README.md

metadata

tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - generated_from_trainer
  - dataset_size:4957
  - loss:AttributeTripletLoss
base_model: Alibaba-NLP/gte-base-en-v1.5
widget:
  - source_sentence: Marianne Williamson
    sentences:
      - Francesco Sedita
      - HarperCollins
      - author
      - publisher
  - source_sentence: Brilliance Audio on CD Unabridged; Unabridged edition (September 22, 2009)
    sentences:
      - Alan Bradley
      - publisher
      - author
      - Bearport Publishing
  - source_sentence: 01 June 2003
    sentences:
      - '9785551553090'
      - Signet (September 4, 1990)
      - publication_date
      - isbn_13
  - source_sentence: The Elusive Pimpernel (Dover Value Editions)
    sentences:
      - HarperCollins; Illustrated. edition (May 15, 1974)
      - publication_date
      - The Real Benjamin Franklin (Vol. 2 of the American classic series)
      - title
  - source_sentence: Diana Gabaldon
    sentences:
      - publisher
      - author
      - Wong Herbert Yee
      - Bloomsbury Publishing PLC
pipeline_tag: sentence-similarity
library_name: sentence-transformers
metrics:
  - cosine_accuracy
  - silhouette_cosine
  - silhouette_euclidean
model-index:
  - name: SentenceTransformer based on Alibaba-NLP/gte-base-en-v1.5
    results:
      - task:
          type: triplet
          name: Triplet
        dataset:
          name: Unknown
          type: unknown
        metrics:
          - type: cosine_accuracy
            value: 0.9509981870651245
            name: Cosine Accuracy
          - type: cosine_accuracy
            value: 0.9656862616539001
            name: Cosine Accuracy
      - task:
          type: silhouette
          name: Silhouette
        dataset:
          name: Unknown
          type: unknown
        metrics:
          - type: silhouette_cosine
            value: 0.5749762654304504
            name: Silhouette Cosine
          - type: silhouette_euclidean
            value: 0.49482977390289307
            name: Silhouette Euclidean
          - type: silhouette_cosine
            value: 0.5844279527664185
            name: Silhouette Cosine
          - type: silhouette_euclidean
            value: 0.5075943470001221
            name: Silhouette Euclidean

SentenceTransformer based on Alibaba-NLP/gte-base-en-v1.5

This is a sentence-transformers model finetuned from Alibaba-NLP/gte-base-en-v1.5. 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: Alibaba-NLP/gte-base-en-v1.5
• 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: NewModel 
  (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})
)

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("albertus-sussex/veriscrape-sbert-book-reference_2_to_verify_8-fold-5")
# Run inference
sentences = [
    'Diana Gabaldon',
    'Wong Herbert Yee',
    'Bloomsbury Publishing PLC',
]
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	0.951

Silhouette

• Evaluated with veriscrape.training.SilhouetteEvaluator

Metric	Value
silhouette_cosine	0.575
silhouette_euclidean	0.4948

Triplet

• Evaluated with TripletEvaluator

Metric	Value
cosine_accuracy	0.9657

Silhouette

• Evaluated with veriscrape.training.SilhouetteEvaluator

Metric	Value
silhouette_cosine	0.5844
silhouette_euclidean	0.5076

Training Details

Training Dataset

Unnamed Dataset

• Size: 4,957 training samples
• Columns: anchor, positive, negative, pos_attr_name, and neg_attr_name
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative	pos_attr_name	neg_attr_name
  type	string	string	string	string	string
  details	min: 3 tokens mean: 8.81 tokens max: 33 tokens	min: 3 tokens mean: 8.62 tokens max: 31 tokens	min: 3 tokens mean: 8.95 tokens max: 30 tokens	min: 3 tokens mean: 3.78 tokens max: 5 tokens	min: 3 tokens mean: 3.78 tokens max: 5 tokens

• Samples:

  anchor	positive	negative	pos_attr_name	neg_attr_name
  Emotions Revealed: Recognizing Faces and Feelings to Improve Communication and Emotional Life	Zen and the Art of Motorcycle Maintenance: An Inquiry Into Values	HarperCollins Publishers Inc	title	publisher
  Dell (2006)	Skyhorse Publishing (March 8, 2010)	Nemesis	publication_date	title
  A Woman's Worth	Bone: v.1	Dell (2006)	title	publication_date

• Loss: veriscrape.training.AttributeTripletLoss with these parameters:

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

Evaluation Dataset

Unnamed Dataset

• Size: 551 evaluation samples
• Columns: anchor, positive, negative, pos_attr_name, and neg_attr_name
• Approximate statistics based on the first 551 samples:

  	anchor	positive	negative	pos_attr_name	neg_attr_name
  type	string	string	string	string	string
  details	min: 3 tokens mean: 8.55 tokens max: 33 tokens	min: 3 tokens mean: 8.63 tokens max: 31 tokens	min: 3 tokens mean: 8.89 tokens max: 25 tokens	min: 3 tokens mean: 3.73 tokens max: 5 tokens	min: 3 tokens mean: 3.83 tokens max: 5 tokens

• Samples:

  anchor	positive	negative	pos_attr_name	neg_attr_name
  As the Future Catches You: How Genomics & Other Forces Are Changing Your Life, Work, Health & Wealth	A Lesson Before Dying (Oprah's Book Club)	978-1402220128	title	isbn_13
  Vintage (July 8, 2003)	HarperCollins Publishers Inc	Vintage (September 8, 2009)	publisher	publication_date
  9780525423645	978-0978785109	Dean Koontz	isbn_13	author

• Loss: veriscrape.training.AttributeTripletLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: epoch
• per_device_train_batch_size: 128
• per_device_eval_batch_size: 128
• num_train_epochs: 5
• warmup_ratio: 0.1

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: epoch
• prediction_loss_only: True
• per_device_train_batch_size: 128
• per_device_eval_batch_size: 128
• 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: 5
• 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: 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: False
• hub_always_push: False
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• include_inputs_for_metrics: False
• 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
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	Validation Loss	cosine_accuracy	silhouette_cosine
-1	-1	-	-	0.3448	0.1061
1.0	39	1.8579	0.7588	0.9474	0.5615
2.0	78	0.2764	0.6407	0.9474	0.5740
3.0	117	0.2342	0.5743	0.9437	0.5770
4.0	156	0.2029	0.5806	0.9510	0.5742
5.0	195	0.1839	0.5671	0.9510	0.5750
-1	-1	-	-	0.9657	0.5844

Framework Versions

• Python: 3.10.16
• Sentence Transformers: 3.4.1
• Transformers: 4.45.2
• PyTorch: 2.5.1+cu124
• Accelerate: 1.5.2
• Datasets: 3.1.0
• Tokenizers: 0.20.3

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",
}

AttributeTripletLoss

@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}
}