leafxyz/main_v2

SentenceTransformer based on prestoai/qwen3-embedding-0.6b-arabic-ecom

This is a sentence-transformers model finetuned from prestoai/qwen3-embedding-0.6b-arabic-ecom on the pairs_with_negatives and positives datasets. It maps sentences & paragraphs to a 1024-dimensional dense vector space and can be used for retrieval.

Model Details

Model Description

• Model Type: Sentence Transformer
• Base model: prestoai/qwen3-embedding-0.6b-arabic-ecom
• Maximum Sequence Length: 128 tokens
• Output Dimensionality: 1024 dimensions
• Similarity Function: Cosine Similarity
• Supported Modality: Text
• Training Datasets:
  • pairs_with_negatives
  • positives

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': 'Qwen3Model'})
  (1): Pooling({'embedding_dimension': 1024, 'pooling_mode': 'lasttoken', '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("leafxyz/main_v2")
# Run inference
queries = [
    'Instruct: Given an Arabic e-commerce search query, retrieve the product that best matches it\nQuery: عطور نسائية',
]
documents = [
    'مجموعة عطر نسائي - ابراهيم القرشي سكر',
    'قبعة رجالية - 07',
    'تن الوفاء سكيب جاك بالزيت الزيتون - 160 غ',
]
query_embeddings = model.encode_query(queries)
document_embeddings = model.encode_document(documents)
print(query_embeddings.shape, document_embeddings.shape)
# [1, 1024] [3, 1024]

# Get the similarity scores for the embeddings
similarities = model.similarity(query_embeddings, document_embeddings)
print(similarities)
# tensor([[ 0.5593, -0.0069,  0.0400]])

Training Details

Training Datasets

pairs_with_negatives

• Dataset: pairs_with_negatives
• Size: 9,900 training samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 24 tokens mean: 29.36 tokens max: 42 tokens	min: 3 tokens mean: 15.82 tokens max: 38 tokens	min: 2 tokens mean: 15.12 tokens max: 44 tokens

• Samples:

  anchor	positive	negative
  Instruct: Given an Arabic e-commerce search query, retrieve the product that best matches it Query: زيت بابايا WKL	زيت جسم - WKL Papaya	زيت جسم - Vaseline Cocoa
  Instruct: Given an Arabic e-commerce search query, retrieve the product that best matches it Query: منكير	اظافر هيفا - TWINKLE	اظافر هيفا - SPARKLE
  Instruct: Given an Arabic e-commerce search query, retrieve the product that best matches it Query: توب فريش حفاضات	توب فريش حفاضات رقم 1 - 44 قطعة	توب فريش حفاضات رقم 2 - 40 قطعة

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim",
      "gather_across_devices": false,
      "directions": [
          "query_to_doc"
      ],
      "partition_mode": "joint",
      "hardness_mode": null,
      "hardness_strength": 0.0
  }
  

positives

• Dataset: positives
• Size: 39,600 training samples
• Columns: anchor and positive
• Approximate statistics based on the first 1000 samples:

  	anchor	positive
  type	string	string
  details	min: 23 tokens mean: 29.52 tokens max: 41 tokens	min: 3 tokens mean: 13.77 tokens max: 39 tokens

• Samples:

  anchor	positive
  Instruct: Given an Arabic e-commerce search query, retrieve the product that best matches it Query: نبي فيكسول ارجواني	منظف ​​الحمام الذكي فيكسول ارجواني - 900 مل
  Instruct: Given an Arabic e-commerce search query, retrieve the product that best matches it Query: شربة نجمة اريغي 500	شربة نجمة اريغي - 500 غ
  Instruct: Given an Arabic e-commerce search query, retrieve the product that best matches it Query: Gas relife drops	Gas relife drops

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim",
      "gather_across_devices": false,
      "directions": [
          "query_to_doc"
      ],
      "partition_mode": "joint",
      "hardness_mode": null,
      "hardness_strength": 0.0
  }
  

Evaluation Datasets

pairs_with_negatives

• Dataset: pairs_with_negatives
• Size: 100 evaluation samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 100 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 24 tokens mean: 29.35 tokens max: 43 tokens	min: 5 tokens mean: 15.71 tokens max: 31 tokens	min: 4 tokens mean: 15.2 tokens max: 34 tokens

• Samples:

  anchor	positive	negative
  Instruct: Given an Arabic e-commerce search query, retrieve the product that best matches it Query: عناية بالجسم	معطر جسم وشعر نسائي - Sol de Janeiro Água Mística	قارورة عصير - AS02
  Instruct: Given an Arabic e-commerce search query, retrieve the product that best matches it Query: بخاخ تشيكو 100 مل	بخاخ تشيكو للحماية من البعوض - 100 مل	مناديل الحماية من البعوض تشيكو - 20 قطعة
  Instruct: Given an Arabic e-commerce search query, retrieve the product that best matches it Query: بخاخ مانع التصاق	بخاخ الطبخ بنكهة الفلفل مانع للالتصاق - 200 مل	بخاخ الطبخ بنكهة الثوم مانع للالتصاق - 200 مل

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim",
      "gather_across_devices": false,
      "directions": [
          "query_to_doc"
      ],
      "partition_mode": "joint",
      "hardness_mode": null,
      "hardness_strength": 0.0
  }
  

positives

• Dataset: positives
• Size: 400 evaluation samples
• Columns: anchor and positive
• Approximate statistics based on the first 400 samples:

  	anchor	positive
  type	string	string
  details	min: 24 tokens mean: 29.43 tokens max: 43 tokens	min: 3 tokens mean: 13.74 tokens max: 35 tokens

• Samples:

  anchor	positive
  Instruct: Given an Arabic e-commerce search query, retrieve the product that best matches it Query: سوار نسائي ذهبي	سوار نسائي - DX052
  Instruct: Given an Arabic e-commerce search query, retrieve the product that best matches it Query: سناكس	شوكلاتة كندر ترونكي 8*48
  Instruct: Given an Arabic e-commerce search query, retrieve the product that best matches it Query: مشروب حليب	حليب - Safi

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim",
      "gather_across_devices": false,
      "directions": [
          "query_to_doc"
      ],
      "partition_mode": "joint",
      "hardness_mode": null,
      "hardness_strength": 0.0
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• gradient_accumulation_steps: 4
• learning_rate: 3e-05
• num_train_epochs: 1
• warmup_steps: 0.05
• fp16: True
• dataloader_num_workers: 2
• gradient_checkpointing: True

All Hyperparameters

Click to expand

• do_predict: False
• prediction_loss_only: True
• per_device_train_batch_size: 8
• per_device_eval_batch_size: 8
• gradient_accumulation_steps: 4
• eval_accumulation_steps: None
• torch_empty_cache_steps: None
• learning_rate: 3e-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: linear
• lr_scheduler_kwargs: None
• warmup_ratio: None
• warmup_steps: 0.05
• log_level: passive
• log_level_replica: warning
• log_on_each_node: True
• logging_nan_inf_filter: True
• enable_jit_checkpoint: False
• save_on_each_node: False
• save_only_model: False
• restore_callback_states_from_checkpoint: False
• use_cpu: False
• seed: 42
• data_seed: None
• bf16: False
• fp16: True
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: None
• local_rank: -1
• ddp_backend: None
• debug: []
• dataloader_drop_last: False
• dataloader_num_workers: 2
• dataloader_prefetch_factor: None
• disable_tqdm: False
• remove_unused_columns: True
• label_names: None
• load_best_model_at_end: False
• ignore_data_skip: False
• fsdp: []
• fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': 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
• deepspeed: None
• label_smoothing_factor: 0.0
• optim: adamw_torch_fused
• optim_args: None
• 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
• push_to_hub: False
• 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: True
• gradient_checkpointing_kwargs: None
• include_for_metrics: []
• eval_do_concat_batches: True
• auto_find_batch_size: False
• full_determinism: False
• ddp_timeout: 1800
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• 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
• use_cache: False
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	pairs with negatives loss	positives loss
0.0323	50	0.2092	-	-
0.0646	100	0.2008	-	-
0.0970	150	0.1948	-	-
0.1293	200	0.1749	-	-
0.1616	250	0.1455	-	-
0.1939	300	0.1924	-	-
0.2262	350	0.1959	-	-
0.2586	400	0.1606	-	-
0.2909	450	0.1679	-	-
0.3232	500	0.1774	0.3304	0.1113
0.3555	550	0.1924	-	-
0.3878	600	0.1487	-	-
0.4202	650	0.1859	-	-
0.4525	700	0.1807	-	-
0.4848	750	0.1785	-	-
0.5171	800	0.1534	-	-
0.5495	850	0.1468	-	-
0.5818	900	0.1566	-	-
0.6141	950	0.1153	-	-
0.6464	1000	0.1322	0.3138	0.0943
0.6787	1050	0.1320	-	-
0.7111	1100	0.1533	-	-
0.7434	1150	0.1358	-	-
0.7757	1200	0.1457	-	-
0.8080	1250	0.1320	-	-
0.8403	1300	0.1680	-	-
0.8727	1350	0.1280	-	-
0.9050	1400	0.1632	-	-
0.9373	1450	0.1656	-	-
0.9696	1500	0.1363	0.3024	0.0914

Training Time

• Training: 1.9 hours

Framework Versions

• Python: 3.12.13
• Sentence Transformers: 5.4.1
• Transformers: 5.0.0
• PyTorch: 2.10.0+cu128
• Accelerate: 1.13.0
• Datasets: 5.0.0
• Tokenizers: 0.22.2

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",
}

MultipleNegativesRankingLoss

@misc{oord2019representationlearningcontrastivepredictive,
      title={Representation Learning with Contrastive Predictive Coding},
      author={Aaron van den Oord and Yazhe Li and Oriol Vinyals},
      year={2019},
      eprint={1807.03748},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/1807.03748},
}