You can deploy this model and start earning money today!

tomaarsen/
csr-mxbai-embed-large-v1-nq-cos-sim-scale-20-gamma-1

Name: csr-mxbai-embed-large-v1-nq-cos-sim-scale-20-gamma-1
Availability: InStock

Views

No views yet

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: Cosine Similarity
Training Dataset:
- natural-questions
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): 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.

python

from sentence_transformers import SparseEncoder

# Download from the 🤗 Hub
model = SparseEncoder("tomaarsen/csr-mxbai-embed-large-v1-nq-cos-sim-scale-20-gamma-1")
# 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([[0.7062, 0.2414, 0.2065]])

Evaluation

Metrics

Sparse Information Retrieval

Datasets: NanoMSMARCO_4 and NanoNQ_4
Evaluated with SparseInformationRetrievalEvaluator with these parameters:
json
1{ 2 "max_active_dims": 4 3}

Metric	NanoMSMARCO_4	NanoNQ_4
cosine_accuracy@1	0.02	0.1
cosine_accuracy@3	0.12	0.16
cosine_accuracy@5	0.18	0.2
cosine_accuracy@10	0.26	0.26
cosine_precision@1	0.02	0.1
cosine_precision@3	0.04	0.0533
cosine_precision@5	0.036	0.04
cosine_precision@10	0.026	0.026
cosine_recall@1	0.02	0.1
cosine_recall@3	0.12	0.16
cosine_recall@5	0.18	0.19
cosine_recall@10	0.26	0.24
cosine_ndcg@10	0.131	0.1618
cosine_mrr@10	0.0911	0.1391
cosine_map@100	0.1006	0.1455
query_active_dims	4.0	4.0
query_sparsity_ratio	0.999	0.999
corpus_active_dims	4.0	4.0
corpus_sparsity_ratio	0.999	0.999

Sparse Nano BEIR

Dataset: NanoBEIR_mean_4
Evaluated with SparseNanoBEIREvaluator with these parameters:
json
1{ 2 "dataset_names": [ 3 "msmarco", 4 "nq" 5 ], 6 "max_active_dims": 4 7}

Metric	Value
cosine_accuracy@1	0.06
cosine_accuracy@3	0.14
cosine_accuracy@5	0.19
cosine_accuracy@10	0.26
cosine_precision@1	0.06
cosine_precision@3	0.0467
cosine_precision@5	0.038
cosine_precision@10	0.026
cosine_recall@1	0.06
cosine_recall@3	0.14
cosine_recall@5	0.185
cosine_recall@10	0.25
cosine_ndcg@10	0.1464
cosine_mrr@10	0.1151
cosine_map@100	0.123
query_active_dims	4.0
query_sparsity_ratio	0.999
corpus_active_dims	4.0
corpus_sparsity_ratio	0.999

Sparse Information Retrieval

Datasets: NanoMSMARCO_16 and NanoNQ_16
Evaluated with SparseInformationRetrievalEvaluator with these parameters:
json
1{ 2 "max_active_dims": 16 3}

Metric	NanoMSMARCO_16	NanoNQ_16
cosine_accuracy@1	0.14	0.14
cosine_accuracy@3	0.32	0.32
cosine_accuracy@5	0.44	0.42
cosine_accuracy@10	0.62	0.54
cosine_precision@1	0.14	0.14
cosine_precision@3	0.1067	0.1067
cosine_precision@5	0.088	0.084
cosine_precision@10	0.062	0.054
cosine_recall@1	0.14	0.14
cosine_recall@3	0.32	0.31
cosine_recall@5	0.44	0.4
cosine_recall@10	0.62	0.51
cosine_ndcg@10	0.3523	0.3159
cosine_mrr@10	0.2692	0.2584
cosine_map@100	0.2835	0.2664
query_active_dims	16.0	16.0
query_sparsity_ratio	0.9961	0.9961
corpus_active_dims	16.0	16.0
corpus_sparsity_ratio	0.9961	0.9961

Sparse Nano BEIR

Dataset: NanoBEIR_mean_16
Evaluated with SparseNanoBEIREvaluator with these parameters:
json
1{ 2 "dataset_names": [ 3 "msmarco", 4 "nq" 5 ], 6 "max_active_dims": 16 7}

Metric	Value
cosine_accuracy@1	0.14
cosine_accuracy@3	0.32
cosine_accuracy@5	0.43
cosine_accuracy@10	0.58
cosine_precision@1	0.14
cosine_precision@3	0.1067
cosine_precision@5	0.086
cosine_precision@10	0.058
cosine_recall@1	0.14
cosine_recall@3	0.315
cosine_recall@5	0.42
cosine_recall@10	0.565
cosine_ndcg@10	0.3341
cosine_mrr@10	0.2638
cosine_map@100	0.2749
query_active_dims	16.0
query_sparsity_ratio	0.9961
corpus_active_dims	16.0
corpus_sparsity_ratio	0.9961

Sparse Information Retrieval

Datasets: NanoMSMARCO_64 and NanoNQ_64
Evaluated with SparseInformationRetrievalEvaluator with these parameters:
json
1{ 2 "max_active_dims": 64 3}

Metric	NanoMSMARCO_64	NanoNQ_64
cosine_accuracy@1	0.42	0.36
cosine_accuracy@3	0.6	0.58
cosine_accuracy@5	0.74	0.74
cosine_accuracy@10	0.78	0.78
cosine_precision@1	0.42	0.36
cosine_precision@3	0.2	0.2
cosine_precision@5	0.148	0.152
cosine_precision@10	0.078	0.082
cosine_recall@1	0.42	0.34
cosine_recall@3	0.6	0.54
cosine_recall@5	0.74	0.68
cosine_recall@10	0.78	0.73
cosine_ndcg@10	0.5989	0.5402
cosine_mrr@10	0.5405	0.4945
cosine_map@100	0.5486	0.4793
query_active_dims	64.0	64.0
query_sparsity_ratio	0.9844	0.9844
corpus_active_dims	64.0	64.0
corpus_sparsity_ratio	0.9844	0.9844

Sparse Nano BEIR

Dataset: NanoBEIR_mean_64
Evaluated with SparseNanoBEIREvaluator with these parameters:
json
1{ 2 "dataset_names": [ 3 "msmarco", 4 "nq" 5 ], 6 "max_active_dims": 64 7}

Metric	Value
cosine_accuracy@1	0.39
cosine_accuracy@3	0.59
cosine_accuracy@5	0.74
cosine_accuracy@10	0.78
cosine_precision@1	0.39
cosine_precision@3	0.2
cosine_precision@5	0.15
cosine_precision@10	0.08
cosine_recall@1	0.38
cosine_recall@3	0.57
cosine_recall@5	0.71
cosine_recall@10	0.755
cosine_ndcg@10	0.5695
cosine_mrr@10	0.5175
cosine_map@100	0.5139
query_active_dims	64.0
query_sparsity_ratio	0.9844
corpus_active_dims	64.0
corpus_sparsity_ratio	0.9844

Sparse Information Retrieval

Datasets: NanoMSMARCO_256 and NanoNQ_256
Evaluated with SparseInformationRetrievalEvaluator with these parameters:
json
1{ 2 "max_active_dims": 256 3}

Metric	NanoMSMARCO_256	NanoNQ_256
cosine_accuracy@1	0.44	0.56
cosine_accuracy@3	0.62	0.72
cosine_accuracy@5	0.68	0.78
cosine_accuracy@10	0.82	0.86
cosine_precision@1	0.44	0.56
cosine_precision@3	0.2067	0.24
cosine_precision@5	0.136	0.16
cosine_precision@10	0.082	0.092
cosine_recall@1	0.44	0.54
cosine_recall@3	0.62	0.67
cosine_recall@5	0.68	0.72
cosine_recall@10	0.82	0.82
cosine_ndcg@10	0.6219	0.6834
cosine_mrr@10	0.5601	0.6571
cosine_map@100	0.5703	0.638
query_active_dims	256.0	256.0
query_sparsity_ratio	0.9375	0.9375
corpus_active_dims	256.0	256.0
corpus_sparsity_ratio	0.9375	0.9375

Sparse Nano BEIR

Dataset: NanoBEIR_mean_256
Evaluated with SparseNanoBEIREvaluator with these parameters:
json
1{ 2 "dataset_names": [ 3 "msmarco", 4 "nq" 5 ], 6 "max_active_dims": 256 7}

Metric	Value
cosine_accuracy@1	0.5
cosine_accuracy@3	0.67
cosine_accuracy@5	0.73
cosine_accuracy@10	0.84
cosine_precision@1	0.5
cosine_precision@3	0.2233
cosine_precision@5	0.148
cosine_precision@10	0.087
cosine_recall@1	0.49
cosine_recall@3	0.645
cosine_recall@5	0.7
cosine_recall@10	0.82
cosine_ndcg@10	0.6527
cosine_mrr@10	0.6086
cosine_map@100	0.6042
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:
json
1{ 2 "beta": 0.1, 3 "gamma": 1.0, 4 "loss": "SparseMultipleNegativesRankingLoss(scale=20.0, similarity_fct='cos_sim')" 5}

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:
json
1{ 2 "beta": 0.1, 3 "gamma": 1.0, 4 "loss": "SparseMultipleNegativesRankingLoss(scale=20.0, similarity_fct='cos_sim')" 5}

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
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: 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: 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	NanoMSMARCO_4_cosine_ndcg@10	NanoNQ_4_cosine_ndcg@10	NanoBEIR_mean_4_cosine_ndcg@10	NanoMSMARCO_16_cosine_ndcg@10	NanoNQ_16_cosine_ndcg@10	NanoBEIR_mean_16_cosine_ndcg@10	NanoMSMARCO_64_cosine_ndcg@10	NanoNQ_64_cosine_ndcg@10	NanoBEIR_mean_64_cosine_ndcg@10	NanoMSMARCO_256_cosine_ndcg@10	NanoNQ_256_cosine_ndcg@10	NanoBEIR_mean_256_cosine_ndcg@10
-1	-1	-	-	0.0850	0.1222	0.1036	0.4256	0.3267	0.3761	0.5827	0.5843	0.5835	0.5987	0.7005	0.6496
0.0646	100	0.6568	-	-	-	-	-	-	-	-	-	-	-	-	-
0.1293	200	0.561	-	-	-	-	-	-	-	-	-	-	-	-	-
0.1939	300	0.5248	0.4118	0.131	0.1618	0.1464	0.3523	0.3159	0.3341	0.5989	0.5402	0.5695	0.6219	0.6834	0.6527
0.2586	400	0.4995	-	-	-	-	-	-	-	-	-	-	-	-	-
0.3232	500	0.484	-	-	-	-	-	-	-	-	-	-	-	-	-
0.3878	600	0.4773	0.3882	0.2023	0.1465	0.1744	0.3397	0.3617	0.3507	0.5710	0.5702	0.5706	0.6091	0.6610	0.6351
0.4525	700	0.464	-	-	-	-	-	-	-	-	-	-	-	-	-
0.5171	800	0.4529	-	-	-	-	-	-	-	-	-	-	-	-	-
0.5818	900	0.4524	0.3753	0.1495	0.1179	0.1337	0.3072	0.3473	0.3272	0.5718	0.5525	0.5622	0.6084	0.6660	0.6372
0.6464	1000	0.4486	-	-	-	-	-	-	-	-	-	-	-	-	-
0.7111	1100	0.4349	-	-	-	-	-	-	-	-	-	-	-	-	-
0.7757	1200	0.4382	0.3690	0.1815	0.0924	0.1370	0.3328	0.3493	0.3410	0.5311	0.5480	0.5396	0.6086	0.6486	0.6286
0.8403	1300	0.4394	-	-	-	-	-	-	-	-	-	-	-	-	-
0.9050	1400	0.427	-	-	-	-	-	-	-	-	-	-	-	-	-
0.9696	1500	0.4312	0.3666	0.1746	0.1350	0.1548	0.3395	0.2952	0.3174	0.5511	0.5252	0.5381	0.6162	0.6494	0.6328
-1	-1	-	-	0.1310	0.1618	0.1464	0.3523	0.3159	0.3341	0.5989	0.5402	0.5695	0.6219	0.6834	0.6527

The bold row denotes the saved checkpoint.

Environmental Impact

Carbon emissions were measured using CodeCarbon.

Energy Consumed: 0.145 kWh
Carbon Emitted: 0.056 kg of CO2
Hours Used: 0.379 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

bibtex

@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

bibtex

@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

bibtex

@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}
}

	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