xin0920/splade-distilbert-base-uncased-msmarco-mrl

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("xin0920/splade-distilbert-base-uncased-msmarco-mrl")
# Run inference
sentences = [
    'Which laptop is best under 25000 INR?',
    'What are the best laptops under 25k?',
    'What is the best laptop under 45k?',
]
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 Information Retrieval

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

Metric	NanoClimateFEVER	NanoDBPedia	NanoFEVER	NanoFiQA2018	NanoHotpotQA	NanoMSMARCO	NanoNFCorpus	NanoNQ	NanoQuoraRetrieval	NanoSCIDOCS	NanoArguAna	NanoSciFact	NanoTouche2020
dot_accuracy@1	0.2	0.46	0.56	0.14	0.46	0.16	0.28	0.18	0.92	0.36	0.1	0.38	0.4898
dot_accuracy@3	0.34	0.66	0.64	0.32	0.66	0.26	0.36	0.34	0.96	0.5	0.28	0.56	0.7551
dot_accuracy@5	0.38	0.76	0.72	0.36	0.72	0.36	0.4	0.4	1.0	0.62	0.32	0.64	0.8367
dot_accuracy@10	0.46	0.82	0.82	0.44	0.84	0.46	0.44	0.48	1.0	0.7	0.38	0.66	0.9388
dot_precision@1	0.2	0.46	0.56	0.14	0.46	0.16	0.28	0.18	0.92	0.36	0.1	0.38	0.4898
dot_precision@3	0.12	0.46	0.2333	0.12	0.2533	0.0867	0.1867	0.1133	0.3733	0.26	0.0933	0.1933	0.4354
dot_precision@5	0.084	0.412	0.156	0.104	0.176	0.072	0.18	0.08	0.256	0.192	0.064	0.14	0.4286
dot_precision@10	0.058	0.348	0.088	0.068	0.11	0.046	0.148	0.048	0.132	0.124	0.038	0.072	0.3367
dot_recall@1	0.0883	0.025	0.5267	0.0678	0.23	0.16	0.01	0.17	0.8207	0.0767	0.1	0.365	0.0323
dot_recall@3	0.1533	0.0861	0.6333	0.1457	0.38	0.26	0.0176	0.32	0.8987	0.1617	0.28	0.54	0.0833
dot_recall@5	0.1717	0.1356	0.7133	0.2	0.44	0.36	0.0312	0.38	0.9727	0.1977	0.32	0.61	0.1362
dot_recall@10	0.2223	0.2109	0.8133	0.2637	0.55	0.46	0.0436	0.46	0.9827	0.2547	0.38	0.63	0.2075
dot_ndcg@10	0.191	0.4008	0.6697	0.1975	0.4642	0.289	0.169	0.3056	0.9457	0.264	0.2465	0.5013	0.3791
dot_mrr@10	0.279	0.5754	0.6316	0.2355	0.5763	0.237	0.3282	0.2675	0.95	0.455	0.2034	0.4667	0.6323
dot_map@100	0.1449	0.2348	0.6282	0.1473	0.3782	0.2547	0.0487	0.2611	0.9233	0.1868	0.2089	0.4647	0.2306
row_non_zero_mean_query	83.12	110.18	96.78	80.34	87.26	96.06	122.94	79.22	73.84	95.92	181.28	90.8	78.7755
row_sparsity_mean_query	0.9973	0.9964	0.9968	0.9974	0.9971	0.9969	0.996	0.9974	0.9976	0.9969	0.9941	0.997	0.9974
row_non_zero_mean_corpus	196.8254	146.9065	219.1213	125.9158	166.4719	105.462	199.5936	145.2502	74.9677	184.4491	160.5598	197.8948	140.811
row_sparsity_mean_corpus	0.9936	0.9952	0.9928	0.9959	0.9945	0.9965	0.9935	0.9952	0.9975	0.994	0.9947	0.9935	0.9954

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"
      ]
  }
  

Metric	Value
dot_accuracy@1	0.3608
dot_accuracy@3	0.5104
dot_accuracy@5	0.5782
dot_accuracy@10	0.6491
dot_precision@1	0.3608
dot_precision@3	0.2253
dot_precision@5	0.1804
dot_precision@10	0.1244
dot_recall@1	0.2056
dot_recall@3	0.3046
dot_recall@5	0.3591
dot_recall@10	0.4214
dot_ndcg@10	0.3864
dot_mrr@10	0.4491
dot_map@100	0.3163
row_non_zero_mean_query	98.1935
row_sparsity_mean_query	0.9968
row_non_zero_mean_corpus	158.7869
row_sparsity_mean_corpus	0.9948

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(
    (model): 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': None})
    )
    (cross_entropy_loss): CrossEntropyLoss()
  ), '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(
    (model): 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': None})
    )
    (cross_entropy_loss): CrossEntropyLoss()
  ), '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

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}
• 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
• 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: batch_sampler
• multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	NanoClimateFEVER_dot_ndcg@10	NanoDBPedia_dot_ndcg@10	NanoFEVER_dot_ndcg@10	NanoFiQA2018_dot_ndcg@10	NanoHotpotQA_dot_ndcg@10	NanoMSMARCO_dot_ndcg@10	NanoNFCorpus_dot_ndcg@10	NanoNQ_dot_ndcg@10	NanoQuoraRetrieval_dot_ndcg@10	NanoSCIDOCS_dot_ndcg@10	NanoArguAna_dot_ndcg@10	NanoSciFact_dot_ndcg@10	NanoTouche2020_dot_ndcg@10	NanoBEIR_mean_dot_ndcg@10
0.1938	200	12.7715	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.3876	400	0.2719	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.5814	600	0.234	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.7752	800	0.2068	-	-	-	-	-	-	-	-	-	-	-	-	-	-
0.9690	1000	0.2041	-	-	-	-	-	-	-	-	-	-	-	-	-	-
-1	-1	-	0.1910	0.4008	0.6697	0.1975	0.4642	0.2890	0.1690	0.3056	0.9457	0.2640	0.2465	0.5013	0.3791	0.3864

Framework Versions

• Python: 3.9.22
• Sentence Transformers: 4.2.0.dev0
• Transformers: 4.52.1
• PyTorch: 2.6.0+cu124
• Accelerate: 1.7.0
• Datasets: 3.6.0
• 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}
    }