sparse-encoder / example-splade-distilbert-base-uncased-quora-duplicates

splade-distilbert-base-uncased trained on Quora Duplicates Questions

This is a SPLADE Sparse Encoder model finetuned from distilbert/distilbert-base-uncased on the quora-duplicates dataset using the sentence-transformers library. It maps sentences & paragraphs to a 30522-dimensional sparse vector space and can be used for semantic search and sparse retrieval.

Model Details

Model Description

• Model Type: SPLADE Sparse Encoder
• Base model: distilbert/distilbert-base-uncased
• Maximum Sequence Length: 256 tokens
• Output Dimensionality: 30522 dimensions
• Similarity Function: Dot Product
• Training Dataset:
  • quora-duplicates
• Language: en
• License: apache-2.0

Model Sources

• Documentation: Sentence Transformers Documentation
• Documentation: Sparse Encoder Documentation
• Repository: Sentence Transformers on GitHub
• Hugging Face: Sparse Encoders on Hugging Face

Full Model Architecture

SparseEncoder(
  (0): MLMTransformer({'max_seq_length': 256, 'do_lower_case': False}) with MLMTransformer model: DistilBertForMaskedLM 
  (1): SpladePooling({'pooling_strategy': 'max', 'activation_function': 'relu', 'word_embedding_dimension': 30522})
)

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 SparseEncoder

# Download from the 🤗 Hub
model = SparseEncoder("arthurbresnu/splade-distilbert-base-uncased-quora-duplicates")
# Run inference
sentences = [
    'What accomplishments did Hillary Clinton achieve during her time as Secretary of State?',
    "What are Hillary Clinton's most recognized accomplishments while Secretary of State?",
    'What are Hillary Clinton’s qualifications to be President?',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# (3, 30522)

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

Evaluation

Metrics

Sparse Binary Classification

• Dataset: quora_duplicates_dev
• Evaluated with SparseBinaryClassificationEvaluator

Sparse Information Retrieval

• Datasets: NanoMSMARCO , NanoNQ , NanoNFCorpus , NanoQuoraRetrieval , NanoClimateFEVER , NanoDBPedia , NanoFEVER , NanoFiQA2018 , NanoHotpotQA , NanoMSMARCO , NanoNFCorpus , NanoNQ , NanoQuoraRetrieval , NanoSCIDOCS , NanoArguAna , NanoSciFact and NanoTouche2020
• Evaluated with SparseInformationRetrievalEvaluator

Sparse Nano BEIR

• Dataset: NanoBEIR_mean
• Evaluated with SparseNanoBEIREvaluator with these parameters:

  {
      "dataset_names": [
          "msmarco",
          "nq",
          "nfcorpus",
          "quoraretrieval"
      ]
  }
  

Sparse Nano BEIR

• Dataset: NanoBEIR_mean
• Evaluated with SparseNanoBEIREvaluator with these parameters:

  {
      "dataset_names": [
          "climatefever",
          "dbpedia",
          "fever",
          "fiqa2018",
          "hotpotqa",
          "msmarco",
          "nfcorpus",
          "nq",
          "quoraretrieval",
          "scidocs",
          "arguana",
          "scifact",
          "touche2020"
      ]
  }
  

Training Details

Training Dataset

quora-duplicates

• Dataset: quora-duplicates at 451a485
• Size: 99,000 training samples
• Columns: anchor , positive , and negative
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 6 tokens mean: 14.1 tokens max: 39 tokens	min: 6 tokens mean: 13.83 tokens max: 41 tokens	min: 6 tokens mean: 15.21 tokens max: 75 tokens

• Samples:

  anchor	positive	negative
  What are the best GMAT coaching institutes in Delhi NCR?	Which are the best GMAT coaching institutes in Delhi/NCR?	What are the best GMAT coaching institutes in Delhi-Noida Area?
  Is a third world war coming?	Is World War 3 more imminent than expected?	Since the UN is unable to control terrorism and groups like ISIS, al-Qaeda and countries that promote terrorism (even though it consumed those countries), can we assume that the world is heading towards World War III?
  Should I build iOS or Android apps first?	Should people choose Android or iOS first to build their App?	How much more effort is it to build your app on both iOS and Android?

• Loss: SpladeLoss with these parameters:

  {
      "loss": "SparseMultipleNegativesRankingLoss(scale=1.0, similarity_fct='dot_score')",
      "lambda_corpus": 3e-05,
      "lambda_query": 5e-05
  }
  

Evaluation Dataset

quora-duplicates

• Dataset: quora-duplicates at 451a485
• Size: 1,000 evaluation samples
• Columns: anchor , positive , and negative
• Approximate statistics based on the first 1000 samples:

  	anchor	positive	negative
  type	string	string	string
  details	min: 6 tokens mean: 14.05 tokens max: 40 tokens	min: 6 tokens mean: 14.14 tokens max: 44 tokens	min: 6 tokens mean: 14.56 tokens max: 60 tokens

• Samples:

  anchor	positive	negative
  What happens if we use petrol in diesel vehicles?	Why can't we use petrol in diesel?	Why are diesel engines noisier than petrol engines?
  Why is Saltwater taffy candy imported in Switzerland?	Why is Saltwater taffy candy imported in Laos?	Is salt a consumer product?
  Which is your favourite film in 2016?	What movie is the best movie of 2016?	What will the best movie of 2017 be?

• Loss: SpladeLoss with these parameters:

  {
      "loss": "SparseMultipleNegativesRankingLoss(scale=1.0, similarity_fct='dot_score')",
      "lambda_corpus": 3e-05,
      "lambda_query": 5e-05
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy : steps
• per_device_train_batch_size : 12
• per_device_eval_batch_size : 12
• learning_rate : 2e-05
• num_train_epochs : 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 : 12
• per_device_eval_batch_size : 12
• 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 : 0.0
• 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 : 0.0
• 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 : 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}
• tp_size : 0
• 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

• The bold row denotes the saved checkpoint.

Environmental Impact

Carbon emissions were measured using CodeCarbon .

• Energy Consumed : 0.025 kWh
• Carbon Emitted : 0.001 kg of CO2
• Hours Used : 0.222 hours

Training Hardware

• On Cloud : No
• GPU Model : 1 x NVIDIA GeForce RTX 3070 Ti Laptop GPU
• CPU Model : AMD Ryzen 9 6900HX with Radeon Graphics
• RAM Size : 30.61 GB

Framework Versions

• Python: 3.12.9
• Sentence Transformers: 4.2.0.dev0
• Transformers: 4.50.3
• PyTorch: 2.6.0+cu124
• Accelerate: 1.6.0
• Datasets: 3.5.0
• Tokenizers: 0.21.1

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",
}

SpladeLoss

@misc{formal2022distillationhardnegativesampling,
      title={From Distillation to Hard Negative Sampling: Making Sparse Neural IR Models More Effective},
      author={Thibault Formal and Carlos Lassance and Benjamin Piwowarski and Stéphane Clinchant},
      year={2022},
      eprint={2205.04733},
      archivePrefix={arXiv},
      primaryClass={cs.IR},
      url={https://arxiv.org/abs/2205.04733},
}

SparseMultipleNegativesRankingLoss

@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}
}

FlopsLoss

@article{paria2020minimizing,
    title={Minimizing flops to learn efficient sparse representations},
    author={Paria, Biswajit and Yeh, Chih-Kuan and Yen, Ian EH and Xu, Ning and Ravikumar, Pradeep and P{'o}czos, Barnab{'a}s},
    journal={arXiv preprint arXiv:2004.05665},
    year={2020}
    }