am-azadi/bilingual-embedding-large_Fine_Tuned_1e

SentenceTransformer based on Lajavaness/bilingual-embedding-large

This is a sentence-transformers model finetuned from Lajavaness/bilingual-embedding-large. 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: Lajavaness/bilingual-embedding-large
• 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 = [
    'Even if you remove JIMENEZ ....... IF THERE IS STILL A SMARTMATIC THAT IS GOOD AT MAGIC THERE IS ALSO NO ..... SMARTMATIC SHOULD BE REMOVED FROM THE COMELEC CONTRACT ..... BECAUSE THE DEMON COMELEC HAS LONG HONORED THE VOTE OF MANY PEOPLE ..... AS LONG AS THERE ARE COMMISSIONERS IN THE COMELEC WHO LOOK LIKE MONEY, WE WILL NOT HAVE A CLEAN ELECTION ....... JUST IMAGINE HOW LONG THE ISSUE SPREADS THAT IF A CANDIDATE WANTS TO WIN, IT WILL PAY THE COMELEC 25 MILLION ???????????????????????????? ? SO ARE THE ELECTION RESULTS HOKOS POKOS ?????????????????????? DEMONS ...... SO ALL THE PUNISHMENT OF HEAVEN HAS BEEN GIVEN IN THE PHILIPPINES BECAUSE TANING LIVES WITH US ...... THE THOUGHT IS PURE MONEY ..... SO EVEN ELECTIONS ARE MONEY ..... ..... 7:08 AM 4G 51% FINALLY, COMELEC OFFICIAL JIMENEZ, REMOVED IN PLACE. BY PRRD AND OTHERS AGAIN THIS. FOR CLEAN NOW ELECTION TO COMING 2022 ELECTION',
    'Philippine President Rodrigo Duterte fired Comelec spokesman James Jimenez in May 2021 Posts misleadingly claim Philippine president fired poll body spokesman',
    'The WHO declared covid-19 an endemic disease Although it considers it probable, the WHO has not yet declared covid-19 an endemic disease',
]
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: 6 tokens mean: 120.55 tokens max: 512 tokens	min: 14 tokens mean: 38.75 tokens max: 148 tokens

• Samples:

  sentence_0	sentence_1
  "On January 1, 1979 New York billionaire Brandon Torrent allowed himself to be photographed while urinating on a homeless man sleeping on the street. This image explains, better than many words, the division of the world into social classes that we must eliminate . Meanwhile, in 21st century Brazil, many 'good citizens', just above the homeless condition, applaud politicians and politicians* who support the predatory elite represented by this abject and unworthy human being, who urinates on people who, in the final analysis, are the builders of the fortune he enjoys. Until we realize which side of this stream of urine we are on, we will not be able to build a truly just society. Class consciousness is the true and most urgent education."	This photo shows a billionaire named Brandon Torrent urinating on a homeless man The real story behind the image of a man who appears to urinate on a homeless person
  French secret service officer jean claude returns from his mission as imam with deash (isis) like others from several countries in Syria.. there are questions	This man is a French intelligence officer No, this man is not a French intelligence officer
  Oh yes! Rohit Sharma Mumbai Indians Burj Khalifa DIEL 82 SAMSUNG MUMBAI INDIANS	Dubai’s Burj Khalifa skyscraper displays photo of Indian cricketer Rohit Sharma This image of the Burj Khalifa has been doctored – the original does not show a projection of Indian cricketer Rohit Sharma

• 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	0.0329
0.0919	1000	0.0296
0.1378	1500	0.0314
0.1837	2000	0.0199
0.2297	2500	0.0435
0.2756	3000	0.0213
0.3215	3500	0.0293
0.3675	4000	0.0387
0.4134	4500	0.0064
0.4593	5000	0.0338
0.5053	5500	0.0317
0.5512	6000	0.0395
0.5972	6500	0.0129
0.6431	7000	0.036
0.6890	7500	0.0292
0.7350	8000	0.0215
0.7809	8500	0.02
0.8268	9000	0.0215
0.8728	9500	0.0139
0.9187	10000	0.0273
0.9646	10500	0.0138

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}
}