metadata

base_model: intfloat/multilingual-e5-small
language:
  - multilingual
library_name: sentence-transformers
license: apache-2.0
metrics:
  - cosine_accuracy@1
  - cosine_accuracy@3
  - cosine_accuracy@5
  - cosine_accuracy@10
  - cosine_precision@1
  - cosine_precision@3
  - cosine_precision@5
  - cosine_precision@10
  - cosine_recall@1
  - cosine_recall@3
  - cosine_recall@5
  - cosine_recall@10
  - cosine_ndcg@10
  - cosine_mrr@10
  - cosine_map@100
pipeline_tag: sentence-similarity
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:2320
  - loss:MatryoshkaLoss
  - loss:MultipleNegativesRankingLoss
widget:
  - source_sentence: |-
      MVGO; medium vacuum
      gas oil
    sentences:
      - 과분해
      - Medium Vacuum Gas Oil(MVGO) ;
      - |-
        선적 전 또는 양하 후에 화물창에 잔존하는 소량의 액체화물 양을 결정하는 수학
        적인 계산 수식
  - source_sentence: PLE; plain large end
    sentences:
      - Plain Large End ;
      - |-
        부하중 변압기 Tap 변환기 ;
        변압기 권선의 Tap을 무정전으로 변경하는 장치
      - Cone Roof Tank에서 Tank내의 Vapor가 외부로 나갈 수 있도록 만들어 놓은 구멍
  - source_sentence: Fluidization
    sentences:
      - |-
        핵심성과지표;
        어떤 계획이나 목표가 성공하였는지 또는 성공하고 있는지를 확인하려면 그 성공
        을 구성하는 요소들을 측정하는 지표를 찾아 측정하여야 하는데, 이들 지표 중 성
        공을 확인할 수 있는 가장 결정적인 지표를 KPI라고 부릅니다.
      - |-
        전압변동에 영향을 주는 무효전력을 줄이기 위한 조상설비의 일종으로 정지형 무
        효전력 보상장치
      - 고체층을 액체나 기체로 확대시키거나 현탁시켜 유통하도록 하는 것
  - source_sentence: |-
      SH; surface hardened
      steel body
    sentences:
      - Surface Hardened Steel Body ;
      - 분산제 ; 슬러지 생성을 방지하기 위하여 Oil에 넣어주는 약품
      - |-
        작업위험성평가;
        현장에서 수행되는 작업을 포함한 전반적인 직무 활동에 대하여 위험요인을 분석
        하여 현재 안전조치를 검토하고 안전대책을 마련하는 기법
  - source_sentence: U-205200
    sentences:
      - 물속의 (-)ion을 OH-로 치환해 주는 이온교환수지탑
      - 차단기, 스위치류 , 스위치
      - 올레핀 송유/동력 Nitrogen Section
model-index:
  - name: Multilingual base soil embedding model (quantized)
    results:
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 256
          type: dim_256
        metrics:
          - type: cosine_accuracy@1
            value: 0.2441860465116279
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.31007751937984496
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.3643410852713178
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.4108527131782946
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.2441860465116279
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.10335917312661498
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.07286821705426358
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.041085271317829464
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.2441860465116279
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.31007751937984496
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.3643410852713178
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.4108527131782946
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.3172493867293268
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.28840746893072483
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.3003133446683658
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 128
          type: dim_128
        metrics:
          - type: cosine_accuracy@1
            value: 0.2054263565891473
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.28294573643410853
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.3178294573643411
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.38372093023255816
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.2054263565891473
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.09431524547803617
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.06356589147286822
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.03837209302325582
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.2054263565891473
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.28294573643410853
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.3178294573643411
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.38372093023255816
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.2850988708112555
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.25465270087363123
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.26532412971784447
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 64
          type: dim_64
        metrics:
          - type: cosine_accuracy@1
            value: 0.1937984496124031
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.2713178294573643
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.29844961240310075
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.3488372093023256
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.1937984496124031
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.0904392764857881
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.059689922480620154
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.03488372093023256
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.1937984496124031
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.2713178294573643
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.29844961240310075
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.3488372093023256
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.26467320016495083
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.2385474344776671
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.2482312240959752
            name: Cosine Map@100

Multilingual base soil embedding model (quantized)

This is a sentence-transformers model finetuned from intfloat/multilingual-e5-small. It maps sentences & paragraphs to a 384-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: intfloat/multilingual-e5-small
Maximum Sequence Length: 512 tokens
Output Dimensionality: 384 tokens
Similarity Function: Cosine Similarity
Language: multilingual
License: apache-2.0

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}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 384, '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})
  (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.

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("ValentinaKim/Multilingual-base-soil-embedding")
# Run inference
sentences = [
    'U-205200',
    '올레핀 송유/동력 Nitrogen Section',
    '차단기, 스위치류 , 스위치',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Information Retrieval

Dataset: dim_256
Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.2442
cosine_accuracy@3	0.3101
cosine_accuracy@5	0.3643
cosine_accuracy@10	0.4109
cosine_precision@1	0.2442
cosine_precision@3	0.1034
cosine_precision@5	0.0729
cosine_precision@10	0.0411
cosine_recall@1	0.2442
cosine_recall@3	0.3101
cosine_recall@5	0.3643
cosine_recall@10	0.4109
cosine_ndcg@10	0.3172
cosine_mrr@10	0.2884
cosine_map@100	0.3003

Information Retrieval

Dataset: dim_128
Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.2054
cosine_accuracy@3	0.2829
cosine_accuracy@5	0.3178
cosine_accuracy@10	0.3837
cosine_precision@1	0.2054
cosine_precision@3	0.0943
cosine_precision@5	0.0636
cosine_precision@10	0.0384
cosine_recall@1	0.2054
cosine_recall@3	0.2829
cosine_recall@5	0.3178
cosine_recall@10	0.3837
cosine_ndcg@10	0.2851
cosine_mrr@10	0.2547
cosine_map@100	0.2653

Information Retrieval

Dataset: dim_64
Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.1938
cosine_accuracy@3	0.2713
cosine_accuracy@5	0.2984
cosine_accuracy@10	0.3488
cosine_precision@1	0.1938
cosine_precision@3	0.0904
cosine_precision@5	0.0597
cosine_precision@10	0.0349
cosine_recall@1	0.1938
cosine_recall@3	0.2713
cosine_recall@5	0.2984
cosine_recall@10	0.3488
cosine_ndcg@10	0.2647
cosine_mrr@10	0.2385
cosine_map@100	0.2482

Training Details

Training Dataset

Unnamed Dataset

Size: 2,320 training samples
Columns: anchor and positive
Approximate statistics based on the first 1000 samples:
anchor positive
type string string
details
min: 3 tokens
mean: 6.72 tokens
max: 16 tokens

min: 3 tokens
mean: 35.77 tokens
max: 408 tokens

	anchor	positive
type	string	string
details	min: 3 tokens mean: 6.72 tokens max: 16 tokens	min: 3 tokens mean: 35.77 tokens max: 408 tokens

Samples:

anchor	positive
`Deionizer`	`탈이온장치 ; Demineralizer와 동일`
`Sub-CC; sub-contracting committee`	`외주 계약의 투명성과 공정성을 확보하기 위한 Sub-계약위원회로서 위원 및 위원 장은 CEO가 임명한다. CC이원원 부문장 이상 임원으로 하고 간사는 구매관리팀 장이 한다.`
`In-line Sampler`	`원유 속의 물과 침전물의 함량을 측정하기 위하여 원유하역 Line에 설치해 놓은 시료채취기`

Loss: MatryoshkaLoss with these parameters:

{
    "loss": "MultipleNegativesRankingLoss",
    "matryoshka_dims": [
        256,
        128,
        64
    ],
    "matryoshka_weights": [
        1,
        1,
        1
    ],
    "n_dims_per_step": -1
}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: epoch
per_device_train_batch_size: 32
per_device_eval_batch_size: 16
gradient_accumulation_steps: 16
learning_rate: 2e-05
num_train_epochs: 10
lr_scheduler_type: cosine
warmup_ratio: 0.1
tf32: False
optim: adamw_torch_fused
batch_sampler: no_duplicates

All Hyperparameters

Click to expand

overwrite_output_dir: False
do_predict: False
eval_strategy: epoch
prediction_loss_only: True
per_device_train_batch_size: 32
per_device_eval_batch_size: 16
per_gpu_train_batch_size: None
per_gpu_eval_batch_size: None
gradient_accumulation_steps: 16
eval_accumulation_steps: None
learning_rate: 2e-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: 10
max_steps: -1
lr_scheduler_type: cosine
lr_scheduler_kwargs: {}
warmup_ratio: 0.1
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: False
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_fused
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: proportional

Training Logs

Epoch	Step	Training Loss	dim_128_cosine_map@100	dim_256_cosine_map@100	dim_64_cosine_map@100
0.8767	4	-	0.2156	0.2448	0.1831
1.9726	9	-	0.2511	0.2765	0.2154
2.1918	10	7.6309	-	-	-
2.8493	13	-	0.2531	0.2852	0.2345
3.9452	18	-	0.2617	0.2914	0.2353
4.3836	20	5.3042	-	-	-
4.8219	22	-	0.2626	0.2946	0.2422
5.9178	27	-	0.2629	0.2987	0.2481
6.5753	30	4.2433	-	-	-
6.7945	31	-	0.2684	0.2988	0.2495
7.8904	36	-	0.2652	0.3003	0.2488
8.7671	40	3.9117	0.2653	0.3003	0.2482

Framework Versions

Python: 3.10.12
Sentence Transformers: 3.1.1
Transformers: 4.41.2
PyTorch: 2.1.2+cu121
Accelerate: 1.0.0
Datasets: 2.19.1
Tokenizers: 0.19.1

Citation

BibTeX

Sentence Transformers

@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",
}

MatryoshkaLoss

@misc{kusupati2024matryoshka,
    title={Matryoshka Representation Learning},
    author={Aditya Kusupati and Gantavya Bhatt and Aniket Rege and Matthew Wallingford and Aditya Sinha and Vivek Ramanujan and William Howard-Snyder and Kaifeng Chen and Sham Kakade and Prateek Jain and Ali Farhadi},
    year={2024},
    eprint={2205.13147},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

MultipleNegativesRankingLoss

@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}
}