This is a sentence-transformers model trained on the olive-phonetic dataset. It maps sentences & paragraphs to a 1024-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

Maximum Sequence Length: 512 tokens
Output Dimensionality: 1024 dimensions
Similarity Function: Cosine Similarity
Training Dataset:
- olive-phonetic

Model Sources

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

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': False, 'architecture': 'PeftModelForFeatureExtraction'})
  (1): Pooling({'word_embedding_dimension': 1024, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, 'pooling_mode_max_tokens': False, 'pooling_mode_mean_sqrt_len_tokens': False, 'pooling_mode_weightedmean_tokens': False, 'pooling_mode_lasttoken': False, '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("dkqjrm/bge-m3-olive-phonetic-incremental-lora")
# Run inference
sentences = [
    '[운동복세탁] 에코두 프랑스 울세제 울샴푸 니트 속옷세제 750ml x 2개',
    '에코도',
    'バークレイ',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.7370, 0.0654],
#         [0.7370, 1.0000, 0.0482],
#         [0.0654, 0.0482, 1.0000]])

Training Details

Training Dataset

olive-phonetic

Dataset: olive-phonetic at 2edf5dd
Size: 348,352 training samples
Columns: anchor and positive
Approximate statistics based on the first 1000 samples:
anchor positive
type string string
details
min: 5 tokens
mean: 23.08 tokens
max: 50 tokens
min: 3 tokens
mean: 6.08 tokens
max: 14 tokens

Samples:

anchor	positive
`필립스 3000 시리즈 듀얼 에어케어 접이식 헤어드라이기 (BHD308/69)`	`飞利浦`
`밀크바오밥 퍼퓸 헤어 화이트머스크 선물세트 (샴푸500ml+트리트먼트500ml)`	`ミルクバオバブー`
`[집들이선물] 에코두 프랑스 청소세제 종합선물세트`	`에코두`

Loss: MultipleNegativesRankingLoss with these parameters:
json
1{ 2 "scale": 20.0, 3 "similarity_fct": "cos_sim", 4 "gather_across_devices": false 5}

Evaluation Dataset

olive-phonetic

Dataset: olive-phonetic at 2edf5dd
Size: 18,334 evaluation samples
Columns: anchor and positive
Approximate statistics based on the first 1000 samples:
anchor positive
type string string
details
min: 7 tokens
mean: 22.86 tokens
max: 52 tokens
min: 3 tokens
mean: 6.13 tokens
max: 14 tokens

Samples:

anchor	positive
`포렌코즈 타투 끌레르 벨벳 틴트`	`フォレンコス`
`텐바이텐 산리오 마이레터 6공 다이어리 (마이멜로디/쿠로미/시나모롤/포차코)`	`텐바이탠`
`RYMD 텐셀 퀵드라이 크로스 요가티`	`リムド`

Loss: MultipleNegativesRankingLoss with these parameters:
json
1{ 2 "scale": 20.0, 3 "similarity_fct": "cos_sim", 4 "gather_across_devices": false 5}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 16
gradient_accumulation_steps: 16
learning_rate: 1e-05
num_train_epochs: 1
lr_scheduler_type: cosine
warmup_ratio: 0.05
fp16: True
push_to_hub: True
batch_sampler: no_duplicates

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: steps
prediction_loss_only: True
per_device_train_batch_size: 16
per_device_eval_batch_size: 8
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 16
eval_accumulation_steps: None
torch_empty_cache_steps: None
learning_rate: 1e-05
weight_decay: 0.0
adam_beta1: 0.9
adam_beta2: 0.999
adam_epsilon: 1e-08
max_grad_norm: 1.0
num_train_epochs: 1
max_steps: -1
lr_scheduler_type: cosine
lr_scheduler_kwargs: None
warmup_ratio: 0.05
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
bf16: False
fp16: True
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}
parallelism_config: None
deepspeed: None
label_smoothing_factor: 0.0
optim: adamw_torch_fused
optim_args: None
adafactor: False
group_by_length: False
length_column_name: length
project: huggingface
trackio_space_id: trackio
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: True
resume_from_checkpoint: None
hub_model_id: None
hub_strategy: every_save
hub_private_repo: None
hub_always_push: False
hub_revision: None
gradient_checkpointing: False
gradient_checkpointing_kwargs: None
include_inputs_for_metrics: False
include_for_metrics: []
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
include_tokens_per_second: False
include_num_input_tokens_seen: no
neftune_noise_alpha: None
optim_target_modules: None
batch_eval_metrics: False
eval_on_start: False
use_liger_kernel: False
liger_kernel_config: None
eval_use_gather_object: False
average_tokens_across_devices: True
prompts: None
batch_sampler: no_duplicates
multi_dataset_batch_sampler: proportional
router_mapping: {}
learning_rate_mapping: {}

Training Logs

Click to expand

Epoch	Step	Training Loss	Validation Loss
0.0073	10	2.0959	-
0.0147	20	2.0514	-
0.0220	30	1.8141	-
0.0294	40	1.6118	-
0.0367	50	1.2453	-
0.0441	60	0.8385	-
0.0514	70	0.6052	-
0.0588	80	0.4456	-
0.0661	90	0.4206	-
0.0735	100	0.3856	0.2105
0.0808	110	0.3858	-
0.0882	120	0.3064	-
0.0955	130	0.3153	-
0.1029	140	0.2906	-
0.1102	150	0.2974	-
0.1176	160	0.293	-
0.1249	170	0.2546	-
0.1323	180	0.267	-
0.1396	190	0.258	-
0.1470	200	0.2742	0.1423
0.1543	210	0.249	-
0.1617	220	0.2486	-
0.1690	230	0.2543	-
0.1764	240	0.249	-
0.1837	250	0.2429	-
0.1911	260	0.2167	-
0.1984	270	0.2419	-
0.2058	280	0.2214	-
0.2131	290	0.2102	-
0.2205	300	0.201	0.1156
0.2278	310	0.2205	-
0.2352	320	0.2109	-
0.2425	330	0.1933	-
0.2499	340	0.2008	-
0.2572	350	0.2041	-
0.2646	360	0.1981	-
0.2719	370	0.2193	-
0.2793	380	0.2111	-
0.2866	390	0.1794	-
0.2940	400	0.1895	0.0982
0.3013	410	0.1997	-
0.3087	420	0.1683	-
0.3160	430	0.1786	-
0.3234	440	0.1811	-
0.3307	450	0.1785	-
0.3380	460	0.1811	-
0.3454	470	0.1933	-
0.3527	480	0.1774	-
0.3601	490	0.1677	-
0.3674	500	0.1787	0.0855
0.3748	510	0.1772	-
0.3821	520	0.1551	-
0.3895	530	0.1788	-
0.3968	540	0.1583	-
0.4042	550	0.1529	-
0.4115	560	0.1691	-
0.4189	570	0.154	-
0.4262	580	0.1592	-
0.4336	590	0.166	-
0.4409	600	0.163	0.0780
0.4483	610	0.1466	-
0.4556	620	0.1579	-
0.4630	630	0.1551	-
0.4703	640	0.142	-
0.4777	650	0.1837	-
0.4850	660	0.1494	-
0.4924	670	0.1582	-
0.4997	680	0.1438	-
0.5071	690	0.1387	-
0.5144	700	0.1682	0.0726
0.5218	710	0.1507	-
0.5291	720	0.1853	-
0.5365	730	0.1392	-
0.5438	740	0.1422	-
0.5512	750	0.1393	-
0.5585	760	0.154	-
0.5659	770	0.1375	-
0.5732	780	0.1405	-
0.5806	790	0.1483	-
0.5879	800	0.135	0.0690
0.5953	810	0.1276	-
0.6026	820	0.142	-
0.6100	830	0.1368	-
0.6173	840	0.1397	-
0.6247	850	0.1354	-
0.6320	860	0.1397	-
0.6394	870	0.1289	-
0.6467	880	0.1596	-
0.6541	890	0.1266	-
0.6614	900	0.1394	0.0666
0.6687	910	0.1434	-
0.6761	920	0.1358	-
0.6834	930	0.1301	-
0.6908	940	0.1232	-
0.6981	950	0.1333	-
0.7055	960	0.1554	-
0.7128	970	0.14	-
0.7202	980	0.1367	-
0.7275	990	0.1397	-
0.7349	1000	0.1486	0.0646
0.7422	1010	0.1126	-
0.7496	1020	0.1432	-
0.7569	1030	0.1234	-
0.7643	1040	0.1583	-
0.7716	1050	0.1274	-
0.7790	1060	0.1314	-
0.7863	1070	0.1163	-
0.7937	1080	0.1512	-
0.8010	1090	0.1392	-
0.8084	1100	0.1401	0.0638
0.8157	1110	0.1366	-
0.8231	1120	0.1471	-
0.8304	1130	0.1341	-
0.8378	1140	0.1495	-
0.8451	1150	0.1297	-
0.8525	1160	0.146	-
0.8598	1170	0.1431	-
0.8672	1180	0.1487	-
0.8745	1190	0.1291	-
0.8819	1200	0.1225	0.0631
0.8892	1210	0.1291	-
0.8966	1220	0.1232	-
0.9039	1230	0.1187	-
0.9113	1240	0.1662	-
0.9186	1250	0.1395	-
0.9260	1260	0.1308	-
0.9333	1270	0.1493	-
0.9407	1280	0.1186	-
0.9480	1290	0.1318	-
0.9554	1300	0.1364	0.0630
0.9627	1310	0.1356	-
0.9701	1320	0.1458	-
0.9774	1330	0.1591	-
0.9848	1340	0.1272	-
0.9921	1350	0.1166	-
0.9994	1360	0.1259	-

Framework Versions

Python: 3.12.12
Sentence Transformers: 5.2.1
Transformers: 4.57.6
PyTorch: 2.10.0+cu128
Accelerate: 1.12.0
Datasets: 4.3.0
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

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}
}

	anchor	positive
type	string	string
details	min: 5 tokens mean: 23.08 tokens max: 50 tokens	min: 3 tokens mean: 6.08 tokens max: 14 tokens