am-azadi/bilingual-embedding-large_Fine_Tuned_3e

SentenceTransformer based on am-azadi/bilingual-embedding-large_Fine_Tuned_2e

This is a sentence-transformers model finetuned from am-azadi/bilingual-embedding-large_Fine_Tuned_2e. 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
• Base model: am-azadi/bilingual-embedding-large_Fine_Tuned_2e
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 1024 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': 512, 'do_lower_case': False}) with Transformer model: BilingualModel 
  (1): Pooling({'word_embedding_dimension': 1024, '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 = [
    'Uuuuu mepa that they killed the real bald guy EXCLUSIVE What are you doing bald, go getting into it jonca that the 12 wants to take pictures with you at any time 14:04 ✓ they found 2 contact cards for this number re add them to your contacts? T SEE CONTACT CARDS CELL PHONES TURNED OFF THEY LOOK FOR THEM EVERYWHERE THE "12" IS LOOKING FOR THEM "serne HD MRASSIA LEGAL: 11 2159 6256 FOR POLICE COMPLAINTS: 11 2159 6256 HERE FOR POLICE COMPLAINTS: 11-',
    'They find Diego Molina murdered in his apartment, the skinny from the funeral home who took photos with Diego Armando Maradona The images of a lacerated body are not of the person who was photographed with the corpse of Maradona',
    'The elected mayor of Medellín does not like ESMAD. WHY WILL IT BE? The original video shows Daniel Quintero in a demonstration against violence in Bogotá',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]

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

Training Details

Training Dataset

Unnamed Dataset

• Size: 21,769 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: 4 tokens mean: 122.97 tokens max: 512 tokens	min: 17 tokens mean: 38.24 tokens max: 109 tokens

• Samples:

  sentence_0	sentence_1
  NEW HANDLING OF ALERT While the achieves 6,101,968 votes (i.e. 26.8%), the Ministry of the Interior only gives it 5,836,202 votes (i.e. 25.7%) to artificially make 's party appear in the lead . Hello Council of State?	The Ministry of the Interior manipulated the results of the legislative elections Legislative: why are the results of the 1st round contested by the Nupes?
  <3<3... Civil Registry Offices in Brazil: The only source that does not lie, as it issues all death certificates daily, for all reasons. This source cannot be disputed by anyone. Only they can say for sure, how many people die each day, and the reason for death. The rest is fake news. Via Jose Mendes Junior Updating... Deaths in Brazil: July 2019 - 119,390 (without pandemic) July 2020 - 113,475 (with pandemic) Source: transparencia.registrocivil.org.br... Now what are they going to say????	More deaths were recorded in Brazil in July 2019, before the pandemic, than in July 2020, during the new coronavirus pandemic. Publications use partial data on deaths recorded in July 2020
  Zimbabwe Police are taking disciplinary action with a church that refused to take closure instructions to prevent the spread of Coronavirus.	Worshipers beaten in Zimbabwe for failing to comply with coronavirus assembly ban No, worshipers have not been beaten by police in Zimbabwe for gathering during the coronavirus outbreak

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim"
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 2
• per_device_eval_batch_size: 2
• num_train_epochs: 1
• multi_dataset_batch_sampler: round_robin

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: no
• prediction_loss_only: True
• per_device_train_batch_size: 2
• per_device_eval_batch_size: 2
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 1
• eval_accumulation_steps: None
• torch_empty_cache_steps: None
• learning_rate: 5e-05
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1
• num_train_epochs: 1
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.0
• 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: 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}
• deepspeed: None
• label_smoothing_factor: 0.0
• optim: adamw_torch
• 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: None
• hub_always_push: False
• 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
• 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
• average_tokens_across_devices: False
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: round_robin

Training Logs

Epoch	Step	Training Loss
0.0459	500	0.0148
0.0919	1000	0.0066
0.1378	1500	0.0245
0.1837	2000	0.0184
0.2297	2500	0.0174
0.2756	3000	0.0053
0.3215	3500	0.025
0.3675	4000	0.0105
0.4134	4500	0.0054
0.4593	5000	0.0076
0.5053	5500	0.0085
0.5512	6000	0.0104
0.5972	6500	0.0208
0.6431	7000	0.0072
0.6890	7500	0.0084
0.7350	8000	0.0053
0.7809	8500	0.0052
0.8268	9000	0.0064
0.8728	9500	0.0074
0.9187	10000	0.0083
0.9646	10500	0.008

Framework Versions

• Python: 3.11.11
• Sentence Transformers: 3.4.1
• Transformers: 4.48.3
• PyTorch: 2.5.1+cu124
• Accelerate: 1.3.0
• Datasets: 3.3.2
• Tokenizers: 0.21.0

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",
}

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}
}