Add new SentenceTransformer model.

README.md

@@ -46,7 +46,7 @@ tags:
 - feature-extraction
 - generated_from_trainer
 - dataset_size:356
-- loss:OnlineContrastiveLoss
+- loss:CoSENTLoss
 widget:
 - source_sentence: これって？
   sentences:
@@ -84,109 +84,109 @@ model-index:
       type: custom-arc-semantics-data
     metrics:
     - type: cosine_accuracy
-      value: 0.9213483146067416
+      value: 0.9550561797752809
       name: Cosine Accuracy
     - type: cosine_accuracy_threshold
-      value: 0.9257684946060181
+      value: 0.5568578243255615
       name: Cosine Accuracy Threshold
     - type: cosine_f1
-      value: 0.9380530973451328
+      value: 0.9655172413793103
       name: Cosine F1
     - type: cosine_f1_threshold
-      value: 0.9257684946060181
+      value: 0.5568578243255615
       name: Cosine F1 Threshold
     - type: cosine_precision
-      value: 0.9814814814814815
+      value: 0.9824561403508771
       name: Cosine Precision
     - type: cosine_recall
-      value: 0.8983050847457628
+      value: 0.9491525423728814
       name: Cosine Recall
     - type: cosine_ap
-      value: 0.9831776961455067
+      value: 0.9932329299017532
       name: Cosine Ap
     - type: dot_accuracy
-      value: 0.898876404494382
+      value: 0.9438202247191011
       name: Dot Accuracy
     - type: dot_accuracy_threshold
-      value: 549.65966796875
+      value: 281.24676513671875
       name: Dot Accuracy Threshold
     - type: dot_f1
-      value: 0.9279999999999999
+      value: 0.957983193277311
       name: Dot F1
     - type: dot_f1_threshold
-      value: 536.9278564453125
+      value: 240.45741271972656
       name: Dot F1 Threshold
     - type: dot_precision
-      value: 0.8787878787878788
+      value: 0.95
       name: Dot Precision
     - type: dot_recall
-      value: 0.9830508474576272
+      value: 0.9661016949152542
       name: Dot Recall
     - type: dot_ap
-      value: 0.9799216592268227
+      value: 0.992060744461618
       name: Dot Ap
     - type: manhattan_accuracy
-      value: 0.9213483146067416
+      value: 0.9550561797752809
       name: Manhattan Accuracy
     - type: manhattan_accuracy_threshold
-      value: 212.17135620117188
+      value: 468.22576904296875
       name: Manhattan Accuracy Threshold
     - type: manhattan_f1
-      value: 0.9380530973451328
+      value: 0.9655172413793103
       name: Manhattan F1
     - type: manhattan_f1_threshold
-      value: 212.17135620117188
+      value: 486.80523681640625
       name: Manhattan F1 Threshold
     - type: manhattan_precision
-      value: 0.9814814814814815
+      value: 0.9824561403508771
       name: Manhattan Precision
     - type: manhattan_recall
-      value: 0.8983050847457628
+      value: 0.9491525423728814
       name: Manhattan Recall
     - type: manhattan_ap
-      value: 0.9831776961455067
+      value: 0.9937064750898389
       name: Manhattan Ap
     - type: euclidean_accuracy
-      value: 0.9213483146067416
+      value: 0.9550561797752809
       name: Euclidean Accuracy
     - type: euclidean_accuracy_threshold
-      value: 9.646502494812012
+      value: 21.117210388183594
       name: Euclidean Accuracy Threshold
     - type: euclidean_f1
-      value: 0.9380530973451328
+      value: 0.9655172413793103
       name: Euclidean F1
     - type: euclidean_f1_threshold
-      value: 9.646502494812012
+      value: 21.95305633544922
       name: Euclidean F1 Threshold
     - type: euclidean_precision
-      value: 0.9814814814814815
+      value: 0.9824561403508771
       name: Euclidean Precision
     - type: euclidean_recall
-      value: 0.8983050847457628
+      value: 0.9491525423728814
       name: Euclidean Recall
     - type: euclidean_ap
-      value: 0.9831776961455067
+      value: 0.9933690931735095
       name: Euclidean Ap
     - type: max_accuracy
-      value: 0.9213483146067416
+      value: 0.9550561797752809
       name: Max Accuracy
     - type: max_accuracy_threshold
-      value: 549.65966796875
+      value: 468.22576904296875
       name: Max Accuracy Threshold
     - type: max_f1
-      value: 0.9380530973451328
+      value: 0.9655172413793103
       name: Max F1
     - type: max_f1_threshold
-      value: 536.9278564453125
+      value: 486.80523681640625
       name: Max F1 Threshold
     - type: max_precision
-      value: 0.9814814814814815
+      value: 0.9824561403508771
       name: Max Precision
     - type: max_recall
-      value: 0.9830508474576272
+      value: 0.9661016949152542
       name: Max Recall
     - type: max_ap
-      value: 0.9831776961455067
+      value: 0.9937064750898389
       name: Max Ap
 ---
 
@@ -287,41 +287,41 @@ You can finetune this model on your own dataset.
 
 | Metric                       | Value      |
 |:-----------------------------|:-----------|
-| cosine_accuracy              | 0.9213     |
-| cosine_accuracy_threshold    | 0.9258     |
-| cosine_f1                    | 0.9381     |
-| cosine_f1_threshold          | 0.9258     |
-| cosine_precision             | 0.9815     |
-| cosine_recall                | 0.8983     |
-| cosine_ap                    | 0.9832     |
-| dot_accuracy                 | 0.8989     |
-| dot_accuracy_threshold       | 549.6597   |
-| dot_f1                       | 0.928      |
-| dot_f1_threshold             | 536.9279   |
-| dot_precision                | 0.8788     |
-| dot_recall                   | 0.9831     |
-| dot_ap                       | 0.9799     |
-| manhattan_accuracy           | 0.9213     |
-| manhattan_accuracy_threshold | 212.1714   |
-| manhattan_f1                 | 0.9381     |
-| manhattan_f1_threshold       | 212.1714   |
-| manhattan_precision          | 0.9815     |
-| manhattan_recall             | 0.8983     |
-| manhattan_ap                 | 0.9832     |
-| euclidean_accuracy           | 0.9213     |
-| euclidean_accuracy_threshold | 9.6465     |
-| euclidean_f1                 | 0.9381     |
-| euclidean_f1_threshold       | 9.6465     |
-| euclidean_precision          | 0.9815     |
-| euclidean_recall             | 0.8983     |
-| euclidean_ap                 | 0.9832     |
-| max_accuracy                 | 0.9213     |
-| max_accuracy_threshold       | 549.6597   |
-| max_f1                       | 0.9381     |
-| max_f1_threshold             | 536.9279   |
-| max_precision                | 0.9815     |
-| max_recall                   | 0.9831     |
-| **max_ap**                   | **0.9832** |
+| cosine_accuracy              | 0.9551     |
+| cosine_accuracy_threshold    | 0.5569     |
+| cosine_f1                    | 0.9655     |
+| cosine_f1_threshold          | 0.5569     |
+| cosine_precision             | 0.9825     |
+| cosine_recall                | 0.9492     |
+| cosine_ap                    | 0.9932     |
+| dot_accuracy                 | 0.9438     |
+| dot_accuracy_threshold       | 281.2468   |
+| dot_f1                       | 0.958      |
+| dot_f1_threshold             | 240.4574   |
+| dot_precision                | 0.95       |
+| dot_recall                   | 0.9661     |
+| dot_ap                       | 0.9921     |
+| manhattan_accuracy           | 0.9551     |
+| manhattan_accuracy_threshold | 468.2258   |
+| manhattan_f1                 | 0.9655     |
+| manhattan_f1_threshold       | 486.8052   |
+| manhattan_precision          | 0.9825     |
+| manhattan_recall             | 0.9492     |
+| manhattan_ap                 | 0.9937     |
+| euclidean_accuracy           | 0.9551     |
+| euclidean_accuracy_threshold | 21.1172    |
+| euclidean_f1                 | 0.9655     |
+| euclidean_f1_threshold       | 21.9531    |
+| euclidean_precision          | 0.9825     |
+| euclidean_recall             | 0.9492     |
+| euclidean_ap                 | 0.9934     |
+| max_accuracy                 | 0.9551     |
+| max_accuracy_threshold       | 468.2258   |
+| max_f1                       | 0.9655     |
+| max_f1_threshold             | 486.8052   |
+| max_precision                | 0.9825     |
+| max_recall                   | 0.9661     |
+| **max_ap**                   | **0.9937** |
 
 <!--
 ## Bias, Risks and Limitations
@@ -355,7 +355,13 @@ You can finetune this model on your own dataset.
   | <code>ジャックはどんな魔法を使うの？</code> | <code>見た目を変える魔法</code>       | <code>0</code> |
   | <code>魔法使い</code>            | <code>魔法をかけられる人</code>       | <code>1</code> |
   | <code>ぬいぐるみが花</code>         | <code>花がぬいぐるみに変えられている</code> | <code>1</code> |
-* Loss: [<code>OnlineContrastiveLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#onlinecontrastiveloss)
+* Loss: [<code>CoSENTLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#cosentloss) with these parameters:
+  ```json
+  {
+      "scale": 20.0,
+      "similarity_fct": "pairwise_cos_sim"
+  }
+  ```
 
 ### Evaluation Dataset
 
@@ -375,7 +381,13 @@ You can finetune this model on your own dataset.
   | <code>トーチ</code>             | <code>なにも要らない</code>         | <code>0</code> |
   | <code>家の外</code>             | <code>家の外へ行こう</code>         | <code>1</code> |
   | <code>お皿に赤い染みがついていたから</code> | <code>棚からトマトがなくなってたから</code> | <code>0</code> |
-* Loss: [<code>OnlineContrastiveLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#onlinecontrastiveloss)
+* Loss: [<code>CoSENTLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#cosentloss) with these parameters:
+  ```json
+  {
+      "scale": 20.0,
+      "similarity_fct": "pairwise_cos_sim"
+  }
+  ```
 
 ### Training Hyperparameters
 #### Non-Default Hyperparameters
@@ -508,19 +520,19 @@ You can finetune this model on your own dataset.
 | Epoch | Step | Training Loss | loss   | custom-arc-semantics-data_max_ap |
 |:-----:|:----:|:-------------:|:------:|:--------------------------------:|
 | None  | 0    | -             | -      | 0.9511                           |
-| 1.0   | 45   | 0.2266        | 0.1131 | 0.9630                           |
-| 2.0   | 90   | 0.0847        | 0.1191 | 0.9745                           |
-| 3.0   | 135  | 0.0227        | 0.0873 | 0.9817                           |
-| 4.0   | 180  | 0.0144        | 0.0887 | 0.9843                           |
-| 5.0   | 225  | 0.006         | 0.0750 | 0.9845                           |
-| 6.0   | 270  | 0.0016        | 0.0920 | 0.9842                           |
-| 7.0   | 315  | 0.0009        | 0.0943 | 0.9837                           |
-| 8.0   | 360  | 0.0044        | 0.0790 | 0.9852                           |
-| 9.0   | 405  | 0.0046        | 0.0772 | 0.9856                           |
-| 10.0  | 450  | 0.005         | 0.0766 | 0.9848                           |
-| 11.0  | 495  | 0.0008        | 0.0766 | 0.9835                           |
-| 12.0  | 540  | 0.0046        | 0.0767 | 0.9835                           |
-| 13.0  | 585  | 0.0008        | 0.0770 | 0.9832                           |
+| 1.0   | 45   | 1.9903        | 1.1863 | 0.9765                           |
+| 2.0   | 90   | 0.8198        | 1.0991 | 0.9873                           |
+| 3.0   | 135  | 0.0806        | 0.9033 | 0.9914                           |
+| 4.0   | 180  | 0.0024        | 0.7569 | 0.9930                           |
+| 5.0   | 225  | 0.0002        | 0.7598 | 0.9937                           |
+| 6.0   | 270  | 0.0001        | 0.7418 | 0.9937                           |
+| 7.0   | 315  | 0.0001        | 0.7322 | 0.9937                           |
+| 8.0   | 360  | 0.0001        | 0.7269 | 0.9937                           |
+| 9.0   | 405  | 0.0001        | 0.7277 | 0.9937                           |
+| 10.0  | 450  | 0.0001        | 0.7289 | 0.9937                           |
+| 11.0  | 495  | 0.0           | 0.7301 | 0.9937                           |
+| 12.0  | 540  | 0.0001        | 0.7299 | 0.9937                           |
+| 13.0  | 585  | 0.0001        | 0.7296 | 0.9937                           |
 
 
 ### Framework Versions
@@ -549,6 +561,17 @@ You can finetune this model on your own dataset.
 }
 ```
 
+#### CoSENTLoss
+```bibtex
+@online{kexuefm-8847,
+    title={CoSENT: A more efficient sentence vector scheme than Sentence-BERT},
+    author={Su Jianlin},
+    year={2022},
+    month={Jan},
+    url={https://kexue.fm/archives/8847},
+}
+```
+
 <!--
 ## Glossary
 

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:656fa6a91d2f1ef8b5068cb095db7dc616a54abbbe033db496b7278e2090f78d
+oid sha256:58c3dab56e3b4f32c43626942ffeec0674e7b11178e1c66e38f95f95fa629978
 size 442491744