codersan/FaLaBSE-v4

SentenceTransformer based on sentence-transformers/LaBSE

This is a sentence-transformers model finetuned from sentence-transformers/LaBSE. 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: sentence-transformers/LaBSE
• Maximum Sequence Length: 256 tokens
• Output Dimensionality: 768 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}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 768, 'pooling_mode_cls_token': True, 'pooling_mode_mean_tokens': False, '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): Dense({'in_features': 768, 'out_features': 768, 'bias': True, 'activation_function': 'torch.nn.modules.activation.Tanh'})
  (3): 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("codersan/FaLaBSE-v4")
# Run inference
sentences = [
    'معنی و هدف زندگی چیست؟',
    'معنی دقیق زندگی چیست؟',
    'چه فیلم هایی را به همه توصیه می کنید که تماشا کنند؟',
]
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]

Training Details

Training Dataset

Unnamed Dataset

• Size: 165,665 training samples
• Columns: anchor and positive
• Approximate statistics based on the first 1000 samples:

  	anchor	positive
  type	string	string
  details	min: 5 tokens mean: 14.65 tokens max: 48 tokens	min: 5 tokens mean: 14.87 tokens max: 53 tokens

• Samples:

  anchor	positive
  طالع بینی: من یک ماه و کلاه درپوش خورشید است ... این در مورد من چه می گوید؟	من یک برج سه گانه (خورشید ، ماه و صعود در برجستگی) هستم که این در مورد من چه می گوید؟
  چگونه می توانم یک زمین شناس خوب باشم؟	چه کاری باید انجام دهم تا یک زمین شناس عالی باشم؟
  چگونه می توانم نظرات YouTube خود را بخوانم و پیدا کنم؟	چگونه می توانم تمام نظرات YouTube خود را ببینم؟

• Loss: MultipleNegativesRankingLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 32
• learning_rate: 2e-05
• weight_decay: 0.01
• batch_sampler: no_duplicates

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: no
• prediction_loss_only: True
• per_device_train_batch_size: 32
• per_device_eval_batch_size: 8
• 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: 2e-05
• weight_decay: 0.01
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1.0
• num_train_epochs: 3
• 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: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss
0.0386	100	0.0863
0.0772	200	0.0652
0.1159	300	0.0595
0.1545	400	0.0614
0.1931	500	0.05
0.2317	600	0.0453
0.2704	700	0.0579
0.3090	800	0.0542
0.3476	900	0.0534
0.3862	1000	0.0532
0.4249	1100	0.0548
0.4635	1200	0.0519
0.5021	1300	0.0547
0.5407	1400	0.0563
0.5794	1500	0.0474
0.6180	1600	0.0433
0.6566	1700	0.0545
0.6952	1800	0.0509
0.7339	1900	0.0453
0.7725	2000	0.0446
0.8111	2100	0.0506
0.8497	2200	0.046
0.8884	2300	0.0413
0.9270	2400	0.149
0.9656	2500	0.6993
1.0039	2600	1.081
1.0425	2700	0.0397
1.0811	2800	0.0337
1.1197	2900	0.0307
1.1584	3000	0.0323
1.1970	3100	0.0273
1.2356	3200	0.0292
1.2742	3300	0.0323
1.3129	3400	0.0352
1.3515	3500	0.0281
1.3901	3600	0.0318
1.4287	3700	0.0281
1.4674	3800	0.0304
1.5060	3900	0.0321
1.5446	4000	0.035
1.5832	4100	0.0279
1.6219	4200	0.0286
1.6605	4300	0.0333
1.6991	4400	0.0323
1.7377	4500	0.0312
1.7764	4600	0.0261
1.8150	4700	0.0361
1.8536	4800	0.0306
1.8922	4900	0.028
1.9309	5000	0.1226
1.9695	5100	0.5625
2.0077	5200	0.8337
2.0463	5300	0.0273
2.0850	5400	0.0242
2.1236	5500	0.0236
2.1622	5600	0.0237
2.2008	5700	0.0197
2.2395	5800	0.0217
2.2781	5900	0.0244
2.3167	6000	0.027
2.3553	6100	0.0235
2.3940	6200	0.0233
2.4326	6300	0.0225
2.4712	6400	0.023
2.5098	6500	0.023
2.5485	6600	0.0243
2.5871	6700	0.0215
2.6257	6800	0.0236
2.6643	6900	0.0234
2.7030	7000	0.0239
2.7416	7100	0.0248
2.7802	7200	0.02
2.8188	7300	0.0271
2.8575	7400	0.0235
2.8961	7500	0.0214
2.9347	7600	0.1147
2.9733	7700	0.5838

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.3.1
• Transformers: 4.47.0
• PyTorch: 2.5.1+cu121
• Accelerate: 1.2.1
• Datasets: 3.2.0
• 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}
}