README.md

---

base_model: jhgan/ko-sroberta-multitask
datasets: []
language: []
library_name: sentence-transformers
metrics:
- pearson_cosine
- spearman_cosine
- pearson_manhattan
- spearman_manhattan
- pearson_euclidean
- spearman_euclidean
- pearson_dot
- spearman_dot
- pearson_max
- spearman_max
pipeline_tag: sentence-similarity
tags:
- sentence-transformers
- sentence-similarity
- feature-extraction
- generated_from_trainer
- dataset_size:574417
- loss:MultipleNegativesRankingLoss
- loss:CosineSimilarityLoss
widget:
- source_sentence: 파타키는 아브라함의 결정을 칭찬했고 리파 회장 리차드 케셀은 케이블이 영구적으로 가동되어야 한다고 말했다.
  sentences:
  - 이스라엘과 하마스 '일시적인 휴전을 받아들이다'
  - 리파 회장 리차드 케셀은 "우리가 보기에 케이블이 사용될 수 있다"고 말했다.
  - 하지만 그들은 그들의 유권자들에게 책임이 있다.
- source_sentence: 돛이 네 개 달린 배가 물 위를 항해하고 있다.
  sentences:
  - 돛단배가 물 위를 항해하고 있다.
  - 보스턴 마라톤 결승선에서 발생한 두 번의 폭발 보고
  - 레바논의 헤즈볼라 거점
- source_sentence: 딱딱한 모자를 쓴 남자가 건물 프레임 앞에 주차된 빨간 트럭의 침대를 쳐다본다.
  sentences:
  - 한 남자가 트럭을 보고 있다.
  - 한 명은 빨간 스웨터를 입고 다른 한 명은 하얀 스웨터를 입은 두 소년은 덤불 근처의 시멘트 블록에 앉아 있었다.
  - 남자가 자고 있다.
- source_sentence: 벽돌 건물 앞 발코니 뒤에 네 사람이 서 있다.
  sentences:
  - 그 사람은 경찰관이다.
  - 그들은 거실에 앉는다
  - 그 단체는 건물 밖에 있다
- source_sentence: 남자가 노래를 부르는 동안 두 남자가 악기를 연주한다.
  sentences:
  - 세 번째 남자가 악기를 연주하는 동안 두 남자가 노래를 부른다.
  - 베이 근처.
  - 3분의 1이 노래하는 동안 두 남자가 악기를 연주한다.
model-index:
- name: SentenceTransformer based on jhgan/ko-sroberta-multitask
  results:
  - task:
      type: semantic-similarity
      name: Semantic Similarity
    dataset:
      name: sts dev
      type: sts-dev
    metrics:
    - type: pearson_cosine
      value: 0.8668233431675435
      name: Pearson Cosine
    - type: spearman_cosine
      value: 0.870259876274258
      name: Spearman Cosine
    - type: pearson_manhattan
      value: 0.8619838546671155
      name: Pearson Manhattan
    - type: spearman_manhattan
      value: 0.8684094795174834
      name: Spearman Manhattan
    - type: pearson_euclidean
      value: 0.8623159159300648
      name: Pearson Euclidean
    - type: spearman_euclidean
      value: 0.8686012195776042
      name: Spearman Euclidean
    - type: pearson_dot
      value: 0.8474110764249254
      name: Pearson Dot
    - type: spearman_dot
      value: 0.8469132619978514
      name: Spearman Dot
    - type: pearson_max
      value: 0.8668233431675435
      name: Pearson Max
    - type: spearman_max
      value: 0.870259876274258
      name: Spearman Max
---


# SentenceTransformer based on jhgan/ko-sroberta-multitask

This is a [sentence-transformers](https://www.SBERT.net) model finetuned from [jhgan/ko-sroberta-multitask](https://huggingface.co/jhgan/ko-sroberta-multitask). 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.

## Model Details

### Model Description
- **Model Type:** Sentence Transformer
- **Base model:** [jhgan/ko-sroberta-multitask](https://huggingface.co/jhgan/ko-sroberta-multitask) <!-- at revision ab957ae6a91e99c4cad36d52063a2a9cf1bf4419 -->
- **Maximum Sequence Length:** 128 tokens
- **Output Dimensionality:** 768 tokens
- **Similarity Function:** Cosine Similarity
<!-- - **Training Dataset:** Unknown -->
<!-- - **Language:** Unknown -->
<!-- - **License:** Unknown -->

### Model Sources

- **Documentation:** [Sentence Transformers Documentation](https://sbert.net)
- **Repository:** [Sentence Transformers on GitHub](https://github.com/UKPLab/sentence-transformers)
- **Hugging Face:** [Sentence Transformers on Hugging Face](https://huggingface.co/models?library=sentence-transformers)

### Full Model Architecture

```

SentenceTransformer(

  (0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: RobertaModel 

  (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})

)

```

## Usage

### Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

```bash

pip install -U sentence-transformers

```

Then you can load this model and run inference.
```python

from sentence_transformers import SentenceTransformer



# Download from the 🤗 Hub

model = SentenceTransformer("sentence_transformers_model_id")

# Run inference

sentences = [

    '남자가 노래를 부르는 동안 두 남자가 악기를 연주한다.',

    '3분의 1이 노래하는 동안 두 남자가 악기를 연주한다.',

    '세 번째 남자가 악기를 연주하는 동안 두 남자가 노래를 부른다.',

]

embeddings = model.encode(sentences)

print(embeddings.shape)

# [3, 768]



# Get the similarity scores for the embeddings

similarities = model.similarity(embeddings, embeddings)

print(similarities.shape)

# [3, 3]

```

<!--
### Direct Usage (Transformers)

<details><summary>Click to see the direct usage in Transformers</summary>

</details>
-->

<!--
### Downstream Usage (Sentence Transformers)

You can finetune this model on your own dataset.

<details><summary>Click to expand</summary>

</details>
-->

<!--
### Out-of-Scope Use

*List how the model may foreseeably be misused and address what users ought not to do with the model.*
-->

## Evaluation

### Metrics

#### Semantic Similarity
* Dataset: `sts-dev`
* Evaluated with [<code>EmbeddingSimilarityEvaluator</code>](https://sbert.net/docs/package_reference/sentence_transformer/evaluation.html#sentence_transformers.evaluation.EmbeddingSimilarityEvaluator)

| Metric             | Value      |
|:-------------------|:-----------|
| pearson_cosine     | 0.8668     |

| spearman_cosine    | 0.8703     |
| pearson_manhattan  | 0.862      |

| spearman_manhattan | 0.8684     |
| pearson_euclidean  | 0.8623     |

| spearman_euclidean | 0.8686     |
| pearson_dot        | 0.8474     |

| spearman_dot       | 0.8469     |
| pearson_max        | 0.8668     |

| **spearman_max**   | **0.8703** |



<!--

## Bias, Risks and Limitations



*What are the known or foreseeable issues stemming from this model? You could also flag here known failure cases or weaknesses of the model.*

-->



<!--

### Recommendations



*What are recommendations with respect to the foreseeable issues? For example, filtering explicit content.*

-->



## Training Details



### Training Datasets



#### Unnamed Dataset





* Size: 568,640 training samples

* Columns: <code>sentence_0</code>, <code>sentence_1</code>, and <code>sentence_2</code>
* Approximate statistics based on the first 1000 samples:
  |         | sentence_0                                                                         | sentence_1                                                                        | sentence_2                                                                        |

  |:--------|:-----------------------------------------------------------------------------------|:----------------------------------------------------------------------------------|:----------------------------------------------------------------------------------|

  | type    | string                                                                             | string                                                                            | string                                                                            |

  | details | <ul><li>min: 4 tokens</li><li>mean: 19.21 tokens</li><li>max: 128 tokens</li></ul> | <ul><li>min: 3 tokens</li><li>mean: 18.31 tokens</li><li>max: 93 tokens</li></ul> | <ul><li>min: 4 tokens</li><li>mean: 14.57 tokens</li><li>max: 54 tokens</li></ul> |

* Samples:

  | sentence_0                              | sentence_1                                       | sentence_2                            |
  |:----------------------------------------|:-------------------------------------------------|:--------------------------------------|
  | <code>발생 부하가 함께 5% 적습니다.</code>         | <code>발생 부하의 5% 감소와 함께 11.</code>                | <code>발생 부하가 5% 증가합니다.</code>         |
  | <code>어떤 행사를 위해 음식과 옷을 배급하는 여성들.</code> | <code>여성들은 음식과 옷을 나눠줌으로써 난민들을 돕고 있다.</code>      | <code>여자들이 사막에서 오토바이를 운전하고 있다.</code> |
  | <code>어린 아이들은 그 지식을 얻을 필요가 있다.</code>   | <code>응, 우리 젊은이들 중 많은 사람들이 그걸 배워야 할 것 같아.</code> | <code>젊은 사람들은 배울 필요가 없다.</code>       |
* Loss: [<code>MultipleNegativesRankingLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#multiplenegativesrankingloss) with these parameters:
  ```json

  {

      "scale": 20.0,

      "similarity_fct": "cos_sim"

  }

  ```

#### Unnamed Dataset


* Size: 5,777 training samples
* Columns: <code>sentence_0</code>, <code>sentence_1</code>, and <code>label</code>
* Approximate statistics based on the first 1000 samples:
  |         | sentence_0                                                                        | sentence_1                                                                        | label                                                          |
  |:--------|:----------------------------------------------------------------------------------|:----------------------------------------------------------------------------------|:---------------------------------------------------------------|
  | type    | string                                                                            | string                                                                            | float                                                          |
  | details | <ul><li>min: 3 tokens</li><li>mean: 17.16 tokens</li><li>max: 66 tokens</li></ul> | <ul><li>min: 3 tokens</li><li>mean: 17.11 tokens</li><li>max: 64 tokens</li></ul> | <ul><li>min: 0.0</li><li>mean: 0.54</li><li>max: 1.0</li></ul> |
* Samples:
  | sentence_0                                                                                       | sentence_1                                                                                      | label             |
  |:-------------------------------------------------------------------------------------------------|:------------------------------------------------------------------------------------------------|:------------------|
  | <code>시는 드램 시장이 2003년에 2.9% 성장하여 157억 달러, 2004년에는 43% 성장하여 225억 달러가 될 것으로 예상하고 있다고 말했습니다.</code> | <code>미국 시장은 2003년에 2.1퍼센트가 감소한 30.6억 달러로, 그리고 나서 2004년에 15.7퍼센트가 증가하여 354억 달러로 성장할 것이다.</code> | <code>0.24</code> |
  | <code>오사마 빈 라덴 부인들 수감</code>                                                                     | <code>인도에서 촬영될 오사마 빈 라덴 영화</code>                                                               | <code>0.16</code> |
  | <code>파키스탄 전투기, '탈리반 은신처' 폭탄 터뜨리기</code>                                                         | <code>파키스탄은 시리아 측에 무기 공급을 중단하기를 원한다.</code>                                                     | <code>0.32</code> |
* Loss: [<code>CosineSimilarityLoss</code>](https://sbert.net/docs/package_reference/sentence_transformer/losses.html#cosinesimilarityloss) with these parameters:
  ```json

  {

      "loss_fct": "torch.nn.modules.loss.MSELoss"

  }

  ```

### Training Hyperparameters
#### Non-Default Hyperparameters

- `eval_strategy`: steps
- `num_train_epochs`: 5
- `batch_sampler`: no_duplicates

- `multi_dataset_batch_sampler`: round_robin



#### All Hyperparameters

<details><summary>Click to expand</summary>



- `overwrite_output_dir`: False

- `do_predict`: False
- `eval_strategy`: steps
- `prediction_loss_only`: True
- `per_device_train_batch_size`: 8
- `per_device_eval_batch_size`: 8
- `per_gpu_train_batch_size`: None
- `per_gpu_eval_batch_size`: None
- `gradient_accumulation_steps`: 1
- `eval_accumulation_steps`: None
- `learning_rate`: 5e-05
- `weight_decay`: 0.0
- `adam_beta1`: 0.9
- `adam_beta2`: 0.999
- `adam_epsilon`: 1e-08
- `max_grad_norm`: 1
- `num_train_epochs`: 5
- `max_steps`: -1
- `lr_scheduler_type`: linear
- `lr_scheduler_kwargs`: {}
- `warmup_ratio`: 0.0
- `warmup_steps`: 0
- `log_level`: passive
- `log_level_replica`: warning
- `log_on_each_node`: True
- `logging_nan_inf_filter`: True
- `save_safetensors`: True
- `save_on_each_node`: False
- `save_only_model`: False
- `restore_callback_states_from_checkpoint`: False
- `no_cuda`: False
- `use_cpu`: False
- `use_mps_device`: False
- `seed`: 42
- `data_seed`: None
- `jit_mode_eval`: False
- `use_ipex`: False
- `bf16`: False
- `fp16`: False
- `fp16_opt_level`: O1
- `half_precision_backend`: auto
- `bf16_full_eval`: False
- `fp16_full_eval`: False
- `tf32`: None
- `local_rank`: 0
- `ddp_backend`: None
- `tpu_num_cores`: None
- `tpu_metrics_debug`: False
- `debug`: []
- `dataloader_drop_last`: False
- `dataloader_num_workers`: 0
- `dataloader_prefetch_factor`: None
- `past_index`: -1
- `disable_tqdm`: False
- `remove_unused_columns`: True
- `label_names`: None
- `load_best_model_at_end`: False
- `ignore_data_skip`: False
- `fsdp`: []
- `fsdp_min_num_params`: 0
- `fsdp_config`: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}

- `fsdp_transformer_layer_cls_to_wrap`: None
- `accelerator_config`: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
- `deepspeed`: None
- `label_smoothing_factor`: 0.0
- `optim`: adamw_torch

- `optim_args`: None
- `adafactor`: False
- `group_by_length`: False
- `length_column_name`: length
- `ddp_find_unused_parameters`: None
- `ddp_bucket_cap_mb`: None
- `ddp_broadcast_buffers`: False
- `dataloader_pin_memory`: True
- `dataloader_persistent_workers`: False
- `skip_memory_metrics`: True
- `use_legacy_prediction_loop`: False
- `push_to_hub`: False
- `resume_from_checkpoint`: None
- `hub_model_id`: None
- `hub_strategy`: every_save

- `hub_private_repo`: False

- `hub_always_push`: False

- `gradient_checkpointing`: False
- `gradient_checkpointing_kwargs`: None
- `include_inputs_for_metrics`: False
- `eval_do_concat_batches`: True
- `fp16_backend`: auto
- `push_to_hub_model_id`: None
- `push_to_hub_organization`: None
- `mp_parameters`: 
- `auto_find_batch_size`: False
- `full_determinism`: False
- `torchdynamo`: None
- `ray_scope`: last
- `ddp_timeout`: 1800
- `torch_compile`: False
- `torch_compile_backend`: None
- `torch_compile_mode`: None
- `dispatch_batches`: None
- `split_batches`: None
- `include_tokens_per_second`: False
- `include_num_input_tokens_seen`: False
- `neftune_noise_alpha`: None
- `optim_target_modules`: None
- `batch_eval_metrics`: False
- `batch_sampler`: no_duplicates

- `multi_dataset_batch_sampler`: round_robin



</details>



### Training Logs

| Epoch  | Step | Training Loss | sts-dev_spearman_max |

|:------:|:----:|:-------------:|:--------------------:|

| 0.3458 | 500  | 0.1504        | -                    |

| 0.6916 | 1000 | 0.1662        | 0.8660               |

| 1.0007 | 1447 | -             | 0.8678               |

| 1.0367 | 1500 | 0.1575        | -                    |

| 1.3824 | 2000 | 0.0539        | 0.8590               |

| 1.7282 | 2500 | 0.0406        | -                    |

| 2.0007 | 2894 | -             | 0.8703               |





### Framework Versions

- Python: 3.11.9

- Sentence Transformers: 3.0.1

- Transformers: 4.41.2

- PyTorch: 2.2.2+cu121

- Accelerate: 0.31.0

- Datasets: 2.20.0

- Tokenizers: 0.19.1



## Citation



### BibTeX



#### Sentence Transformers

```bibtex

@inproceedings{reimers-2019-sentence-bert,

    title = "Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks",

    author = "Reimers, Nils and Gurevych, Iryna",

    booktitle = "Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing",

    month = "11",

    year = "2019",

    publisher = "Association for Computational Linguistics",

    url = "https://arxiv.org/abs/1908.10084",

}

```



#### MultipleNegativesRankingLoss

```bibtex

@misc{henderson2017efficient,

    title={Efficient Natural Language Response Suggestion for Smart Reply}, 

    author={Matthew Henderson and Rami Al-Rfou and Brian Strope and Yun-hsuan Sung and Laszlo Lukacs and Ruiqi Guo and Sanjiv Kumar and Balint Miklos and Ray Kurzweil},

    year={2017},

    eprint={1705.00652},

    archivePrefix={arXiv},

    primaryClass={cs.CL}

}

```



<!--

## Glossary



*Clearly define terms in order to be accessible across audiences.*

-->



<!--

## Model Card Authors



*Lists the people who create the model card, providing recognition and accountability for the detailed work that goes into its construction.*

-->



<!--

## Model Card Contact



*Provides a way for people who have updates to the Model Card, suggestions, or questions, to contact the Model Card authors.*

-->