Hugging Face's logo

tomaarsen
/
csr-mxbai-embed-large-v1-nq-dot-scale-1-gamma-0.1-detach-2

Feature Extraction
sentence-transformers
Safetensors
English
bert
sparse-encoder
sparse
csr
Generated from Trainer
dataset_size:99000
loss:CSRLoss
loss:SparseMultipleNegativesRankingLoss
Eval Results
text-embeddings-inference
Model card Files
xet
Community

Sparse CSR model trained on Natural Questions

This is a CSR Sparse Encoder model finetuned from mixedbread-ai/mxbai-embed-large-v1 on the natural-questions dataset using the sentence-transformers library. It maps sentences & paragraphs to a 4096-dimensional sparse vector space with 256 maximum active dimensions and can be used for semantic search and sparse retrieval.

Model Details

Model Description

  • Model Type: CSR Sparse Encoder
  • Base model: mixedbread-ai/mxbai-embed-large-v1
  • Maximum Sequence Length: 512 tokens
  • Output Dimensionality: 4096 dimensions (trained with 256 maximum active dimensions)
  • Similarity Function: Dot Product
  • Training Dataset:
  • Language: en
  • License: apache-2.0

Model Sources

Full Model Architecture 

SparseEncoder(
  (0): Transformer({'max_seq_length': 512, 'do_lower_case': False}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 1024, '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})
  (2): CSRSparsity({'input_dim': 1024, 'hidden_dim': 4096, 'k': 256, 'k_aux': 512, 'normalize': False, 'dead_threshold': 30})
)

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/csr-mxbai-embed-large-v1-nq-dot-scale-1-gamma-0.1-detach-2")
# Run inference
queries = [
    "who is cornelius in the book of acts",
]
documents = [
    'Cornelius the Centurion Cornelius (Greek: Κορνήλιος) was a Roman centurion who is considered by Christians to be one of the first Gentiles to convert to the faith, as related in Acts of the Apostles.',
    "Joe Ranft Ranft reunited with Lasseter when he was hired by Pixar in 1991 as their head of story.[1] There he worked on all of their films produced up to 2006; this included Toy Story (for which he received an Academy Award nomination) and A Bug's Life, as the co-story writer and others as story supervisor. His final film was Cars. He also voiced characters in many of the films, including Heimlich the caterpillar in A Bug's Life, Wheezy the penguin in Toy Story 2, and Jacques the shrimp in Finding Nemo.[1]",
    'Wonderful Tonight "Wonderful Tonight" is a ballad written by Eric Clapton. It was included on Clapton\'s 1977 album Slowhand. Clapton wrote the song about Pattie Boyd.[1] The female vocal harmonies on the song are provided by Marcella Detroit (then Marcy Levy) and Yvonne Elliman.',
]
query_embeddings = model.encode_query(queries)
document_embeddings = model.encode_document(documents)
print(query_embeddings.shape, document_embeddings.shape)
# [1, 4096] [3, 4096]

# Get the similarity scores for the embeddings
similarities = model.similarity(query_embeddings, document_embeddings)
print(similarities)
# tensor([[119.9615,  28.3687,  21.7583]])

Evaluation

Metrics

Sparse Information Retrieval

Metric Value
dot_accuracy@1 0.213
dot_accuracy@3 0.332
dot_accuracy@5 0.384
dot_accuracy@10 0.471
dot_precision@1 0.213
dot_precision@3 0.1107
dot_precision@5 0.0768
dot_precision@10 0.0471
dot_recall@1 0.213
dot_recall@3 0.332
dot_recall@5 0.384
dot_recall@10 0.471
dot_ndcg@10 0.332
dot_mrr@10 0.2889
dot_map@100 0.2989
query_active_dims 4.0
query_sparsity_ratio 0.999
corpus_active_dims 4.0
corpus_sparsity_ratio 0.999

Sparse Information Retrieval

Metric Value
dot_accuracy@1 0.399
dot_accuracy@3 0.547
dot_accuracy@5 0.605
dot_accuracy@10 0.676
dot_precision@1 0.399
dot_precision@3 0.1823
dot_precision@5 0.121
dot_precision@10 0.0676
dot_recall@1 0.399
dot_recall@3 0.547
dot_recall@5 0.605
dot_recall@10 0.676
dot_ndcg@10 0.5319
dot_mrr@10 0.4862
dot_map@100 0.4942
query_active_dims 8.0
query_sparsity_ratio 0.998
corpus_active_dims 8.0
corpus_sparsity_ratio 0.998

Sparse Information Retrieval

Metric Value
dot_accuracy@1 0.625
dot_accuracy@3 0.772
dot_accuracy@5 0.817
dot_accuracy@10 0.864
dot_precision@1 0.625
dot_precision@3 0.2573
dot_precision@5 0.1634
dot_precision@10 0.0864
dot_recall@1 0.625
dot_recall@3 0.772
dot_recall@5 0.817
dot_recall@10 0.864
dot_ndcg@10 0.7454
dot_mrr@10 0.7073
dot_map@100 0.7111
query_active_dims 16.0
query_sparsity_ratio 0.9961
corpus_active_dims 16.0
corpus_sparsity_ratio 0.9961

Sparse Information Retrieval

Metric Value
dot_accuracy@1 0.796
dot_accuracy@3 0.914
dot_accuracy@5 0.935
dot_accuracy@10 0.96
dot_precision@1 0.796
dot_precision@3 0.3047
dot_precision@5 0.187
dot_precision@10 0.096
dot_recall@1 0.796
dot_recall@3 0.914
dot_recall@5 0.935
dot_recall@10 0.96
dot_ndcg@10 0.8831
dot_mrr@10 0.8579
dot_map@100 0.8592
query_active_dims 32.0
query_sparsity_ratio 0.9922
corpus_active_dims 32.0
corpus_sparsity_ratio 0.9922

Sparse Information Retrieval

Metric Value
dot_accuracy@1 0.874
dot_accuracy@3 0.964
dot_accuracy@5 0.975
dot_accuracy@10 0.984
dot_precision@1 0.874
dot_precision@3 0.3213
dot_precision@5 0.195
dot_precision@10 0.0984
dot_recall@1 0.874
dot_recall@3 0.964
dot_recall@5 0.975
dot_recall@10 0.984
dot_ndcg@10 0.9354
dot_mrr@10 0.9191
dot_map@100 0.9198
query_active_dims 64.0
query_sparsity_ratio 0.9844
corpus_active_dims 64.0
corpus_sparsity_ratio 0.9844

Sparse Information Retrieval

Metric Value
dot_accuracy@1 0.917
dot_accuracy@3 0.982
dot_accuracy@5 0.987
dot_accuracy@10 0.993
dot_precision@1 0.917
dot_precision@3 0.3273
dot_precision@5 0.1974
dot_precision@10 0.0993
dot_recall@1 0.917
dot_recall@3 0.982
dot_recall@5 0.987
dot_recall@10 0.993
dot_ndcg@10 0.9607
dot_mrr@10 0.9498
dot_map@100 0.95
query_active_dims 128.0
query_sparsity_ratio 0.9688
corpus_active_dims 128.0
corpus_sparsity_ratio 0.9688

Sparse Information Retrieval

Metric Value
dot_accuracy@1 0.94
dot_accuracy@3 0.989
dot_accuracy@5 0.992
dot_accuracy@10 0.995
dot_precision@1 0.94
dot_precision@3 0.3297
dot_precision@5 0.1984
dot_precision@10 0.0995
dot_recall@1 0.94
dot_recall@3 0.989
dot_recall@5 0.992
dot_recall@10 0.995
dot_ndcg@10 0.9723
dot_mrr@10 0.9645
dot_map@100 0.9646
query_active_dims 256.0
query_sparsity_ratio 0.9375
corpus_active_dims 256.0
corpus_sparsity_ratio 0.9375

Training Details

Training Dataset

natural-questions

  • Dataset: natural-questions at f9e894e
  • Size: 99,000 training samples
  • Columns: query and answer
  • Approximate statistics based on the first 1000 samples:
    query answer
    type string string
    details
    • min: 10 tokens
    • mean: 11.71 tokens
    • max: 26 tokens
    • min: 4 tokens
    • mean: 131.81 tokens
    • max: 450 tokens
  • Samples:
    query answer
    who played the father in papa don't preach Alex McArthur Alex McArthur (born March 6, 1957) is an American actor.
    where was the location of the battle of hastings Battle of Hastings The Battle of Hastings[a] was fought on 14 October 1066 between the Norman-French army of William, the Duke of Normandy, and an English army under the Anglo-Saxon King Harold Godwinson, beginning the Norman conquest of England. It took place approximately 7 miles (11 kilometres) northwest of Hastings, close to the present-day town of Battle, East Sussex, and was a decisive Norman victory.
    how many puppies can a dog give birth to Canine reproduction The largest litter size to date was set by a Neapolitan Mastiff in Manea, Cambridgeshire, UK on November 29, 2004; the litter was 24 puppies.[22]
  • Loss: CSRLoss with these parameters:
    {
    "beta": 0.1,
    "gamma": 0.1,
    "loss": "SparseMultipleNegativesRankingLoss(scale=1.0, similarity_fct='dot_score')"
}

Evaluation Dataset

natural-questions

  • Dataset: natural-questions at f9e894e
  • Size: 1,000 evaluation samples
  • Columns: query and answer
  • Approximate statistics based on the first 1000 samples:
    query answer
    type string string
    details
    • min: 10 tokens
    • mean: 11.69 tokens
    • max: 23 tokens
    • min: 15 tokens
    • mean: 134.01 tokens
    • max: 512 tokens
  • Samples:
    query answer
    where is the tiber river located in italy Tiber The Tiber (/ˈtaɪbər/, Latin: Tiberis,[1] Italian: Tevere [ˈteːvere])[2] is the third-longest river in Italy, rising in the Apennine Mountains in Emilia-Romagna and flowing 406 kilometres (252 mi) through Tuscany, Umbria and Lazio, where it is joined by the river Aniene, to the Tyrrhenian Sea, between Ostia and Fiumicino.[3] It drains a basin estimated at 17,375 square kilometres (6,709 sq mi). The river has achieved lasting fame as the main watercourse of the city of Rome, founded on its eastern banks.
    what kind of car does jay gatsby drive Jay Gatsby At the Buchanan home, Jordan Baker, Nick, Jay, and the Buchanans decide to visit New York City. Tom borrows Gatsby's yellow Rolls Royce to drive up to the city. On the way to New York City, Tom makes a detour at a gas station in "the Valley of Ashes", a run-down part of Long Island. The owner, George Wilson, shares his concern that his wife, Myrtle, may be having an affair. This unnerves Tom, who has been having an affair with Myrtle, and he leaves in a hurry.
    who sings if i can dream about you I Can Dream About You "I Can Dream About You" is a song performed by American singer Dan Hartman on the soundtrack album of the film Streets of Fire. Released in 1984 as a single from the soundtrack, and included on Hartman's album I Can Dream About You, it reached number 6 on the Billboard Hot 100.[1]
  • Loss: CSRLoss with these parameters:
    {
    "beta": 0.1,
    "gamma": 0.1,
    "loss": "SparseMultipleNegativesRankingLoss(scale=1.0, similarity_fct='dot_score')"
}

Training Hyperparameters

Non-Default Hyperparameters

  • eval_strategy: steps
  • per_device_train_batch_size: 64
  • per_device_eval_batch_size: 64
  • learning_rate: 4e-05
  • num_train_epochs: 1
  • bf16: 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: 64
  • per_device_eval_batch_size: 64
  • 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-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: 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: 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 nq_eval_4_dot_ndcg@10 nq_eval_8_dot_ndcg@10 nq_eval_16_dot_ndcg@10 nq_eval_32_dot_ndcg@10 nq_eval_64_dot_ndcg@10 nq_eval_128_dot_ndcg@10 nq_eval_256_dot_ndcg@10
-1 -1 - - 0.2464 0.4430 0.6677 0.8417 0.9258 0.9545 0.9654
0.0646 100 0.2916 - - - - - - - -
0.1293 200 0.2606 - - - - - - - -
0.1939 300 0.2553 0.2534 0.3060 0.4995 0.7232 0.8641 0.9406 0.9621 0.9697
0.2586 400 0.2496 - - - - - - - -
0.3232 500 0.2488 - - - - - - - -
0.3878 600 0.2475 0.2474 0.3331 0.5247 0.7336 0.8778 0.9350 0.9598 0.9692
0.4525 700 0.2477 - - - - - - - -
0.5171 800 0.2471 - - - - - - - -
0.5818 900 0.246 0.2452 0.3490 0.5252 0.7368 0.8760 0.9326 0.9634 0.9678
0.6464 1000 0.2449 - - - - - - - -
0.7111 1100 0.2444 - - - - - - - -
0.7757 1200 0.2422 0.2418 0.3356 0.5366 0.7431 0.8839 0.9358 0.9601 0.9703
0.8403 1300 0.2438 - - - - - - - -
0.9050 1400 0.2388 - - - - - - - -
0.9696 1500 0.2406 0.2410 0.3318 0.5338 0.7452 0.8818 0.9345 0.9618 0.9713
-1 -1 - - 0.3320 0.5319 0.7454 0.8831 0.9354 0.9607 0.9723

Environmental Impact

Carbon emissions were measured using CodeCarbon.

  • Energy Consumed: 0.115 kWh
  • Carbon Emitted: 0.045 kg of CO2
  • Hours Used: 0.29 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",
}

CSRLoss 

@misc{wen2025matryoshkarevisitingsparsecoding,
      title={Beyond Matryoshka: Revisiting Sparse Coding for Adaptive Representation},
      author={Tiansheng Wen and Yifei Wang and Zequn Zeng and Zhong Peng and Yudi Su and Xinyang Liu and Bo Chen and Hongwei Liu and Stefanie Jegelka and Chenyu You},
      year={2025},
      eprint={2503.01776},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2503.01776},
}

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}
}
Downloads last month
2
Safetensors
Model size
335M params
Tensor type
F32
·
Inference Providers NEW
Feature Extraction
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for tomaarsen/csr-mxbai-embed-large-v1-nq-dot-scale-1-gamma-0.1-detach-2

Base model

mixedbread-ai/mxbai-embed-large-v1
Finetuned
(36)
this model

Dataset used to train tomaarsen/csr-mxbai-embed-large-v1-nq-dot-scale-1-gamma-0.1-detach-2

Evaluation results

View on Papers With Code