slimaneMakh/triplet_CloseHlabel_farLabel_andnegativ-1M-5eps-XLMR_29may

SentenceTransformer based on FacebookAI/xlm-roberta-base

This is a sentence-transformers model finetuned from FacebookAI/xlm-roberta-base. 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: FacebookAI/xlm-roberta-base
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 tokens
• 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: 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})
)

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("slimaneMakh/triplet_CloseHlabel_farLabel_andnegativ-1M-5eps-XLMR_29may")
# Run inference
sentences = [
    'Sales',
    'Revenue',
    'Operating profit',
]
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

Triplet

• Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	0.9988
dot_accuracy	0.0015
manhattan_accuracy	0.9975
euclidean_accuracy	0.9991
max_accuracy	0.9991

Training Details

Training Dataset

Unnamed Dataset

• Size: 660,643 training samples
• Columns: anchor_label, pos_hlabel, and neg_hlabel
• Approximate statistics based on the first 1000 samples:

  	anchor_label	pos_hlabel	neg_hlabel
  type	string	string	string
  details	min: 3 tokens mean: 11.86 tokens max: 39 tokens	min: 3 tokens mean: 9.06 tokens max: 32 tokens	min: 3 tokens mean: 7.99 tokens max: 25 tokens

• Samples:

  anchor_label	pos_hlabel	neg_hlabel
  Basic earnings (loss) per share	Tavakasum kahjum aktsia kohta	II Kapital z nadwyzki wartosci emisyjnej ponad wartosc nominalna
  Comprehensive income	Suma dochodow calkowitych	dont Marques
  Cash and cash equivalents	Cash and cash equivalents	Cars incl prepayments

• Loss: TripletLoss with these parameters:

  {
      "distance_metric": "TripletDistanceMetric.EUCLIDEAN",
      "triplet_margin": 5
  }
  

Evaluation Dataset

Unnamed Dataset

• Size: 283,133 evaluation samples
• Columns: anchor_label, pos_hlabel, and neg_hlabel
• Approximate statistics based on the first 1000 samples:

  	anchor_label	pos_hlabel	neg_hlabel
  type	string	string	string
  details	min: 3 tokens mean: 11.78 tokens max: 37 tokens	min: 3 tokens mean: 9.22 tokens max: 39 tokens	min: 3 tokens mean: 8.12 tokens max: 29 tokens

• Samples:

  anchor_label	pos_hlabel	neg_hlabel
  Deferred tax assets	Deferred tax assets	Immateriella tillgangar
  Equity	EGET KAPITAL inklusive periodens resultat	Materials
  Adjustments for decrease (increase) in other operating receivables	Okning av ovriga rorelsetillgangar	Rorelseresultat

• Loss: TripletLoss with these parameters:

  {
      "distance_metric": "TripletDistanceMetric.EUCLIDEAN",
      "triplet_margin": 5
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• num_train_epochs: 1
• warmup_ratio: 0.1
• batch_sampler: no_duplicates

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• prediction_loss_only: True
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 1
• eval_accumulation_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: 1.0
• num_train_epochs: 1
• 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
• 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}
• 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: False
• hub_always_push: False
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• include_inputs_for_metrics: False
• 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_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	loss	max_accuracy
0.0121	500	3.7705	-	-
0.0242	1000	1.4084	-	-
0.0363	1500	0.7062	-	-
0.0484	2000	0.5236	-	-
0.0605	2500	0.4348	-	-
0.0727	3000	0.3657	-	-
0.0848	3500	0.3657	-	-
0.0969	4000	0.2952	-	-
0.1090	4500	0.3805	-	-
0.1211	5000	0.3255	-	-
0.1332	5500	0.2621	-	-
0.1453	6000	0.2377	-	-
0.1574	6500	0.2139	-	-
0.1695	7000	0.2085	-	-
0.1816	7500	0.1809	-	-
0.1937	8000	0.1711	-	-
0.2059	8500	0.1608	-	-
0.2180	9000	0.1808	-	-
0.2301	9500	0.1553	-	-
0.2422	10000	0.1417	-	-
0.2543	10500	0.1329	-	-
0.2664	11000	0.1689	-	-
0.2785	11500	0.1292	-	-
0.2906	12000	0.1181	-	-
0.3027	12500	0.1223	-	-
0.3148	13000	0.129	-	-
0.3269	13500	0.0911	-	-
0.3391	14000	0.113	-	-
0.3512	14500	0.0955	-	-
0.3633	15000	0.108	-	-
0.3754	15500	0.094	-	-
0.3875	16000	0.0947	-	-
0.3996	16500	0.0748	-	-
0.4117	17000	0.0699	-	-
0.4238	17500	0.0707	-	-
0.4359	18000	0.0768	-	-
0.4480	18500	0.0805	-	-
0.4601	19000	0.0705	-	-
0.4723	19500	0.069	-	-
0.4844	20000	0.072	-	-
0.4965	20500	0.0669	-	-
0.5086	21000	0.066	-	-
0.5207	21500	0.0624	-	-
0.5328	22000	0.0687	-	-
0.5449	22500	0.076	-	-
0.5570	23000	0.0563	-	-
0.5691	23500	0.0594	-	-
0.5812	24000	0.0524	-	-
0.5933	24500	0.0528	-	-
0.6055	25000	0.0448	-	-
0.6176	25500	0.041	-	-
0.6297	26000	0.0397	-	-
0.6418	26500	0.0489	-	-
0.6539	27000	0.0595	-	-
0.6660	27500	0.034	-	-
0.6781	28000	0.0569	-	-
0.6902	28500	0.0467	-	-
0.7023	29000	0.0323	-	-
0.7144	29500	0.0428	-	-
0.7266	30000	0.0344	-	-
0.7387	30500	0.029	-	-
0.7508	31000	0.0418	-	-
0.7629	31500	0.0285	-	-
0.7750	32000	0.0425	-	-
0.7871	32500	0.0266	-	-
0.7992	33000	0.0325	-	-
0.8113	33500	0.0215	-	-
0.8234	34000	0.0316	-	-
0.8355	34500	0.0286	-	-
0.8476	35000	0.0285	-	-
0.8598	35500	0.0284	-	-
0.8719	36000	0.0147	-	-
0.8840	36500	0.0217	-	-
0.8961	37000	0.0311	-	-
0.9082	37500	0.0202	-	-
0.9203	38000	0.0236	-	-
0.9324	38500	0.0201	-	-
0.9445	39000	0.0246	-	-
0.9566	39500	0.0177	-	-
0.9687	40000	0.0173	-	-
0.9808	40500	0.0202	-	-
0.9930	41000	0.017	-	-
1.0	41291	-	0.0140	0.9991

Framework Versions

• Python: 3.10.13
• Sentence Transformers: 3.0.0
• Transformers: 4.39.3
• PyTorch: 2.1.2
• Accelerate: 0.28.0
• Datasets: 2.18.0
• Tokenizers: 0.15.2

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",
}

TripletLoss

@misc{hermans2017defense,
    title={In Defense of the Triplet Loss for Person Re-Identification}, 
    author={Alexander Hermans and Lucas Beyer and Bastian Leibe},
    year={2017},
    eprint={1703.07737},
    archivePrefix={arXiv},
    primaryClass={cs.CV}
}