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:
|
| 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 |
+
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|