This is a Cross Encoder model finetuned from microsoft/MiniLM-L12-H384-uncased on the ms_marco dataset using the sentence-transformers library. It computes scores for pairs of texts, which can be used for text reranking and semantic search.

Model Details

Model Description

Model Type: Cross Encoder
Base model: microsoft/MiniLM-L12-H384-uncased
Maximum Sequence Length: 512 tokens
Number of Output Labels: 1 label
Training Dataset:
- ms_marco
Language: en

Model Sources

Documentation: Sentence Transformers Documentation
Documentation: Cross Encoder Documentation
Repository: Sentence Transformers on GitHub
Hugging Face: Cross Encoders on Hugging Face

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 CrossEncoder

# Download from the 🤗 Hub
model = CrossEncoder("tomaarsen/reranker-msmarco-v1.1-MiniLM-L12-H384-uncased-bce")
# Get scores for pairs of texts
pairs = [
    ['is the calla lily poisonous', 'wholeblossoms.com. Even just a bite or two of the elegant calla lily can cause liver failure in pets and the mouths and tongues of humans to swell. The protein asparagine, found particularly concentrated in the roots of this plant, is poisonous, so be extra prudent with this popular cut flower. Bob Vila. Calla Lily. Even just a bite or two of the elegant calla lily can cause liver failure in pets and the mouths and tongues of humans to swell.'],
    ['how long do you have to go to school to be a nurse practitioner', "A: To become a nurse practitioner at the quickest, you can take up the certificate of nursing which is a one year program and the shortest to get nursing education. Certificate in nursing allows students to be able to practice and have introductory level knowledge and skills in the field of nursing. A: In order to become a nurse, you must complete a 2 year associate's or 4 years bachelor's degree program in nursing. Upon graduation, you should obtain an RN certification and pursue a specialization through an accredited MSN degree program."],
    ['what is true about atp molecule', 'Energy is usually liberated from the ATP molecule to do work in the cell by a reaction that removes one of the phosphate-oxygen groups, leaving adenosine diphosphate (ADP). When the ATP converts to ADP, the ATP is said to be spent. '],
    ['matter ejected from lungs is called', 'sputum. mucous secretion from the lungs, bronchi, and trachea that is ejected through the mouth, in contrast to saliva, which is the secretion of the salivary glands. Called also expectoration. sputum. mucous secretion from the lungs, bronchi and trachea which is ejected through the mouth by humans but not so in animals and it is assumed that it is swallowed. sputum cup. a small—1 inch diameter—cup on a long handle for the collection of sputum from the pharynx of a large animal. sputum specime'],
    ['average iron distances for senior golfers', 'Woods. Although is difficult to take all of these factors into consideration and determine an average distance for each club, there are charts that try. For the driver, a typical woman amateur can expect to get up to 180 yards. The 3-wood can provide 160 yards and the 5-wood 140. While there are charts that list the average distance for each golf club for women, the numbers are only averages and don’t reflect how far individuals actually hit the ball. There are many factors that influence distance, including your physical characteristics and conditions on the course.'],
]
scores = model.predict(pairs)
print(scores.shape)
# (5,)

# Or rank different texts based on similarity to a single text
ranks = model.rank(
    'is the calla lily poisonous',
    [
        'wholeblossoms.com. Even just a bite or two of the elegant calla lily can cause liver failure in pets and the mouths and tongues of humans to swell. The protein asparagine, found particularly concentrated in the roots of this plant, is poisonous, so be extra prudent with this popular cut flower. Bob Vila. Calla Lily. Even just a bite or two of the elegant calla lily can cause liver failure in pets and the mouths and tongues of humans to swell.',
        "A: To become a nurse practitioner at the quickest, you can take up the certificate of nursing which is a one year program and the shortest to get nursing education. Certificate in nursing allows students to be able to practice and have introductory level knowledge and skills in the field of nursing. A: In order to become a nurse, you must complete a 2 year associate's or 4 years bachelor's degree program in nursing. Upon graduation, you should obtain an RN certification and pursue a specialization through an accredited MSN degree program.",
        'Energy is usually liberated from the ATP molecule to do work in the cell by a reaction that removes one of the phosphate-oxygen groups, leaving adenosine diphosphate (ADP). When the ATP converts to ADP, the ATP is said to be spent. ',
        'sputum. mucous secretion from the lungs, bronchi, and trachea that is ejected through the mouth, in contrast to saliva, which is the secretion of the salivary glands. Called also expectoration. sputum. mucous secretion from the lungs, bronchi and trachea which is ejected through the mouth by humans but not so in animals and it is assumed that it is swallowed. sputum cup. a small—1 inch diameter—cup on a long handle for the collection of sputum from the pharynx of a large animal. sputum specime',
        'Woods. Although is difficult to take all of these factors into consideration and determine an average distance for each club, there are charts that try. For the driver, a typical woman amateur can expect to get up to 180 yards. The 3-wood can provide 160 yards and the 5-wood 140. While there are charts that list the average distance for each golf club for women, the numbers are only averages and don’t reflect how far individuals actually hit the ball. There are many factors that influence distance, including your physical characteristics and conditions on the course.',
    ]
)
# [{'corpus_id': ..., 'score': ...}, {'corpus_id': ..., 'score': ...}, ...]

Evaluation

Metrics

Cross Encoder Reranking

Datasets: NanoMSMARCO_R100, NanoNFCorpus_R100 and NanoNQ_R100
Evaluated with CrossEncoderRerankingEvaluator with these parameters:
json
1{ 2 "at_k": 10, 3 "always_rerank_positives": true 4}

Metric	NanoMSMARCO_R100	NanoNFCorpus_R100	NanoNQ_R100
map	0.4241 (-0.0655)	0.2867 (+0.0257)	0.5166 (+0.0970)
mrr@10	0.4173 (-0.0602)	0.4599 (-0.0399)	0.5251 (+0.0984)
ndcg@10	0.5014 (-0.0390)	0.2880 (-0.0370)	0.5684 (+0.0678)

Cross Encoder Nano BEIR

Dataset: NanoBEIR_R100_mean
Evaluated with CrossEncoderNanoBEIREvaluator with these parameters:
json
1{ 2 "dataset_names": [ 3 "msmarco", 4 "nfcorpus", 5 "nq" 6 ], 7 "rerank_k": 100, 8 "at_k": 10, 9 "always_rerank_positives": true 10}

Metric	Value
map	0.4091 (+0.0191)
mrr@10	0.4674 (-0.0006)
ndcg@10	0.4526 (-0.0027)

Training Details

Training Dataset

ms_marco

Dataset: ms_marco at a47ee7a
Size: 666,193 training samples
Columns: query, passage, and label
Approximate statistics based on the first 1000 samples:
query passage label
type string string int
details
min: 11 characters
mean: 34.43 characters
max: 90 characters
min: 62 characters
mean: 423.32 characters
max: 866 characters
0: ~85.50%
1: ~14.50%

Samples:

query	passage	label
`is a meal allowance taxable`	1 An exemption applies of up to three months for meal payments if the employee is required to work away from their normal work location because they're travelling on business. 2 This may be for a specific short-term, work-related journey or for a longer period such as a secondment to a distant work location. The rules also don't apply in the case where fringe benefit tax (FBT) is payable on meals directly provided by the employer such as a discount in the work canteen. The rules that limit employers' deductions of entertainment expenses also continue to apply.	`0`
`average temperature may alaska`	`Precipitation [edit]. Juneau averages over 50 inches (1,270 mm) of precipitation a year, while some other areas in southeast Alaska receive as much as 275 inches (6,980 mm). Average monthly precipitation generally peaks in September or October, and is lowest in May and June.`	`0`
`asic registered fixed and floating charge`	Fixed and Floating Charge. Definition: A fixed and floating charge is a form of security interest usually taken by a lender from a company to secure repayment of a loan. The company granting the charge is usually referred to as the Chargor and the person in whose favour the charge is granted is typically called the Chargee . This form of security is called fixed and floating because the chargee has: The true ingenuity of the fixed and floating charge is that it permits the chargor to deal with and sell their stock etc in the ordinary course of business without requiring consent from the chargor.	`0`

Loss: BinaryCrossEntropyLoss with these parameters:
json
1{ 2 "activation_fct": "torch.nn.modules.linear.Identity", 3 "pos_weight": null 4}

Evaluation Dataset

ms_marco

Dataset: ms_marco at a47ee7a
Size: 10,000 evaluation samples
Columns: query, passage, and label
Approximate statistics based on the first 1000 samples:
query passage label
type string string int
details
min: 11 characters
mean: 34.23 characters
max: 137 characters
min: 88 characters
mean: 427.89 characters
max: 867 characters
0: ~86.70%
1: ~13.30%

Samples:

query	passage	label
`is the calla lily poisonous`	`wholeblossoms.com. Even just a bite or two of the elegant calla lily can cause liver failure in pets and the mouths and tongues of humans to swell. The protein asparagine, found particularly concentrated in the roots of this plant, is poisonous, so be extra prudent with this popular cut flower. Bob Vila. Calla Lily. Even just a bite or two of the elegant calla lily can cause liver failure in pets and the mouths and tongues of humans to swell.`	`1`
`how long do you have to go to school to be a nurse practitioner`	A: To become a nurse practitioner at the quickest, you can take up the certificate of nursing which is a one year program and the shortest to get nursing education. Certificate in nursing allows students to be able to practice and have introductory level knowledge and skills in the field of nursing. A: In order to become a nurse, you must complete a 2 year associate's or 4 years bachelor's degree program in nursing. Upon graduation, you should obtain an RN certification and pursue a specialization through an accredited MSN degree program.	`0`
`what is true about atp molecule`	`Energy is usually liberated from the ATP molecule to do work in the cell by a reaction that removes one of the phosphate-oxygen groups, leaving adenosine diphosphate (ADP). When the ATP converts to ADP, the ATP is said to be spent.`	`0`

Loss: BinaryCrossEntropyLoss with these parameters:
json
1{ 2 "activation_fct": "torch.nn.modules.linear.Identity", 3 "pos_weight": null 4}

Training Hyperparameters

Non-Default Hyperparameters

eval_strategy: steps
per_device_train_batch_size: 32
per_device_eval_batch_size: 32
learning_rate: 2e-05
num_train_epochs: 1
warmup_ratio: 0.1
seed: 12
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: 32
per_device_eval_batch_size: 32
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: 0.1
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: 12
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
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: batch_sampler
multi_dataset_batch_sampler: proportional

Training Logs

Epoch	Step	Training Loss	Validation Loss	NanoMSMARCO_R100_ndcg@10	NanoNFCorpus_R100_ndcg@10	NanoNQ_R100_ndcg@10	NanoBEIR_R100_mean_ndcg@10
-1	-1	-	-	0.0357 (-0.5047)	0.2806 (-0.0444)	0.0527 (-0.4479)	0.1230 (-0.3323)
0.0000	1	0.6838	-	-	-	-	-
0.0480	1000	0.4514	-	-	-	-	-
0.0961	2000	0.385	-	-	-	-	-
0.1441	3000	0.3892	-	-	-	-	-
0.1921	4000	0.3819	0.3777	0.2751 (-0.2653)	0.2555 (-0.0696)	0.2689 (-0.2318)	0.2665 (-0.1889)
0.2402	5000	0.3694	-	-	-	-	-
0.2882	6000	0.3641	-	-	-	-	-
0.3362	7000	0.3679	-	-	-	-	-
0.3843	8000	0.3674	0.3566	0.4745 (-0.0660)	0.2459 (-0.0792)	0.4967 (-0.0040)	0.4057 (-0.0497)
0.4323	9000	0.3677	-	-	-	-	-
0.4803	10000	0.3606	-	-	-	-	-
0.5284	11000	0.3651	-	-	-	-	-
0.5764	12000	0.3586	0.3556	0.4666 (-0.0738)	0.2720 (-0.0531)	0.5513 (+0.0506)	0.4300 (-0.0254)
0.6244	13000	0.3535	-	-	-	-	-
0.6725	14000	0.3551	-	-	-	-	-
0.7205	15000	0.3552	-	-	-	-	-
0.7685	16000	0.3516	0.3439	0.5014 (-0.0390)	0.2880 (-0.0370)	0.5684 (+0.0678)	0.4526 (-0.0027)
0.8166	17000	0.3486	-	-	-	-	-
0.8646	18000	0.3528	-	-	-	-	-
0.9126	19000	0.3526	-	-	-	-	-
0.9607	20000	0.3415	0.3423	0.4898 (-0.0507)	0.2794 (-0.0457)	0.5792 (+0.0785)	0.4494 (-0.0059)
-1	-1	-	-	0.5014 (-0.0390)	0.2880 (-0.0370)	0.5684 (+0.0678)	0.4526 (-0.0027)

The bold row denotes the saved checkpoint.

Environmental Impact

Carbon emissions were measured using CodeCarbon.

Energy Consumed: 0.178 kWh
Carbon Emitted: 0.069 kg of CO2
Hours Used: 0.525 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: 3.5.0.dev0
Transformers: 4.49.0
PyTorch: 2.6.0+cu124
Accelerate: 1.4.0
Datasets: 3.3.2
Tokenizers: 0.21.0

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",
}

	query	passage	label
type	string	string	int
details	min: 11 characters mean: 34.43 characters max: 90 characters	min: 62 characters mean: 423.32 characters max: 866 characters	0: ~85.50% 1: ~14.50%