tomaarsen/splade-cocondenser-ensembledistil-nli

splade-cocondenser-ensembledistil trained on Natural Language Inference (NLI)

This is a SPLADE Sparse Encoder model finetuned from naver/splade-cocondenser-ensembledistil on the all-nli 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: naver/splade-cocondenser-ensembledistil
• Maximum Sequence Length: 256 tokens
• Output Dimensionality: 30522 dimensions
• Similarity Function: Dot Product
• Training Dataset:
  • all-nli
• 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: BertForMaskedLM 
  (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("tomaarsen/splade-cocondenser-ensembledistil-nli")
# Run inference
sentences = [
    'A man is sitting in on the side of the street with brass pots.',
    'A man is playing with the brass pots',
    'A group of adults are swimming at the beach.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 30522]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[16.8617, 12.9505,  0.2749],
#         [12.9505, 20.8479,  0.2440],
#         [ 0.2749,  0.2440, 18.7043]])

Evaluation

Metrics

Semantic Similarity

• Datasets: sts-dev and sts-test
• Evaluated with SparseEmbeddingSimilarityEvaluator

Metric	sts-dev	sts-test
pearson_cosine	0.8541	0.8223
spearman_cosine	0.847	0.8065
active_dims	99.3023	95.7545
sparsity_ratio	0.9967	0.9969

Training Details

Training Dataset

all-nli

• Dataset: all-nli at d482672
• Size: 10,000 training samples
• Columns: sentence1, sentence2, and score
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2	score
  type	string	string	float
  details	min: 6 tokens mean: 17.38 tokens max: 52 tokens	min: 4 tokens mean: 10.7 tokens max: 31 tokens	min: 0.0 mean: 0.5 max: 1.0

• Samples:

  sentence1	sentence2	score
  A person on a horse jumps over a broken down airplane.	A person is training his horse for a competition.	0.5
  A person on a horse jumps over a broken down airplane.	A person is at a diner, ordering an omelette.	0.0
  A person on a horse jumps over a broken down airplane.	A person is outdoors, on a horse.	1.0

• Loss: SpladeLoss with these parameters:

  {
      "loss": "SparseMultipleNegativesRankingLoss(scale=1, similarity_fct='dot_score')",
      "lambda_corpus": 0.003
  }
  

Evaluation Dataset

all-nli

• Dataset: all-nli at d482672
• Size: 1,000 evaluation samples
• Columns: sentence1, sentence2, and score
• Approximate statistics based on the first 1000 samples:

  	sentence1	sentence2	score
  type	string	string	float
  details	min: 6 tokens mean: 18.44 tokens max: 57 tokens	min: 5 tokens mean: 10.57 tokens max: 25 tokens	min: 0.0 mean: 0.5 max: 1.0

• Samples:

  sentence1	sentence2	score
  Two women are embracing while holding to go packages.	The sisters are hugging goodbye while holding to go packages after just eating lunch.	0.5
  Two women are embracing while holding to go packages.	Two woman are holding packages.	1.0
  Two women are embracing while holding to go packages.	The men are fighting outside a deli.	0.0

• Loss: SpladeLoss with these parameters:

  {
      "loss": "SparseMultipleNegativesRankingLoss(scale=1, similarity_fct='dot_score')",
      "lambda_corpus": 0.003
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• learning_rate: 4e-06
• 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: 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: 4e-06
• 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: no_duplicates
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	Validation Loss	sts-dev_spearman_cosine	sts-test_spearman_cosine
-1	-1	-	-	0.8366	-
0.032	20	0.8107	-	-	-
0.064	40	0.7854	-	-	-
0.096	60	0.7015	-	-	-
0.128	80	0.7161	-	-	-
0.16	100	0.724	-	-	-
0.192	120	0.6883	0.7255	0.8454	-
0.224	140	0.6661	-	-	-
0.256	160	0.6786	-	-	-
0.288	180	0.679	-	-	-
0.32	200	0.8013	-	-	-
0.352	220	0.6781	-	-	-
0.384	240	0.667	0.6779	0.8465	-
0.416	260	0.6691	-	-	-
0.448	280	0.7376	-	-	-
0.48	300	0.5601	-	-	-
0.512	320	0.6425	-	-	-
0.544	340	0.7406	-	-	-
0.576	360	0.6033	0.6623	0.8469	-
0.608	380	0.8166	-	-	-
0.64	400	0.5303	-	-	-
0.672	420	0.614	-	-	-
0.704	440	0.6253	-	-	-
0.736	460	0.5467	-	-	-
0.768	480	0.6804	0.6531	0.8470	-
0.8	500	0.6765	-	-	-
0.832	520	0.6522	-	-	-
0.864	540	0.5845	-	-	-
0.896	560	0.6786	-	-	-
0.928	580	0.5232	-	-	-
0.96	600	0.6077	0.6516	0.847	-
0.992	620	0.619	-	-	-
-1	-1	-	-	-	0.8065

• The bold row denotes the saved checkpoint.

Environmental Impact

Carbon emissions were measured using CodeCarbon.

• Energy Consumed: 0.008 kWh
• Carbon Emitted: 0.003 kg of CO2
• Hours Used: 0.033 hours

Training Hardware

• On Cloud: No
• GPU Model: 1 x NVIDIA GeForce RTX 3090
• CPU Model: 13th Gen Intel(R) Core(TM) i7-13700K
• RAM Size: 31.78 GB

Framework Versions

• Python: 3.11.6
• Sentence Transformers: 4.2.0.dev0
• Transformers: 4.52.4
• PyTorch: 2.6.0+cu124
• Accelerate: 1.5.1
• Datasets: 2.21.0
• 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}
    }