Add new SentenceTransformer model.
Browse files- README.md +280 -52
- model.safetensors +1 -1
- runs/Sep17_23-47-31_default/events.out.tfevents.1726616853.default.8433.0 +3 -0
- tokenizer_config.json +14 -64
README.md
CHANGED
@@ -1,6 +1,42 @@
|
|
1 |
---
|
2 |
base_model: colorfulscoop/sbert-base-ja
|
3 |
library_name: sentence-transformers
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
4 |
pipeline_tag: sentence-similarity
|
5 |
tags:
|
6 |
- sentence-transformers
|
@@ -10,38 +46,141 @@ tags:
|
|
10 |
- dataset_size:53
|
11 |
- loss:CosineSimilarityLoss
|
12 |
widget:
|
13 |
-
- source_sentence:
|
14 |
sentences:
|
15 |
-
-
|
16 |
-
-
|
17 |
-
- 人々 は
|
18 |
-
- source_sentence:
|
19 |
-
見え ます 。
|
20 |
sentences:
|
21 |
-
-
|
22 |
-
-
|
23 |
- 木 を 切り 倒した 後 、 木 の 切り株 に 座って いる 少年 。
|
24 |
-
- source_sentence:
|
25 |
-
シャツ を 着た 友人 は 後ろ から 笑って い ます 。
|
26 |
sentences:
|
27 |
-
-
|
28 |
-
-
|
29 |
-
-
|
30 |
-
-
|
31 |
-
|
32 |
-
|
33 |
-
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
40 |
---
|
41 |
|
42 |
# SentenceTransformer based on colorfulscoop/sbert-base-ja
|
43 |
|
44 |
-
This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [colorfulscoop/sbert-base-ja](https://huggingface.co/colorfulscoop/sbert-base-ja). 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.
|
45 |
|
46 |
## Model Details
|
47 |
|
@@ -51,7 +190,8 @@ This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [c
|
|
51 |
- **Maximum Sequence Length:** 512 tokens
|
52 |
- **Output Dimensionality:** 768 tokens
|
53 |
- **Similarity Function:** Cosine Similarity
|
54 |
-
|
|
|
55 |
<!-- - **Language:** Unknown -->
|
56 |
<!-- - **License:** Unknown -->
|
57 |
|
@@ -88,9 +228,9 @@ from sentence_transformers import SentenceTransformer
|
|
88 |
model = SentenceTransformer("sentence_transformers_model_id")
|
89 |
# Run inference
|
90 |
sentences = [
|
91 |
-
'
|
92 |
-
'
|
93 |
-
'
|
94 |
]
|
95 |
embeddings = model.encode(sentences)
|
96 |
print(embeddings.shape)
|
@@ -126,6 +266,52 @@ You can finetune this model on your own dataset.
|
|
126 |
*List how the model may foreseeably be misused and address what users ought not to do with the model.*
|
127 |
-->
|
128 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
129 |
<!--
|
130 |
## Bias, Risks and Limitations
|
131 |
|
@@ -142,22 +328,47 @@ You can finetune this model on your own dataset.
|
|
142 |
|
143 |
### Training Dataset
|
144 |
|
145 |
-
####
|
146 |
-
|
147 |
|
|
|
148 |
* Size: 53 training samples
|
149 |
-
* Columns: <code>
|
150 |
* Approximate statistics based on the first 53 samples:
|
151 |
-
| |
|
152 |
|:--------|:-----------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------|:------------------------------------------------|
|
153 |
| type | string | string | int |
|
154 |
-
| details | <ul><li>min: 14 tokens</li><li>mean: 36
|
155 |
* Samples:
|
156 |
-
|
|
157 |
-
|
158 |
-
| <code
|
159 |
-
| <code
|
160 |
-
| <code
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
161 |
* Loss: [<code>CosineSimilarityLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#cosinesimilarityloss) with these parameters:
|
162 |
```json
|
163 |
{
|
@@ -168,36 +379,38 @@ You can finetune this model on your own dataset.
|
|
168 |
### Training Hyperparameters
|
169 |
#### Non-Default Hyperparameters
|
170 |
|
171 |
-
- `
|
172 |
-
- `
|
173 |
-
- `num_train_epochs`:
|
174 |
-
- `
|
|
|
|
|
175 |
|
176 |
#### All Hyperparameters
|
177 |
<details><summary>Click to expand</summary>
|
178 |
|
179 |
- `overwrite_output_dir`: False
|
180 |
- `do_predict`: False
|
181 |
-
- `eval_strategy`:
|
182 |
- `prediction_loss_only`: True
|
183 |
-
- `per_device_train_batch_size`:
|
184 |
-
- `per_device_eval_batch_size`:
|
185 |
- `per_gpu_train_batch_size`: None
|
186 |
- `per_gpu_eval_batch_size`: None
|
187 |
- `gradient_accumulation_steps`: 1
|
188 |
- `eval_accumulation_steps`: None
|
189 |
- `torch_empty_cache_steps`: None
|
190 |
-
- `learning_rate`:
|
191 |
- `weight_decay`: 0.0
|
192 |
- `adam_beta1`: 0.9
|
193 |
- `adam_beta2`: 0.999
|
194 |
- `adam_epsilon`: 1e-08
|
195 |
-
- `max_grad_norm`: 1
|
196 |
-
- `num_train_epochs`:
|
197 |
- `max_steps`: -1
|
198 |
- `lr_scheduler_type`: linear
|
199 |
- `lr_scheduler_kwargs`: {}
|
200 |
-
- `warmup_ratio`: 0.
|
201 |
- `warmup_steps`: 0
|
202 |
- `log_level`: passive
|
203 |
- `log_level_replica`: warning
|
@@ -215,7 +428,7 @@ You can finetune this model on your own dataset.
|
|
215 |
- `jit_mode_eval`: False
|
216 |
- `use_ipex`: False
|
217 |
- `bf16`: False
|
218 |
-
- `fp16`:
|
219 |
- `fp16_opt_level`: O1
|
220 |
- `half_precision_backend`: auto
|
221 |
- `bf16_full_eval`: False
|
@@ -285,11 +498,26 @@ You can finetune this model on your own dataset.
|
|
285 |
- `batch_eval_metrics`: False
|
286 |
- `eval_on_start`: False
|
287 |
- `eval_use_gather_object`: False
|
288 |
-
- `batch_sampler`:
|
289 |
-
- `multi_dataset_batch_sampler`:
|
290 |
|
291 |
</details>
|
292 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
293 |
### Framework Versions
|
294 |
- Python: 3.10.14
|
295 |
- Sentence Transformers: 3.1.0
|
|
|
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
|
|
|
46 |
- dataset_size:53
|
47 |
- loss:CosineSimilarityLoss
|
48 |
widget:
|
49 |
+
- source_sentence: 黒い タイル の 本当に すてきな カウンター の 前 と 後ろ で 働く 人々 。
|
50 |
sentences:
|
51 |
+
- 男性 は バレエ に 参加 して い ます 。
|
52 |
+
- 岩 の 上 に 座って いる 二 人
|
53 |
+
- 人々 は 宝石 店 で 働いて い ます 。
|
54 |
+
- source_sentence: 少年 は 木 の 切り株 に 座って い ます 。
|
|
|
55 |
sentences:
|
56 |
+
- ストリート ワーカー は 保護 具 を 着用 して い ませ ん 。
|
57 |
+
- 芝生 の エリア で 数 匹 の 犬 が 交流 し ます 。
|
58 |
- 木 を 切り 倒した 後 、 木 の 切り株 に 座って いる 少年 。
|
59 |
+
- source_sentence: 岩 の 多い 景色 を 見て 二 人
|
|
|
60 |
sentences:
|
61 |
+
- 何 か を 見て いる 二 人 が い ます 。
|
62 |
+
- 誰 か が 肖像 画 を 描いて い ます 。
|
63 |
+
- バイカー は 足 を 使って 自転車 を さらに 進め ます 。
|
64 |
+
model-index:
|
65 |
+
- name: SentenceTransformer based on colorfulscoop/sbert-base-ja
|
66 |
+
results:
|
67 |
+
- task:
|
68 |
+
type: binary-classification
|
69 |
+
name: Binary Classification
|
70 |
+
dataset:
|
71 |
+
name: custom arc semantics data jp
|
72 |
+
type: custom-arc-semantics-data-jp
|
73 |
+
metrics:
|
74 |
+
- type: cosine_accuracy
|
75 |
+
value: 0.6363636363636364
|
76 |
+
name: Cosine Accuracy
|
77 |
+
- type: cosine_accuracy_threshold
|
78 |
+
value: 0.32276761531829834
|
79 |
+
name: Cosine Accuracy Threshold
|
80 |
+
- type: cosine_f1
|
81 |
+
value: 0.7777777777777777
|
82 |
+
name: Cosine F1
|
83 |
+
- type: cosine_f1_threshold
|
84 |
+
value: 0.32276761531829834
|
85 |
+
name: Cosine F1 Threshold
|
86 |
+
- type: cosine_precision
|
87 |
+
value: 0.7
|
88 |
+
name: Cosine Precision
|
89 |
+
- type: cosine_recall
|
90 |
+
value: 0.875
|
91 |
+
name: Cosine Recall
|
92 |
+
- type: cosine_ap
|
93 |
+
value: 0.619629329004329
|
94 |
+
name: Cosine Ap
|
95 |
+
- type: dot_accuracy
|
96 |
+
value: 0.6363636363636364
|
97 |
+
name: Dot Accuracy
|
98 |
+
- type: dot_accuracy_threshold
|
99 |
+
value: 180.3168487548828
|
100 |
+
name: Dot Accuracy Threshold
|
101 |
+
- type: dot_f1
|
102 |
+
value: 0.7777777777777777
|
103 |
+
name: Dot F1
|
104 |
+
- type: dot_f1_threshold
|
105 |
+
value: 180.3168487548828
|
106 |
+
name: Dot F1 Threshold
|
107 |
+
- type: dot_precision
|
108 |
+
value: 0.7
|
109 |
+
name: Dot Precision
|
110 |
+
- type: dot_recall
|
111 |
+
value: 0.875
|
112 |
+
name: Dot Recall
|
113 |
+
- type: dot_ap
|
114 |
+
value: 0.650879329004329
|
115 |
+
name: Dot Ap
|
116 |
+
- type: manhattan_accuracy
|
117 |
+
value: 0.6363636363636364
|
118 |
+
name: Manhattan Accuracy
|
119 |
+
- type: manhattan_accuracy_threshold
|
120 |
+
value: 609.3980712890625
|
121 |
+
name: Manhattan Accuracy Threshold
|
122 |
+
- type: manhattan_f1
|
123 |
+
value: 0.7777777777777777
|
124 |
+
name: Manhattan F1
|
125 |
+
- type: manhattan_f1_threshold
|
126 |
+
value: 609.3980712890625
|
127 |
+
name: Manhattan F1 Threshold
|
128 |
+
- type: manhattan_precision
|
129 |
+
value: 0.7
|
130 |
+
name: Manhattan Precision
|
131 |
+
- type: manhattan_recall
|
132 |
+
value: 0.875
|
133 |
+
name: Manhattan Recall
|
134 |
+
- type: manhattan_ap
|
135 |
+
value: 0.619629329004329
|
136 |
+
name: Manhattan Ap
|
137 |
+
- type: euclidean_accuracy
|
138 |
+
value: 0.6363636363636364
|
139 |
+
name: Euclidean Accuracy
|
140 |
+
- type: euclidean_accuracy_threshold
|
141 |
+
value: 27.520790100097656
|
142 |
+
name: Euclidean Accuracy Threshold
|
143 |
+
- type: euclidean_f1
|
144 |
+
value: 0.7777777777777777
|
145 |
+
name: Euclidean F1
|
146 |
+
- type: euclidean_f1_threshold
|
147 |
+
value: 27.520790100097656
|
148 |
+
name: Euclidean F1 Threshold
|
149 |
+
- type: euclidean_precision
|
150 |
+
value: 0.7
|
151 |
+
name: Euclidean Precision
|
152 |
+
- type: euclidean_recall
|
153 |
+
value: 0.875
|
154 |
+
name: Euclidean Recall
|
155 |
+
- type: euclidean_ap
|
156 |
+
value: 0.619629329004329
|
157 |
+
name: Euclidean Ap
|
158 |
+
- type: max_accuracy
|
159 |
+
value: 0.6363636363636364
|
160 |
+
name: Max Accuracy
|
161 |
+
- type: max_accuracy_threshold
|
162 |
+
value: 609.3980712890625
|
163 |
+
name: Max Accuracy Threshold
|
164 |
+
- type: max_f1
|
165 |
+
value: 0.7777777777777777
|
166 |
+
name: Max F1
|
167 |
+
- type: max_f1_threshold
|
168 |
+
value: 609.3980712890625
|
169 |
+
name: Max F1 Threshold
|
170 |
+
- type: max_precision
|
171 |
+
value: 0.7
|
172 |
+
name: Max Precision
|
173 |
+
- type: max_recall
|
174 |
+
value: 0.875
|
175 |
+
name: Max Recall
|
176 |
+
- type: max_ap
|
177 |
+
value: 0.650879329004329
|
178 |
+
name: Max Ap
|
179 |
---
|
180 |
|
181 |
# SentenceTransformer based on colorfulscoop/sbert-base-ja
|
182 |
|
183 |
+
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.
|
184 |
|
185 |
## Model Details
|
186 |
|
|
|
190 |
- **Maximum Sequence Length:** 512 tokens
|
191 |
- **Output Dimensionality:** 768 tokens
|
192 |
- **Similarity Function:** Cosine Similarity
|
193 |
+
- **Training Dataset:**
|
194 |
+
- csv
|
195 |
<!-- - **Language:** Unknown -->
|
196 |
<!-- - **License:** Unknown -->
|
197 |
|
|
|
228 |
model = SentenceTransformer("sentence_transformers_model_id")
|
229 |
# Run inference
|
230 |
sentences = [
|
231 |
+
'岩 の 多い 景色 を 見て 二 人',
|
232 |
+
'何 か を 見て いる 二 人 が い ます 。',
|
233 |
+
'誰 か が 肖像 画 を 描いて い ます 。',
|
234 |
]
|
235 |
embeddings = model.encode(sentences)
|
236 |
print(embeddings.shape)
|
|
|
266 |
*List how the model may foreseeably be misused and address what users ought not to do with the model.*
|
267 |
-->
|
268 |
|
269 |
+
## Evaluation
|
270 |
+
|
271 |
+
### Metrics
|
272 |
+
|
273 |
+
#### Binary Classification
|
274 |
+
* Dataset: `custom-arc-semantics-data-jp`
|
275 |
+
* Evaluated with [<code>BinaryClassificationEvaluator</code>](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.BinaryClassificationEvaluator)
|
276 |
+
|
277 |
+
| Metric | Value |
|
278 |
+
|:-----------------------------|:-----------|
|
279 |
+
| cosine_accuracy | 0.6364 |
|
280 |
+
| cosine_accuracy_threshold | 0.3228 |
|
281 |
+
| cosine_f1 | 0.7778 |
|
282 |
+
| cosine_f1_threshold | 0.3228 |
|
283 |
+
| cosine_precision | 0.7 |
|
284 |
+
| cosine_recall | 0.875 |
|
285 |
+
| cosine_ap | 0.6196 |
|
286 |
+
| dot_accuracy | 0.6364 |
|
287 |
+
| dot_accuracy_threshold | 180.3168 |
|
288 |
+
| dot_f1 | 0.7778 |
|
289 |
+
| dot_f1_threshold | 180.3168 |
|
290 |
+
| dot_precision | 0.7 |
|
291 |
+
| dot_recall | 0.875 |
|
292 |
+
| dot_ap | 0.6509 |
|
293 |
+
| manhattan_accuracy | 0.6364 |
|
294 |
+
| manhattan_accuracy_threshold | 609.3981 |
|
295 |
+
| manhattan_f1 | 0.7778 |
|
296 |
+
| manhattan_f1_threshold | 609.3981 |
|
297 |
+
| manhattan_precision | 0.7 |
|
298 |
+
| manhattan_recall | 0.875 |
|
299 |
+
| manhattan_ap | 0.6196 |
|
300 |
+
| euclidean_accuracy | 0.6364 |
|
301 |
+
| euclidean_accuracy_threshold | 27.5208 |
|
302 |
+
| euclidean_f1 | 0.7778 |
|
303 |
+
| euclidean_f1_threshold | 27.5208 |
|
304 |
+
| euclidean_precision | 0.7 |
|
305 |
+
| euclidean_recall | 0.875 |
|
306 |
+
| euclidean_ap | 0.6196 |
|
307 |
+
| max_accuracy | 0.6364 |
|
308 |
+
| max_accuracy_threshold | 609.3981 |
|
309 |
+
| max_f1 | 0.7778 |
|
310 |
+
| max_f1_threshold | 609.3981 |
|
311 |
+
| max_precision | 0.7 |
|
312 |
+
| max_recall | 0.875 |
|
313 |
+
| **max_ap** | **0.6509** |
|
314 |
+
|
315 |
<!--
|
316 |
## Bias, Risks and Limitations
|
317 |
|
|
|
328 |
|
329 |
### Training Dataset
|
330 |
|
331 |
+
#### csv
|
|
|
332 |
|
333 |
+
* Dataset: csv
|
334 |
* Size: 53 training samples
|
335 |
+
* Columns: <code>text1</code>, <code>text2</code>, and <code>label</code>
|
336 |
* Approximate statistics based on the first 53 samples:
|
337 |
+
| | text1 | text2 | label |
|
338 |
|:--------|:-----------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------|:------------------------------------------------|
|
339 |
| type | string | string | int |
|
340 |
+
| details | <ul><li>min: 14 tokens</li><li>mean: 35.36 tokens</li><li>max: 79 tokens</li></ul> | <ul><li>min: 11 tokens</li><li>mean: 21.33 tokens</li><li>max: 38 tokens</li></ul> | <ul><li>0: ~38.10%</li><li>1: ~61.90%</li></ul> |
|
341 |
* Samples:
|
342 |
+
| text1 | text2 | label |
|
343 |
+
|:---------------------------------------------------------------------------------------|:----------------------------------------------------------|:---------------|
|
344 |
+
| <code>薄紫 色 の ドレス と 明るい ホット ピンク の 靴 を 着た 女性 が 、 水 と コーヒー を 飲んで テーブル に 座って い ます 。</code> | <code>ブラインド デート の 女性 が 座って 、 デート が 現れる の を 待ち ます 。</code> | <code>1</code> |
|
345 |
+
| <code>トラック を 自転車 で 走る 人々 の グループ 。</code> | <code>自転車 の 挑戦 に 勝とう と する 人々 の グループ 。</code> | <code>1</code> |
|
346 |
+
| <code>野球 の 試合 中 に 基地 を 走る 野球 選手 の シャープリー 。</code> | <code>Sharp ley は ゲーム で プレイ して い ます 。</code> | <code>0</code> |
|
347 |
+
* Loss: [<code>CosineSimilarityLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#cosinesimilarityloss) with these parameters:
|
348 |
+
```json
|
349 |
+
{
|
350 |
+
"loss_fct": "torch.nn.modules.loss.MSELoss"
|
351 |
+
}
|
352 |
+
```
|
353 |
+
|
354 |
+
### Evaluation Dataset
|
355 |
+
|
356 |
+
#### csv
|
357 |
+
|
358 |
+
* Dataset: csv
|
359 |
+
* Size: 53 evaluation samples
|
360 |
+
* Columns: <code>text1</code>, <code>text2</code>, and <code>label</code>
|
361 |
+
* Approximate statistics based on the first 53 samples:
|
362 |
+
| | text1 | text2 | label |
|
363 |
+
|:--------|:-----------------------------------------------------------------------------------|:-----------------------------------------------------------------------------------|:------------------------------------------------|
|
364 |
+
| type | string | string | int |
|
365 |
+
| details | <ul><li>min: 19 tokens</li><li>mean: 39.64 tokens</li><li>max: 84 tokens</li></ul> | <ul><li>min: 19 tokens</li><li>mean: 25.27 tokens</li><li>max: 38 tokens</li></ul> | <ul><li>0: ~27.27%</li><li>1: ~72.73%</li></ul> |
|
366 |
+
* Samples:
|
367 |
+
| text1 | text2 | label |
|
368 |
+
|:----------------------------------------------------------------------------------------------------------|:------------------------------------------------|:---------------|
|
369 |
+
| <code>岩 の 多い 景色 を 見て 二 人</code> | <code>何 か を 見て いる 二 人 が い ます 。</code> | <code>0</code> |
|
370 |
+
| <code>白い ヘルメット と オレンジ色 の シャツ 、 ジーンズ 、 白い トラック と オレンジ色 の パイロン の 前 に 反射 ジャケット を 着た 金髪 の ストリート ワーカー 。</code> | <code>ストリート ワーカー は 保護 具 を 着用 して い ませ ん 。</code> | <code>1</code> |
|
371 |
+
| <code>白い 帽子 を かぶった 女性 が 、 鮮やかな 色 の 岩 の 風景 を 描いて い ます 。 岩 層 自体 が 背景 に 見え ます 。</code> | <code>誰 か が 肖像 画 を 描いて い ます 。</code> | <code>1</code> |
|
372 |
* Loss: [<code>CosineSimilarityLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#cosinesimilarityloss) with these parameters:
|
373 |
```json
|
374 |
{
|
|
|
379 |
### Training Hyperparameters
|
380 |
#### Non-Default Hyperparameters
|
381 |
|
382 |
+
- `eval_strategy`: epoch
|
383 |
+
- `learning_rate`: 2e-05
|
384 |
+
- `num_train_epochs`: 10
|
385 |
+
- `warmup_ratio`: 0.4
|
386 |
+
- `fp16`: True
|
387 |
+
- `batch_sampler`: no_duplicates
|
388 |
|
389 |
#### All Hyperparameters
|
390 |
<details><summary>Click to expand</summary>
|
391 |
|
392 |
- `overwrite_output_dir`: False
|
393 |
- `do_predict`: False
|
394 |
+
- `eval_strategy`: epoch
|
395 |
- `prediction_loss_only`: True
|
396 |
+
- `per_device_train_batch_size`: 8
|
397 |
+
- `per_device_eval_batch_size`: 8
|
398 |
- `per_gpu_train_batch_size`: None
|
399 |
- `per_gpu_eval_batch_size`: None
|
400 |
- `gradient_accumulation_steps`: 1
|
401 |
- `eval_accumulation_steps`: None
|
402 |
- `torch_empty_cache_steps`: None
|
403 |
+
- `learning_rate`: 2e-05
|
404 |
- `weight_decay`: 0.0
|
405 |
- `adam_beta1`: 0.9
|
406 |
- `adam_beta2`: 0.999
|
407 |
- `adam_epsilon`: 1e-08
|
408 |
+
- `max_grad_norm`: 1.0
|
409 |
+
- `num_train_epochs`: 10
|
410 |
- `max_steps`: -1
|
411 |
- `lr_scheduler_type`: linear
|
412 |
- `lr_scheduler_kwargs`: {}
|
413 |
+
- `warmup_ratio`: 0.4
|
414 |
- `warmup_steps`: 0
|
415 |
- `log_level`: passive
|
416 |
- `log_level_replica`: warning
|
|
|
428 |
- `jit_mode_eval`: False
|
429 |
- `use_ipex`: False
|
430 |
- `bf16`: False
|
431 |
+
- `fp16`: True
|
432 |
- `fp16_opt_level`: O1
|
433 |
- `half_precision_backend`: auto
|
434 |
- `bf16_full_eval`: False
|
|
|
498 |
- `batch_eval_metrics`: False
|
499 |
- `eval_on_start`: False
|
500 |
- `eval_use_gather_object`: False
|
501 |
+
- `batch_sampler`: no_duplicates
|
502 |
+
- `multi_dataset_batch_sampler`: proportional
|
503 |
|
504 |
</details>
|
505 |
|
506 |
+
### Training Logs
|
507 |
+
| Epoch | Step | Training Loss | loss | custom-arc-semantics-data-jp_max_ap |
|
508 |
+
|:-----:|:----:|:-------------:|:------:|:-----------------------------------:|
|
509 |
+
| 1.0 | 6 | 0.2964 | 0.3110 | 0.7238 |
|
510 |
+
| 2.0 | 12 | 0.2768 | 0.3083 | 0.7238 |
|
511 |
+
| 3.0 | 18 | 0.2389 | 0.2999 | 0.7238 |
|
512 |
+
| 4.0 | 24 | 0.1897 | 0.2843 | 0.6946 |
|
513 |
+
| 5.0 | 30 | 0.1464 | 0.2776 | 0.7134 |
|
514 |
+
| 6.0 | 36 | 0.1112 | 0.2877 | 0.6509 |
|
515 |
+
| 7.0 | 42 | 0.087 | 0.3047 | 0.6509 |
|
516 |
+
| 8.0 | 48 | 0.0754 | 0.3135 | 0.6509 |
|
517 |
+
| 9.0 | 54 | 0.068 | 0.3150 | 0.6509 |
|
518 |
+
| 10.0 | 60 | 0.0588 | 0.3148 | 0.6509 |
|
519 |
+
|
520 |
+
|
521 |
### Framework Versions
|
522 |
- Python: 3.10.14
|
523 |
- Sentence Transformers: 3.1.0
|
model.safetensors
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
size 442491744
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c398d788c3c9eddbd690d96649f0d0fdc6601934c54d199ab895088e46bb5ccd
|
3 |
size 442491744
|
runs/Sep17_23-47-31_default/events.out.tfevents.1726616853.default.8433.0
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9e5714b312f0efb28d1323371ee280fc1148b233cd7c3ae98aea10da8a692d02
|
3 |
+
size 39857
|
tokenizer_config.json
CHANGED
@@ -1,65 +1,15 @@
|
|
1 |
{
|
2 |
-
|
3 |
-
"
|
4 |
-
|
5 |
-
|
6 |
-
|
7 |
-
|
8 |
-
|
9 |
-
|
10 |
-
|
11 |
-
"
|
12 |
-
|
13 |
-
|
14 |
-
|
15 |
-
|
16 |
-
"single_word": false,
|
17 |
-
"special": true
|
18 |
-
},
|
19 |
-
"2": {
|
20 |
-
"content": "[CLS]",
|
21 |
-
"lstrip": false,
|
22 |
-
"normalized": false,
|
23 |
-
"rstrip": false,
|
24 |
-
"single_word": false,
|
25 |
-
"special": false
|
26 |
-
},
|
27 |
-
"3": {
|
28 |
-
"content": "[SEP]",
|
29 |
-
"lstrip": false,
|
30 |
-
"normalized": false,
|
31 |
-
"rstrip": false,
|
32 |
-
"single_word": false,
|
33 |
-
"special": false
|
34 |
-
},
|
35 |
-
"4": {
|
36 |
-
"content": "[MASK]",
|
37 |
-
"lstrip": false,
|
38 |
-
"normalized": false,
|
39 |
-
"rstrip": false,
|
40 |
-
"single_word": false,
|
41 |
-
"special": false
|
42 |
-
},
|
43 |
-
"32000": {
|
44 |
-
"content": "[PAD]",
|
45 |
-
"lstrip": false,
|
46 |
-
"normalized": true,
|
47 |
-
"rstrip": false,
|
48 |
-
"single_word": false,
|
49 |
-
"special": false
|
50 |
-
}
|
51 |
-
},
|
52 |
-
"bos_token": "[CLS]",
|
53 |
-
"clean_up_tokenization_spaces": true,
|
54 |
-
"cls_token": "[CLS]",
|
55 |
-
"do_lower_case": false,
|
56 |
-
"eos_token": "[SEP]",
|
57 |
-
"mask_token": "[MASK]",
|
58 |
-
"model_max_length": 512,
|
59 |
-
"pad_token": "<pad>",
|
60 |
-
"sep_token": "[SEP]",
|
61 |
-
"sp_model_kwargs": {},
|
62 |
-
"split_by_punct": false,
|
63 |
-
"tokenizer_class": "DebertaV2Tokenizer",
|
64 |
-
"unk_token": "<unk>"
|
65 |
-
}
|
|
|
1 |
{
|
2 |
+
"bos_token": "[CLS]",
|
3 |
+
"clean_up_tokenization_spaces": true,
|
4 |
+
"cls_token": "[CLS]",
|
5 |
+
"do_lower_case": false,
|
6 |
+
"eos_token": "[SEP]",
|
7 |
+
"mask_token": "[MASK]",
|
8 |
+
"model_max_length": 512,
|
9 |
+
"pad_token": "<pad>",
|
10 |
+
"sep_token": "[SEP]",
|
11 |
+
"sp_model_kwargs": {},
|
12 |
+
"split_by_punct": false,
|
13 |
+
"tokenizer_class": "DebertaV2Tokenizer",
|
14 |
+
"unk_token": "<unk>"
|
15 |
+
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|