SentenceTransformer based on intfloat/multilingual-e5-base

This is a sentence-transformers model finetuned from intfloat/multilingual-e5-base on the rztk/rozetka_positive_pairs dataset. 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: intfloat/multilingual-e5-base
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 tokens
• Similarity Function: Cosine Similarity
• Training Dataset:
  • rztk/rozetka_positive_pairs

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: XLMRobertaModel 
  (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})
  (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("sentence_transformers_model_id")
# Run inference
sentences = [
    'поилка для детей',
    "<category>Поїльники та непроливайки</category><brand>Nuk</brand><options><option_title>Стать дитини</option_title><option_value>Хлопчик</option_value><option_title>Стать дитини</option_title><option_value>Дівчинка</option_value><option_title>Кількість вантажних місць</option_title><option_value>1</option_value><option_title>Країна реєстрації бренда</option_title><option_value>Німеччина</option_value><option_title>Країна-виробник товару</option_title><option_value>Німеччина</option_value><option_title>Об'єм, мл</option_title><option_value>300</option_value><option_title>Матеріал</option_title><option_value>Пластик</option_value><option_title>Колір</option_title><option_value>Блакитний</option_value><option_title>Тип</option_title><option_value>Поїльник</option_value><option_title>Тип гарантійного талона</option_title><option_value>Гарантія по чеку</option_value><option_title>Доставка Premium</option_title></options>",
    'Шафа розпашній Fenster Оксфорд Лагуна',
]
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]

Evaluation

Metrics

Information Retrieval

• Dataset: rusisms-uk-title
• Evaluated with InformationRetrievalEvaluator

Information Retrieval

• Dataset: rusisms-uk-title--matryoshka_dim-768--
• Evaluated with InformationRetrievalEvaluator

Information Retrieval

• Dataset: rusisms-uk-title--matryoshka_dim-512--
• Evaluated with InformationRetrievalEvaluator

Information Retrieval

• Dataset: rusisms-uk-title--matryoshka_dim-256--
• Evaluated with InformationRetrievalEvaluator

Information Retrieval

• Dataset: rusisms-uk-title--matryoshka_dim-128--
• Evaluated with InformationRetrievalEvaluator

Training Details

Training Dataset

rztk/rozetka_positive_pairs

• Dataset: rztk/rozetka_positive_pairs
• Size: 44,800 training samples
• Columns: query and text
• Approximate statistics based on the first 1000 samples:

  	query	text
  type	string	string
  details	min: 3 tokensmean: 7.18 tokensmax: 16 tokens	min: 9 tokensmean: 158.88 tokensmax: 512 tokens

• Samples:

  query	text
  p smart z	TPU чехол Ultrathin Series 0,33 mm для Huawei P Smart Z Безбарвний (прозорий)
  p smart z	Чохли для мобільних телефонів<option_title>Матеріал</option_title><option_value>Силікон</option_value><option_title>Колір</option_title><option_value>Transparent</option_value><option_title>Сумісна модель</option_title><option_value>P Smart Z</option_value>
  p smart z	TPU чехол Ultrathin Series 0,33mm для Huawei P Smart Z Бесцветный (прозрачный)

• Loss: sentence_transformers_training.model.matryoshka2d_loss.RZTKMatryoshka2dLoss with these parameters:

  {
      "loss": "RZTKMultipleNegativesRankingLoss",
      "n_layers_per_step": 1,
      "last_layer_weight": 1.0,
      "prior_layers_weight": 1.0,
      "kl_div_weight": 1.0,
      "kl_temperature": 0.3,
      "matryoshka_dims": [
          768,
          512,
          256,
          128
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": 1
  }
  

Evaluation Dataset

rztk/rozetka_positive_pairs

• Dataset: rztk/rozetka_positive_pairs
• Size: 4,480 evaluation samples
• Columns: query and text
• Approximate statistics based on the first 1000 samples:

  	query	text
  type	string	string
  details	min: 3 tokensmean: 6.29 tokensmax: 11 tokens	min: 12 tokensmean: 161.36 tokensmax: 512 tokens

• Samples:

  query	text
  кошелек женский	Портмоне BAELLERRY Forever N2345 Черный (020354)
  кошелек женский	ГаманціBaellerry<option_title>Для кого</option_title><option_value>Для жінок</option_value><option_title>Вид</option_title><option_value>Портмоне</option_value><option_title>Матеріал</option_title><option_value>Штучна шкіра</option_value><option_title>Країна-виробник товару</option_title><option_value>Китай</option_value>
  кошелек женский	Портмоне BAELLERRY Forever N2345 Черный (020354)

• Loss: sentence_transformers_training.model.matryoshka2d_loss.RZTKMatryoshka2dLoss with these parameters:

  {
      "loss": "RZTKMultipleNegativesRankingLoss",
      "n_layers_per_step": 1,
      "last_layer_weight": 1.0,
      "prior_layers_weight": 1.0,
      "kl_div_weight": 1.0,
      "kl_temperature": 0.3,
      "matryoshka_dims": [
          768,
          512,
          256,
          128
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": 1
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 112
• per_device_eval_batch_size: 112
• torch_empty_cache_steps: 30
• learning_rate: 2e-05
• num_train_epochs: 1.0
• warmup_ratio: 0.1
• bf16: True
• bf16_full_eval: True
• tf32: True
• dataloader_num_workers: 2
• load_best_model_at_end: True
• optim: adafactor
• push_to_hub: True

All Hyperparameters

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 112
• per_device_eval_batch_size: 112
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 1
• eval_accumulation_steps: None
• torch_empty_cache_steps: 30
• 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: 1.0
• max_steps: -1
• lr_scheduler_type: linear
• 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: True
• fp16: False
• fp16_opt_level: O1
• half_precision_backend: auto
• bf16_full_eval: True
• fp16_full_eval: False
• tf32: True
• local_rank: 0
• ddp_backend: None
• tpu_num_cores: None
• tpu_metrics_debug: False
• debug: []
• dataloader_drop_last: True
• dataloader_num_workers: 2
• dataloader_prefetch_factor: None
• past_index: -1
• disable_tqdm: False
• remove_unused_columns: True
• label_names: None
• load_best_model_at_end: True
• 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: adafactor
• 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: True
• 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
• eval_on_start: False
• use_liger_kernel: False
• eval_use_gather_object: False
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional
• ddp_static_graph: False
• ddp_comm_hook: bf16
• gradient_as_bucket_view: False

Training Logs

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.12.6
• Sentence Transformers: 3.0.1
• Transformers: 4.45.1
• PyTorch: 2.4.1
• Accelerate: 0.34.2
• Datasets: 3.0.0
• Tokenizers: 0.20.0

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",
}