ptpedroVortal/nomic_vortal_v3.3

SentenceTransformer based on nomic-ai/nomic-embed-text-v1

This is a sentence-transformers model finetuned from nomic-ai/nomic-embed-text-v1. 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: nomic-ai/nomic-embed-text-v1
• Maximum Sequence Length: 8192 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': 8192, 'do_lower_case': False}) with Transformer model: NomicBertModel 
  (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})
  (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("ptpedroVortal/nomic_vortal_v3.3")
# Run inference
sentences = [
    'Collect the details that are associated with product \'\' \'Macbook Air 13" com processador M1/M2 e 8 GB de RAM (Telado PT-PT)\', with quantity 1, unit UN',
    'Apresenta -se de seguida a configuração financeira para a fornecimento  dos produtos \\nrequeridos , mediante opções por cor e diferentes características:\\nNOTA:  Valores válidos  até 23 de Fevereiro e mediante adjudicação de 2 ou mais \\nequipamentos  portáteis (excluindo Teclado)\\nPART-NUMBER QTD. DESCRIÇÃOVALOR\\nUNITÁRIOVALOR\\nTOTAL\\nMLY03PO/A 1Apple Macbook AIR  13,6"  (Disco 512GB SSD; 10 core) 1 545,08 €                 1 545,08 €\\nMLXY3PO/A 1Apple Macbook AIR  13,6" (Disco 256GB SSD, 8 core) 1 227,48 €                 1 227,48 €',
    'LOTE 5\n1 MESA APOIO MESA DE APOIO EM INOX AISI 304 2,0 279,000 23,0 558,000\nMesa com 4 rodas , 2 com travão\nTabuleiro inferior\nDimens: C 700 x L 500 x A 800mm\nPrateleira inferior - profundidade 250mm\nFabrico Nacional e por medida\nTotal do do lote 5: 558,00€ Quinhentos e cinquenta e oito euros',
]
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

Semantic Similarity

• Evaluated with main.CustomEvaluator

Metric	Value
pearson_cosine	nan
spearman_cosine	nan

Training Details

Training Dataset

Unnamed Dataset

• Size: 222 training samples
• Columns: query and correct_node
• Approximate statistics based on the first 222 samples:

  	query	correct_node
  type	string	string
  details	min: 15 tokens mean: 55.17 tokens max: 154 tokens	min: 22 tokens mean: 109.22 tokens max: 2920 tokens

• Samples:

  query	correct_node
  Collect the details that are associated with Lot 4 product '' 'Mesas de Mayo', with quantity 2, unit Subcontracting Unit	LOTE 4 1 MESA DE MAYO 82JM 10.ME.1831 2,000 842,00000 23 1 684,00 oitocentos e quarenta e dois euros Origem : Nacional Marca : MOBIT Prazo de entrega: 30 dias Garantia: 2 anos Transporte
  Collect the details that are associated with Lot 7 product '' 'Carro transporte de roupa suja ', with quantity 1, unit US	Lote 7 nan nan nan nan nan\nRef. Description Qt. Un. Un. Price Total\n9856 Carros para Transporte de Roupa Suja e Limpa 1 US 16.23 16.23</code>
  Collect the details that are associated with product '' '2202000014 - FIO SUT. SEDA NÃO ABS. 2/0 MULTIF. SEM AGULHA (CART.)', with quantity 72, unit UN	2202000014 - FIO SUT. SEDA NÃO ABS. 2/0 MULTIF. SEM AGULHA (CART.) 0.36

• Loss: MultipleNegativesRankingLoss with these parameters:

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

Evaluation Dataset

Unnamed Dataset

• Size: 27 evaluation samples
• Columns: query and correct_node
• Approximate statistics based on the first 27 samples:

  	query	correct_node
  type	string	string
  details	min: 17 tokens mean: 56.85 tokens max: 121 tokens	min: 40 tokens mean: 228.15 tokens max: 2963 tokens

• Samples:

  query	correct_node
  Collect the details that are associated with product '' '2202000055 - FIO SUT. POLIAMIDA NÃO ABS. 2/0 MONOF. AG. LANC. 39 MM 3/8 C (CART.)', with quantity 1656, unit UN	2202000055 - FIO SUT. POLIAMIDA NÃO ABS. 2/0 MONOF. AG. LANC. 39 MM 3/8 C (CART.) 1.28
  Collect the details that are associated with Lot 3 product 'Portaria do Parque Coberto dos Olhos de Água' 'Vigilância e segurança humana contínua - Olhos de Água - período de 3 meses - todos os dias da semana, incluindo feriados, total estimado de 2754H', with quantity 1, unit UN
  Collect the details that are associated with Lot 3 product 'Portaria do Parque Coberto dos Olhos de Água' 'Vigilância e segurança humana contínua - Olhos de Água - período de 3 meses - todos os dias da semana, incluindo feriados, total estimado de 2754H', with quantity 1, unit UN	Lote 3:\nPreço Unitário: 10,00€ (dez euros) /hora\nPreço Total: 27.540,00€ (vinte sete mil quinhentos e quarenta euros)

• Loss: MultipleNegativesRankingLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• num_train_epochs: 10
• warmup_ratio: 0.1
• bf16: True
• load_best_model_at_end: True
• batch_sampler: no_duplicates

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• 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
• 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: 1.0
• num_train_epochs: 10
• 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
• 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: True
• 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: True
• 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: False
• 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	Validation Loss	spearman_cosine
7.1429	100	0.0965	0.2395	nan

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.10.14
• Sentence Transformers: 3.3.1
• Transformers: 4.47.0.dev0
• PyTorch: 2.5.1+cu121
• Accelerate: 1.1.1
• Datasets: 3.1.0
• Tokenizers: 0.20.4

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}
}