richie-ghost/sbert_ft_cross-encoder-nli-deberta-v3-large

SentenceTransformer based on cross-encoder/nli-deberta-v3-large

This is a sentence-transformers model finetuned from cross-encoder/nli-deberta-v3-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: cross-encoder/nli-deberta-v3-large
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 1024 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: DebertaV2Model 
  (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})
)

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("richie-ghost/sbert_ft_cross-encoder-nli-deberta-v3-large")
# Run inference
sentences = [
    'Three men riding a bicycle, tow of them are wearing a helmet.',
    'There are at least two helmets.',
    'Accountability measures help establish the financial condition of the government.',
]
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]

Evaluation

Metrics

Information Retrieval

• Dataset: eval
• Evaluated with InformationRetrievalEvaluator

Metric	Value
cosine_accuracy@1	0.0003
cosine_accuracy@3	0.2843
cosine_accuracy@5	0.4288
cosine_accuracy@10	0.5317
cosine_precision@1	0.0003
cosine_precision@3	0.0948
cosine_precision@5	0.0858
cosine_precision@10	0.0532
cosine_recall@1	0.0003
cosine_recall@3	0.2843
cosine_recall@5	0.4288
cosine_recall@10	0.5317
cosine_ndcg@10	0.26
cosine_mrr@10	0.1732
cosine_map@100	0.185
dot_accuracy@1	0.0037
dot_accuracy@3	0.2625
dot_accuracy@5	0.4018
dot_accuracy@10	0.509
dot_precision@1	0.0037
dot_precision@3	0.0875
dot_precision@5	0.0804
dot_precision@10	0.0509
dot_recall@1	0.0037
dot_recall@3	0.2625
dot_recall@5	0.4018
dot_recall@10	0.509
dot_ndcg@10	0.2476
dot_mrr@10	0.1645
dot_map@100	0.1768

Training Details

Training Dataset

Unnamed Dataset

• Size: 40,338 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: 5 tokens mean: 19.64 tokens max: 129 tokens	min: 4 tokens mean: 11.27 tokens max: 36 tokens

• Samples:

  sentence_0	sentence_1
  A group of ladies trying to learn how to belly dance.	Several women learn the art of exotic dancing.
  A man and a woman are having a conversation, while the man drinks a beer.	The man is drinking.
  A brown dog drinks from a water bottle.	A brown cat drinks from a bowl.

• 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: 4
• multi_dataset_batch_sampler: round_robin

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
• num_train_epochs: 4
• 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: 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
• eval_use_gather_object: False
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: round_robin

Training Logs

Epoch	Step	Training Loss	eval_cosine_map@100
0.1983	500	1.2356	0.0873
0.3965	1000	0.4077	0.1200
0.5948	1500	0.3205	0.1280
0.7930	2000	0.2576	0.1416
0.9913	2500	0.2435	0.1476
1.0	2522	-	0.1492
1.1895	3000	0.1821	0.1553
1.3878	3500	0.1237	0.1589
1.5860	4000	0.1074	0.1603
1.7843	4500	0.0905	0.1654
1.9826	5000	0.0783	0.1685
2.0	5044	-	0.1683
2.1808	5500	0.0583	0.1698
2.3791	6000	0.0432	0.1746
2.5773	6500	0.0365	0.1749
2.7756	7000	0.0303	0.1791
2.9738	7500	0.0276	0.1788
3.0	7566	-	0.1805
3.1721	8000	0.02	0.1807
3.3703	8500	0.013	0.1823
3.5686	9000	0.0123	0.1839
3.7669	9500	0.0099	0.1852
3.9651	10000	0.01	0.1850
4.0	10088	-	0.1850

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 3.2.1
• Transformers: 4.44.2
• PyTorch: 2.5.0+cu121
• Accelerate: 1.0.1
• Datasets: 3.0.2
• Tokenizers: 0.19.1

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}
}