_map": 0.23213497322384954,
|
| 223 |
+
"val_map": null,
|
| 224 |
+
"lr": 0.00014694415279957929,
|
| 225 |
+
"train_s": 9.0
|
| 226 |
+
},
|
| 227 |
+
{
|
| 228 |
+
"step": 400,
|
| 229 |
+
"epoch": 0.9346,
|
| 230 |
+
"loss": 0.0176,
|
| 231 |
+
"train_map": 0.23502737978575805,
|
| 232 |
+
"val_map": null,
|
| 233 |
+
"lr": 4.865200911511649e-05,
|
| 234 |
+
"train_s": 9.6
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"step": 425,
|
| 238 |
+
"epoch": 0.993,
|
| 239 |
+
"loss": 0.01759,
|
| 240 |
+
"train_map": 0.23581013416931515,
|
| 241 |
+
"val_map": null,
|
| 242 |
+
"lr": 1.0445612398966596e-05,
|
| 243 |
+
"train_s": 10.2
|
| 244 |
+
},
|
| 245 |
+
{
|
| 246 |
+
"step": 428,
|
| 247 |
+
"epoch": 1.0,
|
| 248 |
+
"loss": 0.01735,
|
| 249 |
+
"train_map": 0.23562746514271118,
|
| 250 |
+
"val_map": 0.14515814030864602,
|
| 251 |
+
"lr": 1e-05,
|
| 252 |
+
"train_s": 10.2
|
| 253 |
+
}
|
| 254 |
+
],
|
| 255 |
+
"final_train_map": 0.23562746514271118,
|
| 256 |
+
"final_val_map": 0.14515814030864602,
|
| 257 |
+
"lora/map": 0.14515814030864602,
|
| 258 |
+
"train_seconds": 10.2,
|
| 259 |
+
"wall_seconds": 61.7
|
| 260 |
+
}
|
p5-mae-15m-d131/exports/step_00040000/lora_1ep_s2.json
ADDED
|
@@ -0,0 +1,260 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"meta": {
|
| 3 |
+
"export": "/workspace/runs/p5-mae-15m-d131/exports/step_00040000",
|
| 4 |
+
"fingerprint": "f1b23242ec8f5601",
|
| 5 |
+
"campaign_id": "phase5",
|
| 6 |
+
"trial_id": "p5-mae-15m-d131",
|
| 7 |
+
"lora_layers_wrapped": 32,
|
| 8 |
+
"params": {
|
| 9 |
+
"trainable": 1002383,
|
| 10 |
+
"total": 15284879,
|
| 11 |
+
"fraction": 0.0655800415561026
|
| 12 |
+
},
|
| 13 |
+
"steps_per_epoch": 428,
|
| 14 |
+
"total_steps": 428,
|
| 15 |
+
"train_clips": 20550,
|
| 16 |
+
"val_clips": 18886,
|
| 17 |
+
"config": {
|
| 18 |
+
"rank": 16,
|
| 19 |
+
"alpha": 32.0,
|
| 20 |
+
"targets": [
|
| 21 |
+
"qkv",
|
| 22 |
+
"proj",
|
| 23 |
+
"fc1",
|
| 24 |
+
"fc2"
|
| 25 |
+
],
|
| 26 |
+
"train_norms": true,
|
| 27 |
+
"epochs": 1.0,
|
| 28 |
+
"batch_size": 48,
|
| 29 |
+
"lr": 0.003,
|
| 30 |
+
"min_lr": 1e-05,
|
| 31 |
+
"warmup_frac": 0.1,
|
| 32 |
+
"betas": [
|
| 33 |
+
0.9,
|
| 34 |
+
0.95
|
| 35 |
+
],
|
| 36 |
+
"weight_decay": 0.05,
|
| 37 |
+
"clip_norm": 4.0,
|
| 38 |
+
"mixup": 0.0,
|
| 39 |
+
"mixup_prob": 0.9,
|
| 40 |
+
"freq_mask": 0,
|
| 41 |
+
"time_mask": 0,
|
| 42 |
+
"roll_aug": false,
|
| 43 |
+
"mask_ratio": 0.0,
|
| 44 |
+
"num_classes": 527,
|
| 45 |
+
"prior_bias_init": true,
|
| 46 |
+
"num_workers": 8,
|
| 47 |
+
"seed": 2,
|
| 48 |
+
"train_probe_clips": 4096,
|
| 49 |
+
"train_eval_every": 25,
|
| 50 |
+
"val_evals": 5,
|
| 51 |
+
"eval_batch_size": 128
|
| 52 |
+
},
|
| 53 |
+
"gpu": "NVIDIA GeForce RTX 4090"
|
| 54 |
+
},
|
| 55 |
+
"history": [
|
| 56 |
+
{
|
| 57 |
+
"step": 0,
|
| 58 |
+
"epoch": 0.0,
|
| 59 |
+
"loss": null,
|
| 60 |
+
"train_map": 0.006204049494510704,
|
| 61 |
+
"val_map": 0.005416988269240371,
|
| 62 |
+
"lr": 0.0,
|
| 63 |
+
"train_s": 0.0
|
| 64 |
+
},
|
| 65 |
+
{
|
| 66 |
+
"step": 25,
|
| 67 |
+
"epoch": 0.0584,
|
| 68 |
+
"loss": 0.02408,
|
| 69 |
+
"train_map": 0.03605000394068666,
|
| 70 |
+
"val_map": null,
|
| 71 |
+
"lr": 0.0017857142857142857,
|
| 72 |
+
"train_s": 0.7
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"step": 50,
|
| 76 |
+
"epoch": 0.1168,
|
| 77 |
+
"loss": 0.02318,
|
| 78 |
+
"train_map": 0.05449432550895909,
|
| 79 |
+
"val_map": null,
|
| 80 |
+
"lr": 0.002996832162669313,
|
| 81 |
+
"train_s": 1.3
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"step": 75,
|
| 85 |
+
"epoch": 0.1752,
|
| 86 |
+
"loss": 0.02206,
|
| 87 |
+
"train_map": 0.06592995083190155,
|
| 88 |
+
"val_map": null,
|
| 89 |
+
"lr": 0.0029464015850084926,
|
| 90 |
+
"train_s": 1.9
|
| 91 |
+
},
|
| 92 |
+
{
|
| 93 |
+
"step": 100,
|
| 94 |
+
"epoch": 0.2336,
|
| 95 |
+
"loss": 0.0211,
|
| 96 |
+
"train_map": 0.08851913837901541,
|
| 97 |
+
"val_map": null,
|
| 98 |
+
"lr": 0.0028365019165406584,
|
| 99 |
+
"train_s": 2.5
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"step": 107,
|
| 103 |
+
"epoch": 0.25,
|
| 104 |
+
"loss": 0.02151,
|
| 105 |
+
"train_map": 0.09253515394522051,
|
| 106 |
+
"val_map": 0.05851254311780776,
|
| 107 |
+
"lr": 0.002795633084492174,
|
| 108 |
+
"train_s": 2.6
|
| 109 |
+
},
|
| 110 |
+
{
|
| 111 |
+
"step": 125,
|
| 112 |
+
"epoch": 0.2921,
|
| 113 |
+
"loss": 0.02138,
|
| 114 |
+
"train_map": 0.09272094685725683,
|
| 115 |
+
"val_map": null,
|
| 116 |
+
"lr": 0.0026716673782849554,
|
| 117 |
+
"train_s": 3.1
|
| 118 |
+
},
|
| 119 |
+
{
|
| 120 |
+
"step": 150,
|
| 121 |
+
"epoch": 0.3505,
|
| 122 |
+
"loss": 0.02106,
|
| 123 |
+
"train_map": 0.11467579004409559,
|
| 124 |
+
"val_map": null,
|
| 125 |
+
"lr": 0.002458698684369641,
|
| 126 |
+
"train_s": 3.7
|
| 127 |
+
},
|
| 128 |
+
{
|
| 129 |
+
"step": 175,
|
| 130 |
+
"epoch": 0.4089,
|
| 131 |
+
"loss": 0.02052,
|
| 132 |
+
"train_map": 0.1255666836919493,
|
| 133 |
+
"val_map": null,
|
| 134 |
+
"lr": 0.0022063824593754194,
|
| 135 |
+
"train_s": 4.3
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"step": 200,
|
| 139 |
+
"epoch": 0.4673,
|
| 140 |
+
"loss": 0.01942,
|
| 141 |
+
"train_map": 0.13507897033580565,
|
| 142 |
+
"val_map": null,
|
| 143 |
+
"lr": 0.0019251287213377403,
|
| 144 |
+
"train_s": 4.9
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"step": 214,
|
| 148 |
+
"epoch": 0.5,
|
| 149 |
+
"loss": 0.01901,
|
| 150 |
+
"train_map": 0.14833897147840153,
|
| 151 |
+
"val_map": 0.09166979949490169,
|
| 152 |
+
"lr": 0.0017592766878384548,
|
| 153 |
+
"train_s": 5.2
|
| 154 |
+
},
|
| 155 |
+
{
|
| 156 |
+
"step": 225,
|
| 157 |
+
"epoch": 0.5257,
|
| 158 |
+
"loss": 0.01909,
|
| 159 |
+
"train_map": 0.15982337880780675,
|
| 160 |
+
"val_map": null,
|
| 161 |
+
"lr": 0.0016265413870700269,
|
| 162 |
+
"train_s": 5.5
|
| 163 |
+
},
|
| 164 |
+
{
|
| 165 |
+
"step": 250,
|
| 166 |
+
"epoch": 0.5841,
|
| 167 |
+
"loss": 0.01881,
|
| 168 |
+
"train_map": 0.1765116685223845,
|
| 169 |
+
"val_map": null,
|
| 170 |
+
"lr": 0.0013229395198229876,
|
| 171 |
+
"train_s": 6.1
|
| 172 |
+
},
|
| 173 |
+
{
|
| 174 |
+
"step": 275,
|
| 175 |
+
"epoch": 0.6425,
|
| 176 |
+
"loss": 0.01899,
|
| 177 |
+
"train_map": 0.18822113390962472,
|
| 178 |
+
"val_map": null,
|
| 179 |
+
"lr": 0.0010268490715591258,
|
| 180 |
+
"train_s": 6.6
|
| 181 |
+
},
|
| 182 |
+
{
|
| 183 |
+
"step": 300,
|
| 184 |
+
"epoch": 0.7009,
|
| 185 |
+
"loss": 0.01844,
|
| 186 |
+
"train_map": 0.19934520773957193,
|
| 187 |
+
"val_map": null,
|
| 188 |
+
"lr": 0.000750486089449494,
|
| 189 |
+
"train_s": 7.2
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"step": 321,
|
| 193 |
+
"epoch": 0.75,
|
| 194 |
+
"loss": 0.01809,
|
| 195 |
+
"train_map": 0.21047947148790677,
|
| 196 |
+
"val_map": 0.13123024838395148,
|
| 197 |
+
"lr": 0.000541962785848462,
|
| 198 |
+
"train_s": 7.7
|
| 199 |
+
},
|
| 200 |
+
{
|
| 201 |
+
"step": 325,
|
| 202 |
+
"epoch": 0.7593,
|
| 203 |
+
"loss": 0.01785,
|
| 204 |
+
"train_map": 0.2120579554720568,
|
| 205 |
+
"val_map": null,
|
| 206 |
+
"lr": 0.0005052527083673369,
|
| 207 |
+
"train_s": 7.8
|
| 208 |
+
},
|
| 209 |
+
{
|
| 210 |
+
"step": 350,
|
| 211 |
+
"epoch": 0.8178,
|
| 212 |
+
"loss": 0.01795,
|
| 213 |
+
"train_map": 0.22381515664523874,
|
| 214 |
+
"val_map": null,
|
| 215 |
+
"lr": 0.00030126672363088163,
|
| 216 |
+
"train_s": 8.4
|
| 217 |
+
},
|
| 218 |
+
{
|
| 219 |
+
"step": 375,
|
| 220 |
+
"epoch": 0.8762,
|
| 221 |
+
"loss": 0.01801,
|
| 222 |
+
"train_map": 0.23167820093866995,
|
| 223 |
+
"val_map": null,
|
| 224 |
+
"lr": 0.00014694415279957929,
|
| 225 |
+
"train_s": 8.9
|
| 226 |
+
},
|
| 227 |
+
{
|
| 228 |
+
"step": 400,
|
| 229 |
+
"epoch": 0.9346,
|
| 230 |
+
"loss": 0.01766,
|
| 231 |
+
"train_map": 0.23196428618349657,
|
| 232 |
+
"val_map": null,
|
| 233 |
+
"lr": 4.865200911511649e-05,
|
| 234 |
+
"train_s": 9.5
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"step": 425,
|
| 238 |
+
"epoch": 0.993,
|
| 239 |
+
"loss": 0.01773,
|
| 240 |
+
"train_map": 0.2339012164399876,
|
| 241 |
+
"val_map": null,
|
| 242 |
+
"lr": 1.0445612398966596e-05,
|
| 243 |
+
"train_s": 10.1
|
| 244 |
+
},
|
| 245 |
+
{
|
| 246 |
+
"step": 428,
|
| 247 |
+
"epoch": 1.0,
|
| 248 |
+
"loss": 0.01801,
|
| 249 |
+
"train_map": 0.23370954823597162,
|
| 250 |
+
"val_map": 0.1438193091975995,
|
| 251 |
+
"lr": 1e-05,
|
| 252 |
+
"train_s": 10.1
|
| 253 |
+
}
|
| 254 |
+
],
|
| 255 |
+
"final_train_map": 0.23370954823597162,
|
| 256 |
+
"final_val_map": 0.1438193091975995,
|
| 257 |
+
"lora/map": 0.1438193091975995,
|
| 258 |
+
"train_seconds": 10.1,
|
| 259 |
+
"wall_seconds": 61.2
|
| 260 |
+
}
|