n_layers_per_step = 1, last_layer_weight = 1 * model_layers,, prior_layers_weight= 0.05, kl_div_weight = 2, kl_temperature= 0.9,
Browse files- README.md +104 -104
- pytorch_model.bin +1 -1
README.md
CHANGED
|
@@ -89,109 +89,109 @@ model-index:
|
|
| 89 |
type: unknown
|
| 90 |
metrics:
|
| 91 |
- type: cosine_accuracy
|
| 92 |
-
value: 0.
|
| 93 |
name: Cosine Accuracy
|
| 94 |
- type: cosine_accuracy_threshold
|
| 95 |
-
value: 0.
|
| 96 |
name: Cosine Accuracy Threshold
|
| 97 |
- type: cosine_f1
|
| 98 |
-
value: 0.
|
| 99 |
name: Cosine F1
|
| 100 |
- type: cosine_f1_threshold
|
| 101 |
-
value: 0.
|
| 102 |
name: Cosine F1 Threshold
|
| 103 |
- type: cosine_precision
|
| 104 |
-
value: 0.
|
| 105 |
name: Cosine Precision
|
| 106 |
- type: cosine_recall
|
| 107 |
-
value: 0.
|
| 108 |
name: Cosine Recall
|
| 109 |
- type: cosine_ap
|
| 110 |
-
value: 0.
|
| 111 |
name: Cosine Ap
|
| 112 |
- type: dot_accuracy
|
| 113 |
-
value: 0.
|
| 114 |
name: Dot Accuracy
|
| 115 |
- type: dot_accuracy_threshold
|
| 116 |
-
value:
|
| 117 |
name: Dot Accuracy Threshold
|
| 118 |
- type: dot_f1
|
| 119 |
-
value: 0.
|
| 120 |
name: Dot F1
|
| 121 |
- type: dot_f1_threshold
|
| 122 |
-
value:
|
| 123 |
name: Dot F1 Threshold
|
| 124 |
- type: dot_precision
|
| 125 |
-
value: 0.
|
| 126 |
name: Dot Precision
|
| 127 |
- type: dot_recall
|
| 128 |
-
value: 0.
|
| 129 |
name: Dot Recall
|
| 130 |
- type: dot_ap
|
| 131 |
-
value: 0.
|
| 132 |
name: Dot Ap
|
| 133 |
- type: manhattan_accuracy
|
| 134 |
-
value: 0.
|
| 135 |
name: Manhattan Accuracy
|
| 136 |
- type: manhattan_accuracy_threshold
|
| 137 |
-
value:
|
| 138 |
name: Manhattan Accuracy Threshold
|
| 139 |
- type: manhattan_f1
|
| 140 |
-
value: 0.
|
| 141 |
name: Manhattan F1
|
| 142 |
- type: manhattan_f1_threshold
|
| 143 |
-
value:
|
| 144 |
name: Manhattan F1 Threshold
|
| 145 |
- type: manhattan_precision
|
| 146 |
-
value: 0.
|
| 147 |
name: Manhattan Precision
|
| 148 |
- type: manhattan_recall
|
| 149 |
-
value: 0.
|
| 150 |
name: Manhattan Recall
|
| 151 |
- type: manhattan_ap
|
| 152 |
-
value: 0.
|
| 153 |
name: Manhattan Ap
|
| 154 |
- type: euclidean_accuracy
|
| 155 |
-
value: 0.
|
| 156 |
name: Euclidean Accuracy
|
| 157 |
- type: euclidean_accuracy_threshold
|
| 158 |
-
value:
|
| 159 |
name: Euclidean Accuracy Threshold
|
| 160 |
- type: euclidean_f1
|
| 161 |
-
value: 0.
|
| 162 |
name: Euclidean F1
|
| 163 |
- type: euclidean_f1_threshold
|
| 164 |
-
value:
|
| 165 |
name: Euclidean F1 Threshold
|
| 166 |
- type: euclidean_precision
|
| 167 |
-
value: 0.
|
| 168 |
name: Euclidean Precision
|
| 169 |
- type: euclidean_recall
|
| 170 |
-
value: 0.
|
| 171 |
name: Euclidean Recall
|
| 172 |
- type: euclidean_ap
|
| 173 |
-
value: 0.
|
| 174 |
name: Euclidean Ap
|
| 175 |
- type: max_accuracy
|
| 176 |
-
value: 0.
|
| 177 |
name: Max Accuracy
|
| 178 |
- type: max_accuracy_threshold
|
| 179 |
-
value:
|
| 180 |
name: Max Accuracy Threshold
|
| 181 |
- type: max_f1
|
| 182 |
-
value: 0.
|
| 183 |
name: Max F1
|
| 184 |
- type: max_f1_threshold
|
| 185 |
-
value:
|
| 186 |
name: Max F1 Threshold
|
| 187 |
- type: max_precision
|
| 188 |
-
value: 0.
|
| 189 |
name: Max Precision
|
| 190 |
- type: max_recall
|
| 191 |
-
value: 0.
|
| 192 |
name: Max Recall
|
| 193 |
- type: max_ap
|
| 194 |
-
value: 0.
|
| 195 |
name: Max Ap
|
| 196 |
---
|
| 197 |
|
|
@@ -293,41 +293,41 @@ You can finetune this model on your own dataset.
|
|
| 293 |
|
| 294 |
| Metric | Value |
|
| 295 |
|:-----------------------------|:-----------|
|
| 296 |
-
| cosine_accuracy | 0.
|
| 297 |
-
| cosine_accuracy_threshold | 0.
|
| 298 |
-
| cosine_f1 | 0.
|
| 299 |
-
| cosine_f1_threshold | 0.
|
| 300 |
-
| cosine_precision | 0.
|
| 301 |
-
| cosine_recall | 0.
|
| 302 |
-
| cosine_ap | 0.
|
| 303 |
-
| dot_accuracy | 0.
|
| 304 |
-
| dot_accuracy_threshold |
|
| 305 |
-
| dot_f1 | 0.
|
| 306 |
-
| dot_f1_threshold |
|
| 307 |
-
| dot_precision | 0.
|
| 308 |
-
| dot_recall | 0.
|
| 309 |
-
| dot_ap | 0.
|
| 310 |
-
| manhattan_accuracy | 0.
|
| 311 |
-
| manhattan_accuracy_threshold |
|
| 312 |
-
| manhattan_f1 | 0.
|
| 313 |
-
| manhattan_f1_threshold |
|
| 314 |
-
| manhattan_precision | 0.
|
| 315 |
-
| manhattan_recall | 0.
|
| 316 |
-
| manhattan_ap | 0.
|
| 317 |
-
| euclidean_accuracy | 0.
|
| 318 |
-
| euclidean_accuracy_threshold |
|
| 319 |
-
| euclidean_f1 | 0.
|
| 320 |
-
| euclidean_f1_threshold |
|
| 321 |
-
| euclidean_precision | 0.
|
| 322 |
-
| euclidean_recall | 0.
|
| 323 |
-
| euclidean_ap | 0.
|
| 324 |
-
| max_accuracy | 0.
|
| 325 |
-
| max_accuracy_threshold |
|
| 326 |
-
| max_f1 | 0.
|
| 327 |
-
| max_f1_threshold |
|
| 328 |
-
| max_precision | 0.
|
| 329 |
-
| max_recall | 0.
|
| 330 |
-
| **max_ap** | **0.
|
| 331 |
|
| 332 |
<!--
|
| 333 |
## Bias, Risks and Limitations
|
|
@@ -366,10 +366,10 @@ You can finetune this model on your own dataset.
|
|
| 366 |
{
|
| 367 |
"loss": "MultipleNegativesRankingLoss",
|
| 368 |
"n_layers_per_step": 1,
|
| 369 |
-
"last_layer_weight": 1
|
| 370 |
-
"prior_layers_weight":
|
| 371 |
"kl_div_weight": 2,
|
| 372 |
-
"kl_temperature":
|
| 373 |
}
|
| 374 |
```
|
| 375 |
|
|
@@ -396,10 +396,10 @@ You can finetune this model on your own dataset.
|
|
| 396 |
{
|
| 397 |
"loss": "MultipleNegativesRankingLoss",
|
| 398 |
"n_layers_per_step": 1,
|
| 399 |
-
"last_layer_weight": 1
|
| 400 |
-
"prior_layers_weight":
|
| 401 |
"kl_div_weight": 2,
|
| 402 |
-
"kl_temperature":
|
| 403 |
}
|
| 404 |
```
|
| 405 |
|
|
@@ -538,34 +538,34 @@ You can finetune this model on your own dataset.
|
|
| 538 |
### Training Logs
|
| 539 |
| Epoch | Step | Training Loss | loss | max_ap |
|
| 540 |
|:------:|:----:|:-------------:|:------:|:------:|
|
| 541 |
-
| 0.1004 | 150 |
|
| 542 |
-
| 0.2001 | 299 | - |
|
| 543 |
-
| 0.2008 | 300 |
|
| 544 |
-
| 0.3012 | 450 |
|
| 545 |
-
| 0.4003 | 598 | - |
|
| 546 |
-
| 0.4016 | 600 |
|
| 547 |
-
| 0.5020 | 750 |
|
| 548 |
-
| 0.6004 | 897 | - |
|
| 549 |
-
| 0.6024 | 900 |
|
| 550 |
-
| 0.7028 | 1050 |
|
| 551 |
-
| 0.8005 | 1196 | - |
|
| 552 |
-
| 0.8032 | 1200 |
|
| 553 |
-
| 0.9036 | 1350 |
|
| 554 |
-
| 1.0007 | 1495 | - |
|
| 555 |
-
| 1.0040 | 1500 |
|
| 556 |
-
| 1.1044 | 1650 |
|
| 557 |
-
| 1.2008 | 1794 | - |
|
| 558 |
-
| 1.2048 | 1800 |
|
| 559 |
-
| 1.3052 | 1950 |
|
| 560 |
-
| 1.4009 | 2093 | - |
|
| 561 |
-
| 1.4056 | 2100 |
|
| 562 |
-
| 1.5060 | 2250 |
|
| 563 |
-
| 1.6011 | 2392 | - |
|
| 564 |
-
| 1.6064 | 2400 |
|
| 565 |
-
| 1.7068 | 2550 |
|
| 566 |
-
| 1.8012 | 2691 | - |
|
| 567 |
-
| 1.8072 | 2700 |
|
| 568 |
-
| 1.9076 | 2850 |
|
| 569 |
|
| 570 |
|
| 571 |
### Framework Versions
|
|
|
|
| 89 |
type: unknown
|
| 90 |
metrics:
|
| 91 |
- type: cosine_accuracy
|
| 92 |
+
value: 0.6651071536371869
|
| 93 |
name: Cosine Accuracy
|
| 94 |
- type: cosine_accuracy_threshold
|
| 95 |
+
value: 0.687929630279541
|
| 96 |
name: Cosine Accuracy Threshold
|
| 97 |
- type: cosine_f1
|
| 98 |
+
value: 0.7077349458301839
|
| 99 |
name: Cosine F1
|
| 100 |
- type: cosine_f1_threshold
|
| 101 |
+
value: 0.6304811239242554
|
| 102 |
name: Cosine F1 Threshold
|
| 103 |
- type: cosine_precision
|
| 104 |
+
value: 0.6222862206468763
|
| 105 |
name: Cosine Precision
|
| 106 |
- type: cosine_recall
|
| 107 |
+
value: 0.8203855140186916
|
| 108 |
name: Cosine Recall
|
| 109 |
- type: cosine_ap
|
| 110 |
+
value: 0.7058220689813709
|
| 111 |
name: Cosine Ap
|
| 112 |
- type: dot_accuracy
|
| 113 |
+
value: 0.6313009357078176
|
| 114 |
name: Dot Accuracy
|
| 115 |
- type: dot_accuracy_threshold
|
| 116 |
+
value: 135.98495483398438
|
| 117 |
name: Dot Accuracy Threshold
|
| 118 |
- type: dot_f1
|
| 119 |
+
value: 0.6997334569475027
|
| 120 |
name: Dot F1
|
| 121 |
- type: dot_f1_threshold
|
| 122 |
+
value: 115.54609680175781
|
| 123 |
name: Dot F1 Threshold
|
| 124 |
- type: dot_precision
|
| 125 |
+
value: 0.5800192122958694
|
| 126 |
name: Dot Precision
|
| 127 |
- type: dot_recall
|
| 128 |
+
value: 0.8817172897196262
|
| 129 |
name: Dot Recall
|
| 130 |
- type: dot_ap
|
| 131 |
+
value: 0.6554755795160082
|
| 132 |
name: Dot Ap
|
| 133 |
- type: manhattan_accuracy
|
| 134 |
+
value: 0.6708421370359191
|
| 135 |
name: Manhattan Accuracy
|
| 136 |
- type: manhattan_accuracy_threshold
|
| 137 |
+
value: 219.32388305664062
|
| 138 |
name: Manhattan Accuracy Threshold
|
| 139 |
- type: manhattan_f1
|
| 140 |
+
value: 0.7119951778179626
|
| 141 |
name: Manhattan F1
|
| 142 |
- type: manhattan_f1_threshold
|
| 143 |
+
value: 262.314697265625
|
| 144 |
name: Manhattan F1 Threshold
|
| 145 |
- type: manhattan_precision
|
| 146 |
+
value: 0.6062410182714022
|
| 147 |
name: Manhattan Precision
|
| 148 |
- type: manhattan_recall
|
| 149 |
+
value: 0.8624415887850467
|
| 150 |
name: Manhattan Recall
|
| 151 |
- type: manhattan_ap
|
| 152 |
+
value: 0.7135236162968746
|
| 153 |
name: Manhattan Ap
|
| 154 |
- type: euclidean_accuracy
|
| 155 |
+
value: 0.6652580742529429
|
| 156 |
name: Euclidean Accuracy
|
| 157 |
- type: euclidean_accuracy_threshold
|
| 158 |
+
value: 11.506816864013672
|
| 159 |
name: Euclidean Accuracy Threshold
|
| 160 |
- type: euclidean_f1
|
| 161 |
+
value: 0.7080090384132564
|
| 162 |
name: Euclidean F1
|
| 163 |
- type: euclidean_f1_threshold
|
| 164 |
+
value: 12.478536605834961
|
| 165 |
name: Euclidean F1 Threshold
|
| 166 |
- type: euclidean_precision
|
| 167 |
+
value: 0.6208718626155878
|
| 168 |
name: Euclidean Precision
|
| 169 |
- type: euclidean_recall
|
| 170 |
+
value: 0.8235981308411215
|
| 171 |
name: Euclidean Recall
|
| 172 |
- type: euclidean_ap
|
| 173 |
+
value: 0.7090362803652147
|
| 174 |
name: Euclidean Ap
|
| 175 |
- type: max_accuracy
|
| 176 |
+
value: 0.6708421370359191
|
| 177 |
name: Max Accuracy
|
| 178 |
- type: max_accuracy_threshold
|
| 179 |
+
value: 219.32388305664062
|
| 180 |
name: Max Accuracy Threshold
|
| 181 |
- type: max_f1
|
| 182 |
+
value: 0.7119951778179626
|
| 183 |
name: Max F1
|
| 184 |
- type: max_f1_threshold
|
| 185 |
+
value: 262.314697265625
|
| 186 |
name: Max F1 Threshold
|
| 187 |
- type: max_precision
|
| 188 |
+
value: 0.6222862206468763
|
| 189 |
name: Max Precision
|
| 190 |
- type: max_recall
|
| 191 |
+
value: 0.8817172897196262
|
| 192 |
name: Max Recall
|
| 193 |
- type: max_ap
|
| 194 |
+
value: 0.7135236162968746
|
| 195 |
name: Max Ap
|
| 196 |
---
|
| 197 |
|
|
|
|
| 293 |
|
| 294 |
| Metric | Value |
|
| 295 |
|:-----------------------------|:-----------|
|
| 296 |
+
| cosine_accuracy | 0.6651 |
|
| 297 |
+
| cosine_accuracy_threshold | 0.6879 |
|
| 298 |
+
| cosine_f1 | 0.7077 |
|
| 299 |
+
| cosine_f1_threshold | 0.6305 |
|
| 300 |
+
| cosine_precision | 0.6223 |
|
| 301 |
+
| cosine_recall | 0.8204 |
|
| 302 |
+
| cosine_ap | 0.7058 |
|
| 303 |
+
| dot_accuracy | 0.6313 |
|
| 304 |
+
| dot_accuracy_threshold | 135.985 |
|
| 305 |
+
| dot_f1 | 0.6997 |
|
| 306 |
+
| dot_f1_threshold | 115.5461 |
|
| 307 |
+
| dot_precision | 0.58 |
|
| 308 |
+
| dot_recall | 0.8817 |
|
| 309 |
+
| dot_ap | 0.6555 |
|
| 310 |
+
| manhattan_accuracy | 0.6708 |
|
| 311 |
+
| manhattan_accuracy_threshold | 219.3239 |
|
| 312 |
+
| manhattan_f1 | 0.712 |
|
| 313 |
+
| manhattan_f1_threshold | 262.3147 |
|
| 314 |
+
| manhattan_precision | 0.6062 |
|
| 315 |
+
| manhattan_recall | 0.8624 |
|
| 316 |
+
| manhattan_ap | 0.7135 |
|
| 317 |
+
| euclidean_accuracy | 0.6653 |
|
| 318 |
+
| euclidean_accuracy_threshold | 11.5068 |
|
| 319 |
+
| euclidean_f1 | 0.708 |
|
| 320 |
+
| euclidean_f1_threshold | 12.4785 |
|
| 321 |
+
| euclidean_precision | 0.6209 |
|
| 322 |
+
| euclidean_recall | 0.8236 |
|
| 323 |
+
| euclidean_ap | 0.709 |
|
| 324 |
+
| max_accuracy | 0.6708 |
|
| 325 |
+
| max_accuracy_threshold | 219.3239 |
|
| 326 |
+
| max_f1 | 0.712 |
|
| 327 |
+
| max_f1_threshold | 262.3147 |
|
| 328 |
+
| max_precision | 0.6223 |
|
| 329 |
+
| max_recall | 0.8817 |
|
| 330 |
+
| **max_ap** | **0.7135** |
|
| 331 |
|
| 332 |
<!--
|
| 333 |
## Bias, Risks and Limitations
|
|
|
|
| 366 |
{
|
| 367 |
"loss": "MultipleNegativesRankingLoss",
|
| 368 |
"n_layers_per_step": 1,
|
| 369 |
+
"last_layer_weight": 1,
|
| 370 |
+
"prior_layers_weight": 0.05,
|
| 371 |
"kl_div_weight": 2,
|
| 372 |
+
"kl_temperature": 0.9
|
| 373 |
}
|
| 374 |
```
|
| 375 |
|
|
|
|
| 396 |
{
|
| 397 |
"loss": "MultipleNegativesRankingLoss",
|
| 398 |
"n_layers_per_step": 1,
|
| 399 |
+
"last_layer_weight": 1,
|
| 400 |
+
"prior_layers_weight": 0.05,
|
| 401 |
"kl_div_weight": 2,
|
| 402 |
+
"kl_temperature": 0.9
|
| 403 |
}
|
| 404 |
```
|
| 405 |
|
|
|
|
| 538 |
### Training Logs
|
| 539 |
| Epoch | Step | Training Loss | loss | max_ap |
|
| 540 |
|:------:|:----:|:-------------:|:------:|:------:|
|
| 541 |
+
| 0.1004 | 150 | 4.5827 | - | - |
|
| 542 |
+
| 0.2001 | 299 | - | 3.5735 | 0.6133 |
|
| 543 |
+
| 0.2008 | 300 | 3.5451 | - | - |
|
| 544 |
+
| 0.3012 | 450 | 2.9066 | - | - |
|
| 545 |
+
| 0.4003 | 598 | - | 2.8785 | 0.6561 |
|
| 546 |
+
| 0.4016 | 600 | 2.5141 | - | - |
|
| 547 |
+
| 0.5020 | 750 | 2.0248 | - | - |
|
| 548 |
+
| 0.6004 | 897 | - | 2.1300 | 0.6917 |
|
| 549 |
+
| 0.6024 | 900 | 1.6782 | - | - |
|
| 550 |
+
| 0.7028 | 1050 | 1.4187 | - | - |
|
| 551 |
+
| 0.8005 | 1196 | - | 1.7111 | 0.7051 |
|
| 552 |
+
| 0.8032 | 1200 | 1.2446 | - | - |
|
| 553 |
+
| 0.9036 | 1350 | 1.1078 | - | - |
|
| 554 |
+
| 1.0007 | 1495 | - | 1.4859 | 0.7108 |
|
| 555 |
+
| 1.0040 | 1500 | 0.9827 | - | - |
|
| 556 |
+
| 1.1044 | 1650 | 0.9335 | - | - |
|
| 557 |
+
| 1.2008 | 1794 | - | 1.3516 | 0.7121 |
|
| 558 |
+
| 1.2048 | 1800 | 0.8595 | - | - |
|
| 559 |
+
| 1.3052 | 1950 | 0.8362 | - | - |
|
| 560 |
+
| 1.4009 | 2093 | - | 1.2659 | 0.7147 |
|
| 561 |
+
| 1.4056 | 2100 | 0.8167 | - | - |
|
| 562 |
+
| 1.5060 | 2250 | 0.7695 | - | - |
|
| 563 |
+
| 1.6011 | 2392 | - | 1.2218 | 0.7135 |
|
| 564 |
+
| 1.6064 | 2400 | 0.7544 | - | - |
|
| 565 |
+
| 1.7068 | 2550 | 0.7625 | - | - |
|
| 566 |
+
| 1.8012 | 2691 | - | 1.2073 | 0.7135 |
|
| 567 |
+
| 1.8072 | 2700 | 0.7366 | - | - |
|
| 568 |
+
| 1.9076 | 2850 | 0.7348 | - | - |
|
| 569 |
|
| 570 |
|
| 571 |
### Framework Versions
|
pytorch_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 565251810
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d987747af13557ed9ce9d67bddcbcc93242a3bea2f8ca559470e5c3ae052aa72
|
| 3 |
size 565251810
|