LeoChiuu commited on
Commit
23e9052
1 Parent(s): 1f0919d

Add new SentenceTransformer model.

Browse files
README.md CHANGED
@@ -1,201 +1,587 @@
1
  ---
2
  base_model: colorfulscoop/sbert-base-ja
3
- language: ja
4
- license: cc-by-sa-4.0
5
- model_name: LeoChiuu/sbert-base-ja-arc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6
  ---
7
 
8
- # Model Card for LeoChiuu/sbert-base-ja-arc
9
-
10
- <!-- Provide a quick summary of what the model is/does. -->
11
-
12
 
 
13
 
14
  ## Model Details
15
 
16
  ### Model Description
 
 
 
 
 
 
 
 
 
17
 
18
- <!-- Provide a longer summary of what this model is. -->
19
-
20
- Generates similarity embeddings
21
-
22
- - **Developed by:** [More Information Needed]
23
- - **Funded by [optional]:** [More Information Needed]
24
- - **Shared by [optional]:** [More Information Needed]
25
- - **Model type:** [More Information Needed]
26
- - **Language(s) (NLP):** ja
27
- - **License:** cc-by-sa-4.0
28
- - **Finetuned from model [optional]:** colorfulscoop/sbert-base-ja
29
-
30
- ### Model Sources [optional]
31
-
32
- <!-- Provide the basic links for the model. -->
33
-
34
- - **Repository:** [More Information Needed]
35
- - **Paper [optional]:** [More Information Needed]
36
- - **Demo [optional]:** [More Information Needed]
37
-
38
- ## Uses
39
-
40
- <!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
41
 
42
- ### Direct Use
 
 
43
 
44
- <!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
45
 
46
- [More Information Needed]
 
 
 
 
 
47
 
48
- ### Downstream Use [optional]
49
 
50
- <!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
51
 
52
- [More Information Needed]
53
 
54
- ### Out-of-Scope Use
55
-
56
- <!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
57
-
58
- [More Information Needed]
59
 
60
- ## Bias, Risks, and Limitations
 
 
61
 
62
- <!-- This section is meant to convey both technical and sociotechnical limitations. -->
 
 
 
 
 
 
 
 
 
 
63
 
64
- [More Information Needed]
65
-
66
- ### Recommendations
 
 
67
 
68
- <!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
 
69
 
70
- Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
71
 
72
- ## How to Get Started with the Model
73
-
74
- Use the code below to get started with the model.
75
-
76
- [More Information Needed]
77
-
78
- ## Training Details
79
 
80
- ### Training Data
 
81
 
82
- <!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
83
 
84
- [More Information Needed]
85
 
86
- ### Training Procedure
 
87
 
88
- <!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
89
-
90
- #### Preprocessing [optional]
91
-
92
- [More Information Needed]
93
-
94
-
95
- #### Training Hyperparameters
96
-
97
- - **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
98
-
99
- #### Speeds, Sizes, Times [optional]
100
-
101
- <!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
102
 
103
- [More Information Needed]
 
104
 
105
  ## Evaluation
106
 
107
- <!-- This section describes the evaluation protocols and provides the results. -->
108
-
109
- ### Testing Data, Factors & Metrics
110
-
111
- #### Testing Data
112
-
113
- <!-- This should link to a Dataset Card if possible. -->
114
-
115
- [More Information Needed]
116
-
117
- #### Factors
118
-
119
- <!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
120
-
121
- [More Information Needed]
122
-
123
- #### Metrics
124
-
125
- <!-- These are the evaluation metrics being used, ideally with a description of why. -->
126
-
127
- [More Information Needed]
128
-
129
- ### Results
130
-
131
- [More Information Needed]
132
-
133
- #### Summary
134
-
135
-
136
-
137
- ## Model Examination [optional]
138
-
139
- <!-- Relevant interpretability work for the model goes here -->
140
-
141
- [More Information Needed]
142
-
143
- ## Environmental Impact
144
-
145
- <!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
146
-
147
- Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
148
-
149
- - **Hardware Type:** [More Information Needed]
150
- - **Hours used:** [More Information Needed]
151
- - **Cloud Provider:** [More Information Needed]
152
- - **Compute Region:** [More Information Needed]
153
- - **Carbon Emitted:** [More Information Needed]
154
-
155
- ## Technical Specifications [optional]
156
-
157
- ### Model Architecture and Objective
158
-
159
- [More Information Needed]
160
-
161
- ### Compute Infrastructure
162
-
163
- [More Information Needed]
164
-
165
- #### Hardware
166
-
167
- [More Information Needed]
168
-
169
- #### Software
170
-
171
- [More Information Needed]
172
-
173
- ## Citation [optional]
174
-
175
- <!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
176
-
177
- **BibTeX:**
178
-
179
- [More Information Needed]
180
-
181
- **APA:**
182
-
183
- [More Information Needed]
184
-
185
- ## Glossary [optional]
186
-
187
- <!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
188
-
189
- [More Information Needed]
190
-
191
- ## More Information [optional]
192
-
193
- [More Information Needed]
194
 
195
- ## Model Card Authors [optional]
 
196
 
197
- [More Information Needed]
198
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
199
  ## Model Card Contact
200
 
201
- [More Information Needed]
 
 
1
  ---
2
  base_model: colorfulscoop/sbert-base-ja
3
+ library_name: sentence-transformers
4
+ metrics:
5
+ - cosine_accuracy
6
+ - cosine_accuracy_threshold
7
+ - cosine_f1
8
+ - cosine_f1_threshold
9
+ - cosine_precision
10
+ - cosine_recall
11
+ - cosine_ap
12
+ - dot_accuracy
13
+ - dot_accuracy_threshold
14
+ - dot_f1
15
+ - dot_f1_threshold
16
+ - dot_precision
17
+ - dot_recall
18
+ - dot_ap
19
+ - manhattan_accuracy
20
+ - manhattan_accuracy_threshold
21
+ - manhattan_f1
22
+ - manhattan_f1_threshold
23
+ - manhattan_precision
24
+ - manhattan_recall
25
+ - manhattan_ap
26
+ - euclidean_accuracy
27
+ - euclidean_accuracy_threshold
28
+ - euclidean_f1
29
+ - euclidean_f1_threshold
30
+ - euclidean_precision
31
+ - euclidean_recall
32
+ - euclidean_ap
33
+ - max_accuracy
34
+ - max_accuracy_threshold
35
+ - max_f1
36
+ - max_f1_threshold
37
+ - max_precision
38
+ - max_recall
39
+ - max_ap
40
+ pipeline_tag: sentence-similarity
41
+ tags:
42
+ - sentence-transformers
43
+ - sentence-similarity
44
+ - feature-extraction
45
+ - generated_from_trainer
46
+ - dataset_size:680
47
+ - loss:ContrastiveLoss
48
+ widget:
49
+ - source_sentence: 両方はだめ?
50
+ sentences:
51
+ - 両方欲しい
52
+ - あほ
53
+ - キッチンを調べよう
54
+ - source_sentence: どっちも欲しくない
55
+ sentences:
56
+ - 誰かが魔法の呪文で花をぬいぐるみに変えた
57
+ - 呪文を試すため
58
+ - 家の中を調べよう
59
+ - source_sentence: この本は?
60
+ sentences:
61
+ - お鍋から匂いがしたから
62
+ - なんでここに本が?
63
+ - 両方行きたい
64
+ - source_sentence: 他のは選べる?
65
+ sentences:
66
+ - 昨日夕飯にチキンヌードル食べた?
67
+ - 別のは選べる?
68
+ - チキンヌードル作った?
69
+ - source_sentence: 猫のぬいぐるみ
70
+ sentences:
71
+ - 両方はだめ?
72
+ - ぬいぐるみ
73
+ - 夜ご飯を食べる前
74
+ model-index:
75
+ - name: SentenceTransformer based on colorfulscoop/sbert-base-ja
76
+ results:
77
+ - task:
78
+ type: binary-classification
79
+ name: Binary Classification
80
+ dataset:
81
+ name: custom arc semantics data jp
82
+ type: custom-arc-semantics-data-jp
83
+ metrics:
84
+ - type: cosine_accuracy
85
+ value: 0.8897058823529411
86
+ name: Cosine Accuracy
87
+ - type: cosine_accuracy_threshold
88
+ value: 0.6581918001174927
89
+ name: Cosine Accuracy Threshold
90
+ - type: cosine_f1
91
+ value: 0.9044585987261147
92
+ name: Cosine F1
93
+ - type: cosine_f1_threshold
94
+ value: 0.6180122494697571
95
+ name: Cosine F1 Threshold
96
+ - type: cosine_precision
97
+ value: 0.9466666666666667
98
+ name: Cosine Precision
99
+ - type: cosine_recall
100
+ value: 0.8658536585365854
101
+ name: Cosine Recall
102
+ - type: cosine_ap
103
+ value: 0.9692848872766847
104
+ name: Cosine Ap
105
+ - type: dot_accuracy
106
+ value: 0.8897058823529411
107
+ name: Dot Accuracy
108
+ - type: dot_accuracy_threshold
109
+ value: 374.541748046875
110
+ name: Dot Accuracy Threshold
111
+ - type: dot_f1
112
+ value: 0.9019607843137255
113
+ name: Dot F1
114
+ - type: dot_f1_threshold
115
+ value: 374.541748046875
116
+ name: Dot F1 Threshold
117
+ - type: dot_precision
118
+ value: 0.971830985915493
119
+ name: Dot Precision
120
+ - type: dot_recall
121
+ value: 0.8414634146341463
122
+ name: Dot Recall
123
+ - type: dot_ap
124
+ value: 0.9691104975300342
125
+ name: Dot Ap
126
+ - type: manhattan_accuracy
127
+ value: 0.8970588235294118
128
+ name: Manhattan Accuracy
129
+ - type: manhattan_accuracy_threshold
130
+ value: 453.2839660644531
131
+ name: Manhattan Accuracy Threshold
132
+ - type: manhattan_f1
133
+ value: 0.9102564102564101
134
+ name: Manhattan F1
135
+ - type: manhattan_f1_threshold
136
+ value: 453.2839660644531
137
+ name: Manhattan F1 Threshold
138
+ - type: manhattan_precision
139
+ value: 0.9594594594594594
140
+ name: Manhattan Precision
141
+ - type: manhattan_recall
142
+ value: 0.8658536585365854
143
+ name: Manhattan Recall
144
+ - type: manhattan_ap
145
+ value: 0.9687920395428105
146
+ name: Manhattan Ap
147
+ - type: euclidean_accuracy
148
+ value: 0.8897058823529411
149
+ name: Euclidean Accuracy
150
+ - type: euclidean_accuracy_threshold
151
+ value: 19.75204086303711
152
+ name: Euclidean Accuracy Threshold
153
+ - type: euclidean_f1
154
+ value: 0.9047619047619047
155
+ name: Euclidean F1
156
+ - type: euclidean_f1_threshold
157
+ value: 23.66771125793457
158
+ name: Euclidean F1 Threshold
159
+ - type: euclidean_precision
160
+ value: 0.8837209302325582
161
+ name: Euclidean Precision
162
+ - type: euclidean_recall
163
+ value: 0.926829268292683
164
+ name: Euclidean Recall
165
+ - type: euclidean_ap
166
+ value: 0.9690811253492324
167
+ name: Euclidean Ap
168
+ - type: max_accuracy
169
+ value: 0.8970588235294118
170
+ name: Max Accuracy
171
+ - type: max_accuracy_threshold
172
+ value: 453.2839660644531
173
+ name: Max Accuracy Threshold
174
+ - type: max_f1
175
+ value: 0.9102564102564101
176
+ name: Max F1
177
+ - type: max_f1_threshold
178
+ value: 453.2839660644531
179
+ name: Max F1 Threshold
180
+ - type: max_precision
181
+ value: 0.971830985915493
182
+ name: Max Precision
183
+ - type: max_recall
184
+ value: 0.926829268292683
185
+ name: Max Recall
186
+ - type: max_ap
187
+ value: 0.9692848872766847
188
+ name: Max Ap
189
  ---
190
 
191
+ # SentenceTransformer based on colorfulscoop/sbert-base-ja
 
 
 
192
 
193
+ This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [colorfulscoop/sbert-base-ja](https://huggingface.co/colorfulscoop/sbert-base-ja) on the csv dataset. It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.
194
 
195
  ## Model Details
196
 
197
  ### Model Description
198
+ - **Model Type:** Sentence Transformer
199
+ - **Base model:** [colorfulscoop/sbert-base-ja](https://huggingface.co/colorfulscoop/sbert-base-ja) <!-- at revision ecb8a98cd5176719ff7ab0d770a27420118732cf -->
200
+ - **Maximum Sequence Length:** 512 tokens
201
+ - **Output Dimensionality:** 768 tokens
202
+ - **Similarity Function:** Cosine Similarity
203
+ - **Training Dataset:**
204
+ - csv
205
+ <!-- - **Language:** Unknown -->
206
+ <!-- - **License:** Unknown -->
207
 
208
+ ### Model Sources
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
209
 
210
+ - **Documentation:** [Sentence Transformers Documentation](https://sbert.net)
211
+ - **Repository:** [Sentence Transformers on GitHub](https://github.com/UKPLab/sentence-transformers)
212
+ - **Hugging Face:** [Sentence Transformers on Hugging Face](https://huggingface.co/models?library=sentence-transformers)
213
 
214
+ ### Full Model Architecture
215
 
216
+ ```
217
+ SentenceTransformer(
218
+ (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel
219
+ (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, 'include_prompt': True})
220
+ )
221
+ ```
222
 
223
+ ## Usage
224
 
225
+ ### Direct Usage (Sentence Transformers)
226
 
227
+ First install the Sentence Transformers library:
228
 
229
+ ```bash
230
+ pip install -U sentence-transformers
231
+ ```
 
 
232
 
233
+ Then you can load this model and run inference.
234
+ ```python
235
+ from sentence_transformers import SentenceTransformer
236
 
237
+ # Download from the 🤗 Hub
238
+ model = SentenceTransformer("sentence_transformers_model_id")
239
+ # Run inference
240
+ sentences = [
241
+ '猫のぬいぐるみ',
242
+ 'ぬいぐるみ',
243
+ '両方はだめ?',
244
+ ]
245
+ embeddings = model.encode(sentences)
246
+ print(embeddings.shape)
247
+ # [3, 768]
248
 
249
+ # Get the similarity scores for the embeddings
250
+ similarities = model.similarity(embeddings, embeddings)
251
+ print(similarities.shape)
252
+ # [3, 3]
253
+ ```
254
 
255
+ <!--
256
+ ### Direct Usage (Transformers)
257
 
258
+ <details><summary>Click to see the direct usage in Transformers</summary>
259
 
260
+ </details>
261
+ -->
 
 
 
 
 
262
 
263
+ <!--
264
+ ### Downstream Usage (Sentence Transformers)
265
 
266
+ You can finetune this model on your own dataset.
267
 
268
+ <details><summary>Click to expand</summary>
269
 
270
+ </details>
271
+ -->
272
 
273
+ <!--
274
+ ### Out-of-Scope Use
 
 
 
 
 
 
 
 
 
 
 
 
275
 
276
+ *List how the model may foreseeably be misused and address what users ought not to do with the model.*
277
+ -->
278
 
279
  ## Evaluation
280
 
281
+ ### Metrics
282
+
283
+ #### Binary Classification
284
+ * Dataset: `custom-arc-semantics-data-jp`
285
+ * Evaluated with [<code>BinaryClassificationEvaluator</code>](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.BinaryClassificationEvaluator)
286
+
287
+ | Metric | Value |
288
+ |:-----------------------------|:-----------|
289
+ | cosine_accuracy | 0.8897 |
290
+ | cosine_accuracy_threshold | 0.6582 |
291
+ | cosine_f1 | 0.9045 |
292
+ | cosine_f1_threshold | 0.618 |
293
+ | cosine_precision | 0.9467 |
294
+ | cosine_recall | 0.8659 |
295
+ | cosine_ap | 0.9693 |
296
+ | dot_accuracy | 0.8897 |
297
+ | dot_accuracy_threshold | 374.5417 |
298
+ | dot_f1 | 0.902 |
299
+ | dot_f1_threshold | 374.5417 |
300
+ | dot_precision | 0.9718 |
301
+ | dot_recall | 0.8415 |
302
+ | dot_ap | 0.9691 |
303
+ | manhattan_accuracy | 0.8971 |
304
+ | manhattan_accuracy_threshold | 453.284 |
305
+ | manhattan_f1 | 0.9103 |
306
+ | manhattan_f1_threshold | 453.284 |
307
+ | manhattan_precision | 0.9595 |
308
+ | manhattan_recall | 0.8659 |
309
+ | manhattan_ap | 0.9688 |
310
+ | euclidean_accuracy | 0.8897 |
311
+ | euclidean_accuracy_threshold | 19.752 |
312
+ | euclidean_f1 | 0.9048 |
313
+ | euclidean_f1_threshold | 23.6677 |
314
+ | euclidean_precision | 0.8837 |
315
+ | euclidean_recall | 0.9268 |
316
+ | euclidean_ap | 0.9691 |
317
+ | max_accuracy | 0.8971 |
318
+ | max_accuracy_threshold | 453.284 |
319
+ | max_f1 | 0.9103 |
320
+ | max_f1_threshold | 453.284 |
321
+ | max_precision | 0.9718 |
322
+ | max_recall | 0.9268 |
323
+ | **max_ap** | **0.9693** |
324
+
325
+ <!--
326
+ ## Bias, Risks and Limitations
327
+
328
+ *What are the known or foreseeable issues stemming from this model? You could also flag here known failure cases or weaknesses of the model.*
329
+ -->
330
+
331
+ <!--
332
+ ### Recommendations
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
333
 
334
+ *What are recommendations with respect to the foreseeable issues? For example, filtering explicit content.*
335
+ -->
336
 
337
+ ## Training Details
338
 
339
+ ### Training Dataset
340
+
341
+ #### csv
342
+
343
+ * Dataset: csv
344
+ * Size: 680 training samples
345
+ * Columns: <code>text1</code>, <code>text2</code>, and <code>label</code>
346
+ * Approximate statistics based on the first 680 samples:
347
+ | | text1 | text2 | label |
348
+ |:--------|:---------------------------------------------------------------------------------|:--------------------------------------------------------------------------------|:------------------------------------------------|
349
+ | type | string | string | int |
350
+ | details | <ul><li>min: 4 tokens</li><li>mean: 8.32 tokens</li><li>max: 15 tokens</li></ul> | <ul><li>min: 4 tokens</li><li>mean: 8.0 tokens</li><li>max: 14 tokens</li></ul> | <ul><li>0: ~41.73%</li><li>1: ~58.27%</li></ul> |
351
+ * Samples:
352
+ | text1 | text2 | label |
353
+ |:----------------------|:---------------------------|:---------------|
354
+ | <code>試すため</code> | <code>ためすため</code> | <code>1</code> |
355
+ | <code>お鍋からの香り</code> | <code>お鍋から辛い匂いがしたから</code> | <code>1</code> |
356
+ | <code>なんで話せるの?</code> | <code>なんでしゃべれるの?</code> | <code>1</code> |
357
+ * Loss: [<code>ContrastiveLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#contrastiveloss) with these parameters:
358
+ ```json
359
+ {
360
+ "distance_metric": "SiameseDistanceMetric.COSINE_DISTANCE",
361
+ "margin": 0.8,
362
+ "size_average": true
363
+ }
364
+ ```
365
+
366
+ ### Evaluation Dataset
367
+
368
+ #### csv
369
+
370
+ * Dataset: csv
371
+ * Size: 680 evaluation samples
372
+ * Columns: <code>text1</code>, <code>text2</code>, and <code>label</code>
373
+ * Approximate statistics based on the first 680 samples:
374
+ | | text1 | text2 | label |
375
+ |:--------|:---------------------------------------------------------------------------------|:---------------------------------------------------------------------------------|:------------------------------------------------|
376
+ | type | string | string | int |
377
+ | details | <ul><li>min: 4 tokens</li><li>mean: 8.21 tokens</li><li>max: 13 tokens</li></ul> | <ul><li>min: 4 tokens</li><li>mean: 8.04 tokens</li><li>max: 14 tokens</li></ul> | <ul><li>0: ~39.71%</li><li>1: ~60.29%</li></ul> |
378
+ * Samples:
379
+ | text1 | text2 | label |
380
+ |:-----------------------|:-----------------------|:---------------|
381
+ | <code>村人について教えて</code> | <code>猫のぬいぐるみ</code> | <code>0</code> |
382
+ | <code>ハロー</code> | <code>やあ</code> | <code>1</code> |
383
+ | <code>窓から出て行った</code> | <code>オブリビオンの魔法</code> | <code>0</code> |
384
+ * Loss: [<code>ContrastiveLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#contrastiveloss) with these parameters:
385
+ ```json
386
+ {
387
+ "distance_metric": "SiameseDistanceMetric.COSINE_DISTANCE",
388
+ "margin": 0.8,
389
+ "size_average": true
390
+ }
391
+ ```
392
+
393
+ ### Training Hyperparameters
394
+ #### Non-Default Hyperparameters
395
+
396
+ - `eval_strategy`: epoch
397
+ - `learning_rate`: 2e-05
398
+ - `num_train_epochs`: 5
399
+ - `warmup_ratio`: 0.1
400
+ - `fp16`: True
401
+ - `batch_sampler`: no_duplicates
402
+
403
+ #### All Hyperparameters
404
+ <details><summary>Click to expand</summary>
405
+
406
+ - `overwrite_output_dir`: False
407
+ - `do_predict`: False
408
+ - `eval_strategy`: epoch
409
+ - `prediction_loss_only`: True
410
+ - `per_device_train_batch_size`: 8
411
+ - `per_device_eval_batch_size`: 8
412
+ - `per_gpu_train_batch_size`: None
413
+ - `per_gpu_eval_batch_size`: None
414
+ - `gradient_accumulation_steps`: 1
415
+ - `eval_accumulation_steps`: None
416
+ - `torch_empty_cache_steps`: None
417
+ - `learning_rate`: 2e-05
418
+ - `weight_decay`: 0.0
419
+ - `adam_beta1`: 0.9
420
+ - `adam_beta2`: 0.999
421
+ - `adam_epsilon`: 1e-08
422
+ - `max_grad_norm`: 1.0
423
+ - `num_train_epochs`: 5
424
+ - `max_steps`: -1
425
+ - `lr_scheduler_type`: linear
426
+ - `lr_scheduler_kwargs`: {}
427
+ - `warmup_ratio`: 0.1
428
+ - `warmup_steps`: 0
429
+ - `log_level`: passive
430
+ - `log_level_replica`: warning
431
+ - `log_on_each_node`: True
432
+ - `logging_nan_inf_filter`: True
433
+ - `save_safetensors`: True
434
+ - `save_on_each_node`: False
435
+ - `save_only_model`: False
436
+ - `restore_callback_states_from_checkpoint`: False
437
+ - `no_cuda`: False
438
+ - `use_cpu`: False
439
+ - `use_mps_device`: False
440
+ - `seed`: 42
441
+ - `data_seed`: None
442
+ - `jit_mode_eval`: False
443
+ - `use_ipex`: False
444
+ - `bf16`: False
445
+ - `fp16`: True
446
+ - `fp16_opt_level`: O1
447
+ - `half_precision_backend`: auto
448
+ - `bf16_full_eval`: False
449
+ - `fp16_full_eval`: False
450
+ - `tf32`: None
451
+ - `local_rank`: 0
452
+ - `ddp_backend`: None
453
+ - `tpu_num_cores`: None
454
+ - `tpu_metrics_debug`: False
455
+ - `debug`: []
456
+ - `dataloader_drop_last`: False
457
+ - `dataloader_num_workers`: 0
458
+ - `dataloader_prefetch_factor`: None
459
+ - `past_index`: -1
460
+ - `disable_tqdm`: False
461
+ - `remove_unused_columns`: True
462
+ - `label_names`: None
463
+ - `load_best_model_at_end`: False
464
+ - `ignore_data_skip`: False
465
+ - `fsdp`: []
466
+ - `fsdp_min_num_params`: 0
467
+ - `fsdp_config`: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
468
+ - `fsdp_transformer_layer_cls_to_wrap`: None
469
+ - `accelerator_config`: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
470
+ - `deepspeed`: None
471
+ - `label_smoothing_factor`: 0.0
472
+ - `optim`: adamw_torch
473
+ - `optim_args`: None
474
+ - `adafactor`: False
475
+ - `group_by_length`: False
476
+ - `length_column_name`: length
477
+ - `ddp_find_unused_parameters`: None
478
+ - `ddp_bucket_cap_mb`: None
479
+ - `ddp_broadcast_buffers`: False
480
+ - `dataloader_pin_memory`: True
481
+ - `dataloader_persistent_workers`: False
482
+ - `skip_memory_metrics`: True
483
+ - `use_legacy_prediction_loop`: False
484
+ - `push_to_hub`: False
485
+ - `resume_from_checkpoint`: None
486
+ - `hub_model_id`: None
487
+ - `hub_strategy`: every_save
488
+ - `hub_private_repo`: False
489
+ - `hub_always_push`: False
490
+ - `gradient_checkpointing`: False
491
+ - `gradient_checkpointing_kwargs`: None
492
+ - `include_inputs_for_metrics`: False
493
+ - `eval_do_concat_batches`: True
494
+ - `fp16_backend`: auto
495
+ - `push_to_hub_model_id`: None
496
+ - `push_to_hub_organization`: None
497
+ - `mp_parameters`:
498
+ - `auto_find_batch_size`: False
499
+ - `full_determinism`: False
500
+ - `torchdynamo`: None
501
+ - `ray_scope`: last
502
+ - `ddp_timeout`: 1800
503
+ - `torch_compile`: False
504
+ - `torch_compile_backend`: None
505
+ - `torch_compile_mode`: None
506
+ - `dispatch_batches`: None
507
+ - `split_batches`: None
508
+ - `include_tokens_per_second`: False
509
+ - `include_num_input_tokens_seen`: False
510
+ - `neftune_noise_alpha`: None
511
+ - `optim_target_modules`: None
512
+ - `batch_eval_metrics`: False
513
+ - `eval_on_start`: False
514
+ - `eval_use_gather_object`: False
515
+ - `batch_sampler`: no_duplicates
516
+ - `multi_dataset_batch_sampler`: proportional
517
+
518
+ </details>
519
+
520
+ ### Training Logs
521
+ | Epoch | Step | Training Loss | loss | custom-arc-semantics-data-jp_max_ap |
522
+ |:-----:|:----:|:-------------:|:------:|:-----------------------------------:|
523
+ | None | 0 | - | - | 0.9118 |
524
+ | 1.0 | 68 | 0.0481 | 0.0342 | 0.9611 |
525
+ | 2.0 | 136 | 0.0307 | 0.0318 | 0.9656 |
526
+ | 3.0 | 204 | 0.0218 | 0.0282 | 0.9728 |
527
+ | 4.0 | 272 | 0.0169 | 0.0285 | 0.9706 |
528
+ | 5.0 | 340 | 0.0144 | 0.0289 | 0.9693 |
529
+
530
+
531
+ ### Framework Versions
532
+ - Python: 3.10.14
533
+ - Sentence Transformers: 3.1.0
534
+ - Transformers: 4.44.2
535
+ - PyTorch: 2.4.1+cu121
536
+ - Accelerate: 0.34.2
537
+ - Datasets: 2.20.0
538
+ - Tokenizers: 0.19.1
539
+
540
+ ## Citation
541
+
542
+ ### BibTeX
543
+
544
+ #### Sentence Transformers
545
+ ```bibtex
546
+ @inproceedings{reimers-2019-sentence-bert,
547
+ title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",
548
+ author = "Reimers, Nils and Gurevych, Iryna",
549
+ booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",
550
+ month = "11",
551
+ year = "2019",
552
+ publisher = "Association for Computational Linguistics",
553
+ url = "https://arxiv.org/abs/1908.10084",
554
+ }
555
+ ```
556
+
557
+ #### ContrastiveLoss
558
+ ```bibtex
559
+ @inproceedings{hadsell2006dimensionality,
560
+ author={Hadsell, R. and Chopra, S. and LeCun, Y.},
561
+ booktitle={2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06)},
562
+ title={Dimensionality Reduction by Learning an Invariant Mapping},
563
+ year={2006},
564
+ volume={2},
565
+ number={},
566
+ pages={1735-1742},
567
+ doi={10.1109/CVPR.2006.100}
568
+ }
569
+ ```
570
+
571
+ <!--
572
+ ## Glossary
573
+
574
+ *Clearly define terms in order to be accessible across audiences.*
575
+ -->
576
+
577
+ <!--
578
+ ## Model Card Authors
579
+
580
+ *Lists the people who create the model card, providing recognition and accountability for the detailed work that goes into its construction.*
581
+ -->
582
+
583
+ <!--
584
  ## Model Card Contact
585
 
586
+ *Provides a way for people who have updates to the Model Card, suggestions, or questions, to contact the Model Card authors.*
587
+ -->
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:6ea971648a5f2a8763054e8e39824b1f5acc795ea1e4e11c3565210e0f89f56c
3
  size 442491744
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7b664b71c9092c8de7fd9ee4d7b22efbd8830a5a65a7b79a3772e033a198b2c5
3
  size 442491744
runs/Sep12_17-23-58_default/events.out.tfevents.1726161847.default.701.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:466ec0e1ba224f2f6c329fdc6de7d178d40471885e3f725f308eadd0e27bd182
3
+ size 4184
runs/Sep12_17-26-39_default/events.out.tfevents.1726162005.default.764.0 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ccc15be72347419a945e301703bb4d984c6fd5ddbf9fa2ea13849a5a3dbb2eba
3
+ size 22423