End of training
Browse files
all_results.json
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 2.99,
|
| 3 |
+
"eval_accuracy": 0.9973333333333333,
|
| 4 |
+
"eval_loss": 0.005204513669013977,
|
| 5 |
+
"eval_runtime": 4.3172,
|
| 6 |
+
"eval_samples_per_second": 86.862,
|
| 7 |
+
"eval_steps_per_second": 10.887,
|
| 8 |
+
"total_flos": 2.5044894502148506e+17,
|
| 9 |
+
"train_loss": 0.10708669147321156,
|
| 10 |
+
"train_runtime": 1544.9876,
|
| 11 |
+
"train_samples_per_second": 6.552,
|
| 12 |
+
"train_steps_per_second": 0.204
|
| 13 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 2.99,
|
| 3 |
+
"eval_accuracy": 0.9973333333333333,
|
| 4 |
+
"eval_loss": 0.005204513669013977,
|
| 5 |
+
"eval_runtime": 4.3172,
|
| 6 |
+
"eval_samples_per_second": 86.862,
|
| 7 |
+
"eval_steps_per_second": 10.887
|
| 8 |
+
}
|
runs/Apr16_00-47-58_e141a17aebe5/events.out.tfevents.1681607659.e141a17aebe5.1112.9
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:de90de234c71470940bc64a6d4ff3798502499d9bc29f4253b96ef848b4dffa6
|
| 3 |
+
size 411
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 2.99,
|
| 3 |
+
"total_flos": 2.5044894502148506e+17,
|
| 4 |
+
"train_loss": 0.10708669147321156,
|
| 5 |
+
"train_runtime": 1544.9876,
|
| 6 |
+
"train_samples_per_second": 6.552,
|
| 7 |
+
"train_steps_per_second": 0.204
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,238 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": 0.9973333333333333,
|
| 3 |
+
"best_model_checkpoint": "swin-tiny-patch4-window7-224-short_sleeve/checkpoint-315",
|
| 4 |
+
"epoch": 2.985781990521327,
|
| 5 |
+
"global_step": 315,
|
| 6 |
+
"is_hyper_param_search": false,
|
| 7 |
+
"is_local_process_zero": true,
|
| 8 |
+
"is_world_process_zero": true,
|
| 9 |
+
"log_history": [
|
| 10 |
+
{
|
| 11 |
+
"epoch": 0.09,
|
| 12 |
+
"learning_rate": 1.5625e-05,
|
| 13 |
+
"loss": 0.6458,
|
| 14 |
+
"step": 10
|
| 15 |
+
},
|
| 16 |
+
{
|
| 17 |
+
"epoch": 0.19,
|
| 18 |
+
"learning_rate": 3.125e-05,
|
| 19 |
+
"loss": 0.5051,
|
| 20 |
+
"step": 20
|
| 21 |
+
},
|
| 22 |
+
{
|
| 23 |
+
"epoch": 0.28,
|
| 24 |
+
"learning_rate": 4.6875e-05,
|
| 25 |
+
"loss": 0.2574,
|
| 26 |
+
"step": 30
|
| 27 |
+
},
|
| 28 |
+
{
|
| 29 |
+
"epoch": 0.38,
|
| 30 |
+
"learning_rate": 4.858657243816255e-05,
|
| 31 |
+
"loss": 0.1243,
|
| 32 |
+
"step": 40
|
| 33 |
+
},
|
| 34 |
+
{
|
| 35 |
+
"epoch": 0.47,
|
| 36 |
+
"learning_rate": 4.6819787985865726e-05,
|
| 37 |
+
"loss": 0.0954,
|
| 38 |
+
"step": 50
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"epoch": 0.57,
|
| 42 |
+
"learning_rate": 4.50530035335689e-05,
|
| 43 |
+
"loss": 0.1013,
|
| 44 |
+
"step": 60
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.66,
|
| 48 |
+
"learning_rate": 4.328621908127209e-05,
|
| 49 |
+
"loss": 0.0827,
|
| 50 |
+
"step": 70
|
| 51 |
+
},
|
| 52 |
+
{
|
| 53 |
+
"epoch": 0.76,
|
| 54 |
+
"learning_rate": 4.1519434628975266e-05,
|
| 55 |
+
"loss": 0.1021,
|
| 56 |
+
"step": 80
|
| 57 |
+
},
|
| 58 |
+
{
|
| 59 |
+
"epoch": 0.85,
|
| 60 |
+
"learning_rate": 3.975265017667845e-05,
|
| 61 |
+
"loss": 0.0351,
|
| 62 |
+
"step": 90
|
| 63 |
+
},
|
| 64 |
+
{
|
| 65 |
+
"epoch": 0.95,
|
| 66 |
+
"learning_rate": 3.7985865724381624e-05,
|
| 67 |
+
"loss": 0.1194,
|
| 68 |
+
"step": 100
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"epoch": 1.0,
|
| 72 |
+
"eval_accuracy": 0.992,
|
| 73 |
+
"eval_loss": 0.018144262954592705,
|
| 74 |
+
"eval_runtime": 127.7806,
|
| 75 |
+
"eval_samples_per_second": 2.935,
|
| 76 |
+
"eval_steps_per_second": 0.368,
|
| 77 |
+
"step": 105
|
| 78 |
+
},
|
| 79 |
+
{
|
| 80 |
+
"epoch": 1.04,
|
| 81 |
+
"learning_rate": 3.621908127208481e-05,
|
| 82 |
+
"loss": 0.0895,
|
| 83 |
+
"step": 110
|
| 84 |
+
},
|
| 85 |
+
{
|
| 86 |
+
"epoch": 1.14,
|
| 87 |
+
"learning_rate": 3.445229681978799e-05,
|
| 88 |
+
"loss": 0.0846,
|
| 89 |
+
"step": 120
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"epoch": 1.23,
|
| 93 |
+
"learning_rate": 3.2685512367491163e-05,
|
| 94 |
+
"loss": 0.0791,
|
| 95 |
+
"step": 130
|
| 96 |
+
},
|
| 97 |
+
{
|
| 98 |
+
"epoch": 1.33,
|
| 99 |
+
"learning_rate": 3.0918727915194346e-05,
|
| 100 |
+
"loss": 0.0712,
|
| 101 |
+
"step": 140
|
| 102 |
+
},
|
| 103 |
+
{
|
| 104 |
+
"epoch": 1.42,
|
| 105 |
+
"learning_rate": 2.915194346289753e-05,
|
| 106 |
+
"loss": 0.0972,
|
| 107 |
+
"step": 150
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 1.52,
|
| 111 |
+
"learning_rate": 2.738515901060071e-05,
|
| 112 |
+
"loss": 0.0666,
|
| 113 |
+
"step": 160
|
| 114 |
+
},
|
| 115 |
+
{
|
| 116 |
+
"epoch": 1.61,
|
| 117 |
+
"learning_rate": 2.5618374558303885e-05,
|
| 118 |
+
"loss": 0.1107,
|
| 119 |
+
"step": 170
|
| 120 |
+
},
|
| 121 |
+
{
|
| 122 |
+
"epoch": 1.71,
|
| 123 |
+
"learning_rate": 2.3851590106007068e-05,
|
| 124 |
+
"loss": 0.0231,
|
| 125 |
+
"step": 180
|
| 126 |
+
},
|
| 127 |
+
{
|
| 128 |
+
"epoch": 1.8,
|
| 129 |
+
"learning_rate": 2.2084805653710246e-05,
|
| 130 |
+
"loss": 0.0533,
|
| 131 |
+
"step": 190
|
| 132 |
+
},
|
| 133 |
+
{
|
| 134 |
+
"epoch": 1.9,
|
| 135 |
+
"learning_rate": 2.031802120141343e-05,
|
| 136 |
+
"loss": 0.0713,
|
| 137 |
+
"step": 200
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.99,
|
| 141 |
+
"learning_rate": 1.855123674911661e-05,
|
| 142 |
+
"loss": 0.1087,
|
| 143 |
+
"step": 210
|
| 144 |
+
},
|
| 145 |
+
{
|
| 146 |
+
"epoch": 2.0,
|
| 147 |
+
"eval_accuracy": 0.992,
|
| 148 |
+
"eval_loss": 0.017423858866095543,
|
| 149 |
+
"eval_runtime": 5.2579,
|
| 150 |
+
"eval_samples_per_second": 71.321,
|
| 151 |
+
"eval_steps_per_second": 8.939,
|
| 152 |
+
"step": 211
|
| 153 |
+
},
|
| 154 |
+
{
|
| 155 |
+
"epoch": 2.09,
|
| 156 |
+
"learning_rate": 1.678445229681979e-05,
|
| 157 |
+
"loss": 0.0626,
|
| 158 |
+
"step": 220
|
| 159 |
+
},
|
| 160 |
+
{
|
| 161 |
+
"epoch": 2.18,
|
| 162 |
+
"learning_rate": 1.501766784452297e-05,
|
| 163 |
+
"loss": 0.0639,
|
| 164 |
+
"step": 230
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"epoch": 2.27,
|
| 168 |
+
"learning_rate": 1.3250883392226149e-05,
|
| 169 |
+
"loss": 0.023,
|
| 170 |
+
"step": 240
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"epoch": 2.37,
|
| 174 |
+
"learning_rate": 1.148409893992933e-05,
|
| 175 |
+
"loss": 0.0404,
|
| 176 |
+
"step": 250
|
| 177 |
+
},
|
| 178 |
+
{
|
| 179 |
+
"epoch": 2.46,
|
| 180 |
+
"learning_rate": 9.717314487632508e-06,
|
| 181 |
+
"loss": 0.0645,
|
| 182 |
+
"step": 260
|
| 183 |
+
},
|
| 184 |
+
{
|
| 185 |
+
"epoch": 2.56,
|
| 186 |
+
"learning_rate": 7.950530035335689e-06,
|
| 187 |
+
"loss": 0.0322,
|
| 188 |
+
"step": 270
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"epoch": 2.65,
|
| 192 |
+
"learning_rate": 6.18374558303887e-06,
|
| 193 |
+
"loss": 0.0505,
|
| 194 |
+
"step": 280
|
| 195 |
+
},
|
| 196 |
+
{
|
| 197 |
+
"epoch": 2.75,
|
| 198 |
+
"learning_rate": 4.41696113074205e-06,
|
| 199 |
+
"loss": 0.0103,
|
| 200 |
+
"step": 290
|
| 201 |
+
},
|
| 202 |
+
{
|
| 203 |
+
"epoch": 2.84,
|
| 204 |
+
"learning_rate": 2.65017667844523e-06,
|
| 205 |
+
"loss": 0.0503,
|
| 206 |
+
"step": 300
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 2.94,
|
| 210 |
+
"learning_rate": 8.8339222614841e-07,
|
| 211 |
+
"loss": 0.0131,
|
| 212 |
+
"step": 310
|
| 213 |
+
},
|
| 214 |
+
{
|
| 215 |
+
"epoch": 2.99,
|
| 216 |
+
"eval_accuracy": 0.9973333333333333,
|
| 217 |
+
"eval_loss": 0.005204513669013977,
|
| 218 |
+
"eval_runtime": 4.0686,
|
| 219 |
+
"eval_samples_per_second": 92.17,
|
| 220 |
+
"eval_steps_per_second": 11.552,
|
| 221 |
+
"step": 315
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 2.99,
|
| 225 |
+
"step": 315,
|
| 226 |
+
"total_flos": 2.5044894502148506e+17,
|
| 227 |
+
"train_loss": 0.10708669147321156,
|
| 228 |
+
"train_runtime": 1544.9876,
|
| 229 |
+
"train_samples_per_second": 6.552,
|
| 230 |
+
"train_steps_per_second": 0.204
|
| 231 |
+
}
|
| 232 |
+
],
|
| 233 |
+
"max_steps": 315,
|
| 234 |
+
"num_train_epochs": 3,
|
| 235 |
+
"total_flos": 2.5044894502148506e+17,
|
| 236 |
+
"trial_name": null,
|
| 237 |
+
"trial_params": null
|
| 238 |
+
}
|