This is a sentence-transformers model finetuned from sentence-transformers/all-MiniLM-L6-v2. 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: sentence-transformers/all-MiniLM-L6-v2
Maximum Sequence Length: 256 tokens
Output Dimensionality: 384 dimensions
Similarity Function: Cosine Similarity

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': 256, 'do_lower_case': False, 'architecture': '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.

python

from sentence_transformers import SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("sentence_transformers_model_id")
# Run inference
sentences = [
    '[OFFER] Лен | 10.0 упаковка',
    '[ORDER] Лен | 10.0 упаковка',
    '[OFFER] Арматура 8А500С L=11,7 м | 8.5 т',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[ 1.0000,  0.9970, -0.0349],
#         [ 0.9970,  1.0000, -0.0279],
#         [-0.0349, -0.0279,  1.0000]])

Evaluation

Metrics

Triplet

Dataset: goods-val
Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	0.9561

Training Details

Training Dataset

Unnamed Dataset

Size: 204,810 training samples
Columns: sentence_0 and sentence_1
Approximate statistics based on the first 1000 samples:
sentence_0 sentence_1
type string string
details
min: 13 tokens
mean: 40.47 tokens
max: 256 tokens
min: 14 tokens
mean: 43.67 tokens
max: 256 tokens

Samples:

sentence_0	sentence_1
`[OFFER] 231625 Болт \| 24.0 шт`	`[ORDER] Болт для крепления углового карданного вала МАНИТУ МТ732 231625 \| 24.0 шт`
`[OFFER] BPL2930 Профиль П-обр. PSL L3000 1,5мм ДКС \| 30.0 м`	`[ORDER] Профиль П-образный PSL 29х48х3000 1.5 мм (BPL2930) / DKC \| 30.0 м`
`[OFFER] Счетчик электроэнергии трехфазный многотарифный (2 тарифа) Меркурий-231 АТ-011 5-60А IrDA DIN ЖКИ Инкотекс \| 2.0 шт`	`[ORDER] Электросчетчик "Меркурий 231AT-01 I", мн.т. ЖКИ, IrDA 5(60)A \| 2.0 шт`

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

per_device_train_batch_size: 24
num_train_epochs: 4
eval_strategy: steps
per_device_eval_batch_size: 24
multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

per_device_train_batch_size: 24
num_train_epochs: 4
max_steps: -1
learning_rate: 5e-05
lr_scheduler_type: linear
lr_scheduler_kwargs: None
warmup_steps: 0
optim: adamw_torch
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
label_smoothing_factor: 0.0
bf16: False
fp16: False
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: trackio
eval_strategy: steps
per_device_eval_batch_size: 24
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: False
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_backend: None
ddp_timeout: 1800
fsdp: []
fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}
deepspeed: None
debug: []
skip_memory_metrics: True
do_predict: False
resume_from_checkpoint: None
warmup_ratio: None
local_rank: -1
prompts: None
batch_sampler: batch_sampler
multi_dataset_batch_sampler: round_robin
router_mapping: {}
learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	goods-val_cosine_accuracy
0.0586	500	0.8426	-
0.1172	1000	0.2536	-
0.1758	1500	0.2174	-
0.2344	2000	0.1768	-
0.2929	2500	0.1701	-
0.3515	3000	0.1486	-
0.4101	3500	0.1418	-
0.4687	4000	0.1197	-
0.5	4267	-	0.9440
0.5273	4500	0.1159	-
0.5859	5000	0.1029	-
0.6445	5500	0.0910	-
0.7031	6000	0.0887	-
0.7617	6500	0.0819	-
0.8202	7000	0.0709	-
0.8788	7500	0.0708	-
0.9374	8000	0.0708	-
0.9960	8500	0.0608	-
1.0	8534	-	0.9530
1.0546	9000	0.0520	-
1.1132	9500	0.0492	-
1.1718	10000	0.0520	-
1.2304	10500	0.0487	-
1.2890	11000	0.0503	-
1.3476	11500	0.0455	-
1.4061	12000	0.0444	-
1.4647	12500	0.0454	-
1.5	12801	-	0.9550
1.5233	13000	0.0421	-
1.5819	13500	0.0412	-
1.6405	14000	0.0449	-
1.6991	14500	0.0374	-
1.7577	15000	0.0358	-
1.8163	15500	0.0360	-
1.8749	16000	0.0349	-
1.9334	16500	0.0386	-
1.9920	17000	0.0365	-
2.0	17068	-	0.9520
2.0506	17500	0.0287	-
2.1092	18000	0.0336	-
2.1678	18500	0.0274	-
2.2264	19000	0.0256	-
2.2850	19500	0.0278	-
2.3436	20000	0.0240	-
2.4022	20500	0.0276	-
2.4607	21000	0.0266	-
2.5	21335	-	0.9557
2.5193	21500	0.0274	-
2.5779	22000	0.0240	-
2.6365	22500	0.0249	-
2.6951	23000	0.0253	-
2.7537	23500	0.0244	-
2.8123	24000	0.0256	-
2.8709	24500	0.0261	-
2.9295	25000	0.0225	-
2.9880	25500	0.0231	-
3.0	25602	-	0.9549
3.0466	26000	0.0250	-
3.1052	26500	0.0201	-
3.1638	27000	0.0159	-
3.2224	27500	0.0206	-
3.2810	28000	0.0230	-
3.3396	28500	0.0193	-
3.3982	29000	0.0185	-
3.4568	29500	0.0178	-
3.5	29869	-	0.9557
3.5154	30000	0.0209	-
3.5739	30500	0.0193	-
3.6325	31000	0.0194	-
3.6911	31500	0.0170	-
3.7497	32000	0.0191	-
3.8083	32500	0.0161	-
3.8669	33000	0.0190	-
3.9255	33500	0.0197	-
3.9841	34000	0.0189	-
4.0	34136	-	0.9561

Framework Versions

Python: 3.11.10
Sentence Transformers: 5.2.3
Transformers: 5.2.0
PyTorch: 2.5.1+cu124
Accelerate: 1.12.0
Datasets: 4.6.1
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