This is a sentence-transformers model finetuned from dragonkue/multilingual-e5-small-ko-v2 on the json dataset. It maps sentences & paragraphs to a 384-dimensional dense vector space and can be used for retrieval.

Model Details

Model Description

Model Type: Sentence Transformer
Base model: dragonkue/multilingual-e5-small-ko-v2
Maximum Sequence Length: 512 tokens
Output Dimensionality: 384 dimensions
Similarity Function: Cosine Similarity
Supported Modality: Text
Training Dataset:
- json

Model Sources

Documentation: Sentence Transformers Documentation
Repository: Sentence Transformers on GitHub
Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'transformer_task': 'feature-extraction', 'modality_config': {'text': {'method': 'forward', 'method_output_name': 'last_hidden_state'}}, 'module_output_name': 'token_embeddings', 'architecture': 'BertModel'})
  (1): Pooling({'embedding_dimension': 384, 'pooling_mode': 'mean', 'include_prompt': True})
  (2): Normalize({})
)

Usage

Direct Usage (Sentence Transformers)

First install the Sentence Transformers library:

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
queries = [
    'query: 병과명',
]
documents = [
    'passage: 한글명: 병과명\n영문약어명: BR_NM\n단어조합: 병과 + 명\n단어영문약어조합: BR + NM\n단어도메인분류조합: 명\n마지막단어: 명\n마지막단어도메인분류명: 명\n설명: 군사조직에서 군인이 능력에 따라 수행하는 주요 임무를 분류한 것의 이름',
    'passage: 한글명: 변환파일명\n영문약어명: TRSF_FILE_NM\n단어조합: 변환 + 파일 + 명\n단어영문약어조합: TRSF + FILE + NM\n단어도메인분류조합: 명\n마지막단어: 명\n마지막단어도메인분류명: 명\n설명: 다르게 하여 바꾼 파일의 이름',
    'passage: 한글명: 국가명\n영문약어명: NTN_NM\n단어조합: 국가 + 명\n단어영문약어조합: NTN + NM\n단어도메인분류조합: 명\n마지막단어: 명\n마지막단어도메인분류명: 명\n설명: 일정한 영토와 거기에 사는 사람들로 구성되고, 주권(主權)에 따른 하나의 통치 조직을 가지고 있는 사회 집단의 이름\n이음동의어: 나라명, 국가명칭, 국가이름, 나라명칭, 나라이름',
]
query_embeddings = model.encode_query(queries)
document_embeddings = model.encode_document(documents)
print(query_embeddings.shape, document_embeddings.shape)
# [1, 384] [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(query_embeddings, document_embeddings)
print(similarities)
# tensor([[0.9488, 0.0378, 0.2989]])

Evaluation

Metrics

Information Retrieval

Dataset: meta-assistant-terms
Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.9889
cosine_accuracy@3	0.9997
cosine_accuracy@5	0.9997
cosine_precision@1	0.9889
cosine_precision@3	0.3332
cosine_precision@5	0.1999
cosine_precision@10	0.1
cosine_recall@1	0.9889
cosine_recall@3	0.9997
cosine_recall@5	0.9997
cosine_recall@10	1.0
cosine_ndcg@10	0.9955
cosine_mrr@10	0.994
cosine_map@100	0.994

Training Details

Training Dataset

json

Dataset: json
Size: 25,358 training samples
Columns: anchor, positive, and negative

Approximate statistics based on the first 100 samples:

	anchor	positive	negative
type	string	string	string
modality	text	text	text
details	min: 6 tokens mean: 7.65 tokens max: 10 tokens	min: 64 tokens mean: 84.45 tokens max: 105 tokens	min: 73 tokens mean: 102.46 tokens max: 150 tokens

Samples:

anchor	positive	negative
`query: 고도`	`passage: 한글명: 고도[object Object]영문약어명: ALTD[object Object]단어조합: 고도[object Object]단어영문약어조합: ALTD[object Object]단어도메인분류조합: 수[object Object]마지막단어: 고도[object Object]마지막단어도메인분류명: 수[object Object]설명: 평균 해수면 따위를 0으로 하여 측정한 대상 물체의 높이`	`passage: 한글명: 해발고도[object Object]영문약어명: ELVT_ALTD[object Object]단어조합: 해발 + 고도[object Object]단어영문약어조합: ELVT + ALTD[object Object]단어도메인분류조합: 수[object Object]마지막단어: 고도[object Object]마지막단어도메인분류명: 수[object Object]설명: 평균 해수면을 기준으로 하여 잰 어떤 지점의 높이`
`query: 국비`	`passage: 한글명: 국비[object Object]영문약어명: GVEXP[object Object]단어조합: 국비[object Object]단어영문약어조합: GVEXP[object Object]단어도메인분류조합: 비용[object Object]마지막단어: 국비[object Object]마지막단어도메인분류명: 비용[object Object]설명: 나라의 재정으로 부담하는 비용`	passage: 한글명: 국민주택기금금액[object Object]영문약어명: NHFND_AMT[object Object]단어조합: 국민주택기금 + 금액[object Object]단어영문약어조합: NHFND + AMT[object Object]단어도메인분류조합: 금액[object Object]마지막단어: 금액[object Object]마지막단어도메인분류명: 금액[object Object]설명: 한 세대당 85㎡ 이하로 규정된 국민 주택의 건설하는 것을 지원할 목적으로 마련한 기금에서 지원·융자 또는 보조하는 돈의 액수
`query: 군번`	`passage: 한글명: 군번[object Object]영문약어명: SRVNO[object Object]단어조합: 군번[object Object]단어영문약어조합: SRVNO[object Object]단어도메인분류조합: 번호[object Object]마지막단어: 군번[object Object]마지막단어도메인분류명: 번호[object Object]설명: 군인 개인마다 주어지는 고유 번호`	`passage: 한글명: 군경력개월수[object Object]영문약어명: MLCR_MCNT[object Object]단어조합: 군경력 + 개월수[object Object]단어영문약어조합: MLCR + MCNT[object Object]단어도메인분류조합: 수[object Object]마지막단어: 개월수[object Object]마지막단어도메인분류명: 수[object Object]설명: 군에 복무한 경력에 대한 개월의 수`

Loss: MultipleNegativesRankingLoss with these parameters:
json
1{ 2 "scale": 20.0, 3 "similarity_fct": "cos_sim", 4 "gather_across_devices": false, 5 "directions": [ 6 "query_to_doc" 7 ], 8 "partition_mode": "joint", 9 "hardness_mode": null, 10 "hardness_strength": 0.0 11}

Training Hyperparameters

Non-Default Hyperparameters

per_device_train_batch_size: 32
num_train_epochs: 2
learning_rate: 2e-05
warmup_steps: 0.1
fp16: True
dataloader_drop_last: True
batch_sampler: no_duplicates

All Hyperparameters

Click to expand

per_device_train_batch_size: 32
num_train_epochs: 2
max_steps: -1
learning_rate: 2e-05
lr_scheduler_type: linear
lr_scheduler_kwargs: None
warmup_steps: 0.1
optim: adamw_torch_fused
optim_args: None
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
optim_target_modules: None
gradient_accumulation_steps: 1
average_tokens_across_devices: True
max_grad_norm: 1.0
label_smoothing_factor: 0.0
bf16: False
fp16: True
bf16_full_eval: False
fp16_full_eval: False
tf32: None
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
torch_compile: False
torch_compile_backend: None
torch_compile_mode: None
use_liger_kernel: False
liger_kernel_config: None
use_cache: False
neftune_noise_alpha: None
torch_empty_cache_steps: None
auto_find_batch_size: False
log_on_each_node: True
logging_nan_inf_filter: True
include_num_input_tokens_seen: no
log_level: passive
log_level_replica: warning
disable_tqdm: False
project: huggingface
trackio_space_id: None
trackio_bucket_id: None
trackio_static_space_id: None
per_device_eval_batch_size: 8
prediction_loss_only: True
eval_on_start: False
eval_do_concat_batches: True
eval_use_gather_object: False
eval_accumulation_steps: None
include_for_metrics: []
batch_eval_metrics: False
save_only_model: False
save_on_each_node: False
enable_jit_checkpoint: False
push_to_hub: False
hub_private_repo: None
hub_model_id: None
hub_strategy: every_save
hub_always_push: False
hub_revision: None
load_best_model_at_end: False
ignore_data_skip: False
restore_callback_states_from_checkpoint: False
full_determinism: False
seed: 42
data_seed: None
use_cpu: False
accelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}
parallelism_config: None
dataloader_drop_last: True
dataloader_num_workers: 0
dataloader_pin_memory: True
dataloader_persistent_workers: False
dataloader_prefetch_factor: None
remove_unused_columns: True
label_names: None
train_sampling_strategy: random
length_column_name: length
ddp_find_unused_parameters: None
ddp_bucket_cap_mb: None
ddp_broadcast_buffers: False
ddp_static_graph: None
ddp_backend: None
ddp_timeout: 1800
fsdp: None
fsdp_config: None
deepspeed: None
debug: []
skip_memory_metrics: True
do_predict: False
resume_from_checkpoint: None
warmup_ratio: None
local_rank: -1
prompts: None
batch_sampler: no_duplicates
multi_dataset_batch_sampler: proportional
router_mapping: {}
learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	meta-assistant-terms_cosine_ndcg@10
-1	-1	-	0.9082
0.0253	20	1.0291	-
0.0505	40	0.4264	-
0.0758	60	0.1148	-
0.1010	80	0.0439	-
0.1263	100	0.0415	-
0.1515	120	0.0254	-
0.1768	140	0.0248	-
0.2020	160	0.0183	-
0.2273	180	0.0152	-
0.2525	200	0.0155	-
0.2778	220	0.0180	-
0.3030	240	0.0240	-
0.3283	260	0.0119	-
0.3535	280	0.0235	-
0.3788	300	0.0164	-
0.4040	320	0.0094	-
0.4293	340	0.0075	-
0.4545	360	0.0067	-
0.4798	380	0.0072	-
0.5051	400	0.0081	-
0.5303	420	0.0071	-
0.5556	440	0.0078	-
0.5808	460	0.0077	-
0.6061	480	0.0093	-
0.6313	500	0.0109	-
0.6566	520	0.0075	-
0.6818	540	0.0036	-
0.7071	560	0.0046	-
0.7323	580	0.0056	-
0.7576	600	0.0061	-
0.7828	620	0.0052	-
0.8081	640	0.0056	-
0.8333	660	0.0081	-
0.8586	680	0.0065	-
0.8838	700	0.0043	-
0.9091	720	0.0067	-
0.9343	740	0.0049	-
0.9596	760	0.0038	-
0.9848	780	0.0054	-
1.0	792	-	0.9951
1.0101	800	0.0088	-
1.0354	820	0.0053	-
1.0606	840	0.0049	-
1.0859	860	0.0070	-
1.1111	880	0.0038	-
1.1364	900	0.0037	-
1.1616	920	0.0041	-
1.1869	940	0.0028	-
1.2121	960	0.0111	-
1.2374	980	0.0041	-
1.2626	1000	0.0040	-
1.2879	1020	0.0043	-
1.3131	1040	0.0034	-
1.3384	1060	0.0068	-
1.3636	1080	0.0040	-
1.3889	1100	0.0027	-
1.4141	1120	0.0036	-
1.4394	1140	0.0047	-
1.4646	1160	0.0051	-
1.4899	1180	0.0073	-
1.5152	1200	0.0052	-
1.5404	1220	0.0034	-
1.5657	1240	0.0056	-
1.5909	1260	0.0054	-
1.6162	1280	0.0100	-
1.6414	1300	0.0049	-
1.6667	1320	0.0034	-
1.6919	1340	0.0045	-
1.7172	1360	0.0027	-
1.7424	1380	0.0036	-
1.7677	1400	0.0024	-
1.7929	1420	0.0071	-
1.8182	1440	0.0050	-
1.8434	1460	0.0020	-
1.8687	1480	0.0028	-
1.8939	1500	0.0065	-
1.9192	1520	0.0028	-
1.9444	1540	0.0051	-
1.9697	1560	0.0075	-
1.9949	1580	0.0034	-
2.0	1584	-	0.9955

Training Time

Training: 7.6 minutes

Framework Versions

Python: 3.12.13
Sentence Transformers: 5.6.0
Transformers: 5.12.1
PyTorch: 2.11.0+cu128
Accelerate: 1.14.0
Datasets: 4.0.0
Tokenizers: 0.22.2

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