/ Модель

Характеристики

Контекст
7 999 токенов
Параметры
31,9 млн
Архитектура
modernbert · dense
Базовая модель
ettin-encoder-32m
Библиотека
sentence-transformers
Опубликована
24 ноя 2025

ms-marco-ettin-32m-reranker

Ранжирование текстаТекст → score6 профессий

Англоязычный Cross Encoder на ModernBERT для оценки релевантности пар запрос–документ и переранжирования результатов поиска.

233скачиваний за 30 дней0лайковвеса обновлены24 ноя 2025

Бенчмарки

Как модель справилась с задачами пользователей ailist и что о ней публикуют открытые лидерборды

Тесты пользователейпока нет

Никто ещё не опубликовал тест этой модели. Соберите бенчмарк из своих задач — результат появится здесь.

  1. Добавьте модель в бенчмарк

    Вместе с другими моделями категории «Ранжирование» — сравнение идёт внутри одной категории.

  2. Опишите свои задачи

    Напишите запросы, при необходимости приложите файлы и эталонные ответы.

  3. Запустите и опубликуйте

    Модели ответят на каждую задачу, судья оценит ответы. Опубликованный результат появится на этой странице.

Публичные оценки

Из открытых лидербордов. Сравнивайте результаты одного теста при одинаковых условиях — отсутствие оценки не значит, что модель хуже.

В подключённых источниках результатов для этой модели пока нет.

Что умеет модель

Описание составлено по карточке модели и метаданным Hugging Face

Это англоязычный Cross Encoder на базе jhu-clsp/ettin-encoder-32m, дообученный авторами Sentence Transformers на датасете MS MARCO. Модель использует архитектуру ModernBERT, содержит 31,9 млн параметров и принимает пару текстов, чтобы оценить их релевантность; результатом служит один числовой балл. Для этой модели указано максимальное ограничение в 512 токенов на последовательность.

Модель предназначена для переранжирования результатов поиска и семантического поиска: её можно применять для переоценки кандидатов, уже найденных поисковой системой, чтобы поднять наиболее подходящие документы выше. В README приведены результаты оценки на наборах NanoMSMARCO, NanoNFCorpus, NanoNQ и NanoBEIR.

Модель распространяется в формате safetensors и рассчитана на использование с библиотекой sentence-transformers. Сведения о лицензии в предоставленных материалах отсутствуют.

Сценарии применения

  • переранжирование документов по запросу
  • уточнение порядка результатов поисковой выдачи
  • оценка релевантности пары запрос–документ
  • семантический поиск по корпусу текстов

Ограничения

Модель рассчитана на английский язык и ограничивает входную последовательность 512 токенами. Она выдаёт оценку релевантности, а не текстовый ответ; в предоставленных материалах лицензия не указана.

Для каких профессий подходит6

Разделы каталога профессий, в задачах которых используется эта модель

Все профессии

Также подходит для

Карточка модели

README от автора на Hugging Face, как он его опубликовал

Открыть на Hugging Face

CrossEncoder based on jhu-clsp/ettin-encoder-32m

This is a Cross Encoder model finetuned from jhu-clsp/ettin-encoder-32m on the msmarco 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: jhu-clsp/ettin-encoder-32m
  • Maximum Sequence Length: 512 tokens
  • Number of Output Labels: 1 label
  • Training Dataset:
  • Language: en

Model Sources

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 CrossEncoder

# Download from the 🤗 Hub
model = CrossEncoder("tomaarsen/ms-marco-ettin-32m-reranker")
# Get scores for pairs of texts
pairs = [
    ['what is honor society', 'In the United States, an honor society is a rank organization that recognizes excellence among peers. Numerous societies recognize various fields and circumstances. The Order of the Arrow, for example, is the national honor society of the Boy Scouts of America.'],
    ['what happens to blood pressure when you raise your arm', 'Well, you measured blood pressure in that arm would drop because while the pressure that your heart puts out will stay pretty constant, your arm is now higher and requires more energy to reach the end, so the pressure seen there will be lower. The Doc · 7 years ago.'],
    ['what country is the name astrid from', 'Comment by silkfire. Astrid, Estrid, Ã\x86striðr æstriðr Ã\x81strÃ\xadðr astriðr ástrÃ\xadðr is a given Name Of north. Germanic origin it Comes From Ã\x81sfrÃ\xadðr (Norse Asfriðr), ásfrÃ\xadðr divine (beauty) + from (ass). áss god friðr FrÃ\xadðr beautiful names derived From, astrid include the name astrida which is a somewhat common name for Girls. in the country of latviat comes from Old Norse Ã\x81sfrÃ\xadðr (ásfrÃ\xadðr Divine), beauty from (ass) + áss (god). Friðr frÃ\xadðr beautiful Names derived from astrid Include, the name astrida which is a somewhat common name for girls in The. country of latvia'],
    ['where the latin people came from', 'Latina is a city in Italy and the Latin people come from southern Italy around Rome.'],
    ['how long do cocker spaniels', "the cocker spaniel's life span is 12-15 years. a dog that is kept fit and on nutritious dog food has a better chance of living longer than those who aren't. I would assume roughly 10-15 years. My cocker spaniel is 13.5 years old and just starting to lose her health."],
]
scores = model.predict(pairs)
print(scores.shape)
# (5,)

# Or rank different texts based on similarity to a single text
ranks = model.rank(
    'what is honor society',
    [
        'In the United States, an honor society is a rank organization that recognizes excellence among peers. Numerous societies recognize various fields and circumstances. The Order of the Arrow, for example, is the national honor society of the Boy Scouts of America.',
        'Well, you measured blood pressure in that arm would drop because while the pressure that your heart puts out will stay pretty constant, your arm is now higher and requires more energy to reach the end, so the pressure seen there will be lower. The Doc · 7 years ago.',
        'Comment by silkfire. Astrid, Estrid, Ã\x86striðr æstriðr Ã\x81strÃ\xadðr astriðr ástrÃ\xadðr is a given Name Of north. Germanic origin it Comes From Ã\x81sfrÃ\xadðr (Norse Asfriðr), ásfrÃ\xadðr divine (beauty) + from (ass). áss god friðr FrÃ\xadðr beautiful names derived From, astrid include the name astrida which is a somewhat common name for Girls. in the country of latviat comes from Old Norse Ã\x81sfrÃ\xadðr (ásfrÃ\xadðr Divine), beauty from (ass) + áss (god). Friðr frÃ\xadðr beautiful Names derived from astrid Include, the name astrida which is a somewhat common name for girls in The. country of latvia',
        'Latina is a city in Italy and the Latin people come from southern Italy around Rome.',
        "the cocker spaniel's life span is 12-15 years. a dog that is kept fit and on nutritious dog food has a better chance of living longer than those who aren't. I would assume roughly 10-15 years. My cocker spaniel is 13.5 years old and just starting to lose her health.",
    ]
)
# [{'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:
    {
        "at_k": 10,
        "always_rerank_positives": true
    }
    
MetricNanoMSMARCO_R100NanoNFCorpus_R100NanoNQ_R100
map0.6380 (+0.1484)0.3547 (+0.0937)0.6748 (+0.2552)
mrr@100.6319 (+0.1544)0.6102 (+0.1104)0.6919 (+0.2652)
ndcg@100.6951 (+0.1547)0.4138 (+0.0888)0.7294 (+0.2287)

Cross Encoder Nano BEIR

  • Dataset: NanoBEIR_R100_mean
  • Evaluated with CrossEncoderNanoBEIREvaluator with these parameters:
    {
        "dataset_names": [
            "msmarco",
            "nfcorpus",
            "nq"
        ],
        "rerank_k": 100,
        "at_k": 10,
        "always_rerank_positives": true
    }
    
MetricValue
map0.5558 (+0.1658)
mrr@100.6446 (+0.1766)
ndcg@100.6128 (+0.1574)

Training Details

Training Dataset

msmarco

  • Dataset: msmarco at 9e329ed
  • Size: 39,770,704 training samples
  • Columns: query_id, positive_id, negative_id, and score
  • Approximate statistics based on the first 1000 samples:
    query_idpositive_idnegative_idscore
    typestringstringstringfloat
    detailsmin: 11 charactersmean: 33.98 charactersmax: 124 charactersmin: 55 charactersmean: 348.53 charactersmax: 926 charactersmin: 80 charactersmean: 340.54 charactersmax: 818 charactersmin: -3.12mean: 13.18max: 22.49
  • Samples:
    query_idpositive_idnegative_idscore
    which artist started fauves movementLes Fauves (Wild Beasts) • Fauvism was the first twentieth-century movement in modern art. Inspired by. the examples of van Gogh, Gauguin, Seurat, and Cézanne. • It grew out of a loosely allied group of French painters with shared interests. Henri Matisse was eventually recognized as the leader of Les Fauves, or The. Wild Beasts”.Let’s get a reflection onto your glass, for added realism. Architectural 2D to 3D visualizer and graphic artist Jonathan Pagaduan Ignas shows you how in this great little tutorial he has contributed to us here at SketchUpArtists.Here are a few easy steps to a glass reflection effect using Google SketchUp and V-Ray for SketchUp..let’s get started!Step 1 - Import Image File. Step 2 - Place Image in Glass. Step 3 - Position Texture. Step 4 - Sample Texture with Dropper.rchitectural 2D to 3D visualizer and graphic artist Jonathan Pagaduan Ignas shows you how in this great little tutorial he has contributed to us here at SketchUpArtists. Here are a few easy steps to a glass reflection effect using Google SketchUp and V-Ray for SketchUp..let’s get started!18.45222806930542
    what foods are high in uric acid listAvoid high-purine foods. See attached lists. Avoid or limit alcohol. Alcohol increases purine production, leading to higher uric acid levels in your blood and urine. Limit meat to 3 ounces per meal. Limit high-fat foods such as salad dressings, ice cream, fried foods, gravies, and dressings.Fat holds onto uric acid in your kidneys.lcohol increases purine production, leading to higher uric acid levels in your blood and urine. Limit meat to 3 ounces per meal. Limit high-fat foods such as salad dressings, ice cream, fried foods, gravies, and dressings. Fat holds onto uric acid in your kidneys.You can develop gout if you have too much uric acid in your body. Uric acid is a chemical that everyone has in their blood. It's a waste product that forms from substances called purines. These are found in every cell in your body, and in certain foods, such as seafood and liver.7.823118130366007
    what battle was the turning point in the pacific theatreThe campaign lasted for nearly six months and cost thousands of American and Japanese lives. This campaign, from the naval perspective, and indeed from a strategic perspective, was the turning point of World War Two in the Pacific Theater. The experienced and well trained Japanese fleet was still superior to the United States fleet after the Battle of Midway, and was still a threat to the US fleet until it was checked in the actions off Guadalcanal.It is famous as the setting of the February–March 1945 Battle of Iwo Jima involving the United States and a small number of elements of the British Pacific Fleet versus the Empire of Japan during World War II.7.8578096230824785
  • Loss: MarginMSELoss with these parameters:
    {
        "activation_fn": "torch.nn.modules.linear.Identity"
    }
    

Evaluation Dataset

msmarco

  • Dataset: msmarco at 9e329ed
  • Size: 10,000 evaluation samples
  • Columns: query_id, positive_id, negative_id, and score
  • Approximate statistics based on the first 1000 samples:
    query_idpositive_idnegative_idscore
    typestringstringstringfloat
    detailsmin: 11 charactersmean: 34.75 charactersmax: 124 charactersmin: 69 charactersmean: 346.11 charactersmax: 920 charactersmin: 79 charactersmean: 350.68 charactersmax: 1014 charactersmin: -2.41mean: 13.41max: 22.48
  • Samples:
    query_idpositive_idnegative_idscore
    what is honor societyIn the United States, an honor society is a rank organization that recognizes excellence among peers. Numerous societies recognize various fields and circumstances. The Order of the Arrow, for example, is the national honor society of the Boy Scouts of America.Pi Gamma Mu Headlines. Pi Gamma Mu is the oldest and preeminent honor society in the social sciences. Our mission is to encourage and recognize superior scholarship in social science disciplines and to foster cooperation and social service among its members.7.578772008419037
    what happens to blood pressure when you raise your armWell, you measured blood pressure in that arm would drop because while the pressure that your heart puts out will stay pretty constant, your arm is now higher and requires more energy to reach the end, so the pressure seen there will be lower. The Doc · 7 years ago.Constriction of blood vessels and increase in heart rate does raise blood pressure, but only temporarily; when the stress reaction goes away, blood pressure returns to its pre-stress level.This is called situational stress, and its effects are generally short-lived and disappear when the stressful event is over.hronic (constant) stress causes our bodies to go into high gear on and off for days or weeks at a time. The links between chronic stress and blood pressure are not clear. Although stress does not clearly cause heart disease, it can play a role in general wellness.8.200530961155891
    what country is the name astrid fromComment by silkfire. Astrid, Estrid, Æstriðr æstriðr Ástríðr astriðr ástríðr is a given Name Of north. Germanic origin it Comes From Ásfríðr (Norse Asfriðr), ásfríðr divine (beauty) + from (ass). áss god friðr Fríðr beautiful names derived From, astrid include the name astrida which is a somewhat common name for Girls. in the country of latviat comes from Old Norse Ásfríðr (ásfríðr Divine), beauty from (ass) + áss (god). Friðr fríðr beautiful Names derived from astrid Include, the name astrida which is a somewhat common name for girls in The. country of latvia» Follow author. » Share. Astrid Ayrianto was on omeegle one day on the video chat, little did she know, she had almost everything in common with the lead singer of her favorite band Get Scared, Nicholas Matthews.10.449922064940136
  • Loss: MarginMSELoss with these parameters:
    {
        "activation_fn": "torch.nn.modules.linear.Identity"
    }
    

Training Hyperparameters

Non-Default Hyperparameters

  • eval_strategy: steps
  • per_device_train_batch_size: 256
  • per_device_eval_batch_size: 256
  • learning_rate: 2e-05
  • num_train_epochs: 1
  • warmup_ratio: 0.1
  • seed: 12
  • bf16: True
  • load_best_model_at_end: True

All Hyperparameters

  • overwrite_output_dir: False
  • do_predict: False
  • eval_strategy: steps
  • prediction_loss_only: True
  • per_device_train_batch_size: 256
  • per_device_eval_batch_size: 256
  • 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
  • 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: True
  • 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}
  • parallelism_config: None
  • deepspeed: None
  • label_smoothing_factor: 0.0
  • optim: adamw_torch_fused
  • optim_args: None
  • adafactor: False
  • group_by_length: False
  • length_column_name: length
  • project: huggingface
  • trackio_space_id: trackio
  • 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
  • hub_revision: None
  • 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: no
  • neftune_noise_alpha: None
  • optim_target_modules: None
  • batch_eval_metrics: False
  • eval_on_start: False
  • use_liger_kernel: False
  • liger_kernel_config: None
  • eval_use_gather_object: False
  • average_tokens_across_devices: True
  • prompts: None
  • batch_sampler: batch_sampler
  • multi_dataset_batch_sampler: proportional
  • router_mapping: {}
  • learning_rate_mapping: {}

Training Logs

EpochStepTraining LossValidation LossNanoMSMARCO_R100_ndcg@10NanoNFCorpus_R100_ndcg@10NanoNQ_R100_ndcg@10NanoBEIR_R100_mean_ndcg@10
-1-1--0.0694 (-0.4711)0.2964 (-0.0286)0.0281 (-0.4725)0.1313 (-0.3241)
0.00011209.9035-----
0.002039207.8315-----
0.004078204.3864-----
0.0060117197.3435-----
0.0080156188.2296-----
0.0100195176.2927169.83260.1222 (-0.4182)0.2790 (-0.0460)0.0840 (-0.4166)0.1618 (-0.2936)
0.0121234161.1845-----
0.0141273145.5738-----
0.0161312122.6948-----
0.018135181.3822-----
0.020139043.111735.89880.5569 (+0.0164)0.3702 (+0.0452)0.4567 (-0.0439)0.4613 (+0.0059)
0.022142932.5851-----
0.024146827.7726-----
0.026150724.6193-----
0.028154622.2331-----
0.030158520.815819.36420.6007 (+0.0603)0.3694 (+0.0444)0.5467 (+0.0460)0.5056 (+0.0502)
0.032162419.5039-----
0.034166318.151-----
0.036270217.1317-----
0.038274116.6977-----
0.040278015.674915.10130.6167 (+0.0763)0.3794 (+0.0544)0.5682 (+0.0675)0.5214 (+0.0661)
0.042281915.2219-----
0.044285814.7046-----
0.046289714.1086-----
0.048293613.5811-----
0.050297513.220212.56010.6616 (+0.1212)0.3763 (+0.0513)0.5839 (+0.0833)0.5406 (+0.0852)
0.0522101412.7417-----
0.0542105312.4341-----
0.0562109212.1748-----
0.0582113111.8355-----
0.0603117011.328811.05950.6582 (+0.1178)0.3847 (+0.0596)0.6125 (+0.1118)0.5518 (+0.0964)
0.0623120911.2907-----
0.0643124810.9028-----
0.0663128710.627-----
0.0683132610.4322-----
0.0703136510.08319.75910.6565 (+0.1161)0.3910 (+0.0659)0.6118 (+0.1111)0.5531 (+0.0977)
0.0723140410.0484-----
0.074314439.7691-----
0.076314829.5597-----
0.078315219.47-----
0.080315609.25288.99100.6667 (+0.1263)0.3815 (+0.0564)0.6247 (+0.1240)0.5576 (+0.1023)
0.082315999.1445-----
0.084416389.0069-----
0.086416778.6523-----
0.088417168.5901-----
0.090417558.7158.31660.6402 (+0.0997)0.3853 (+0.0603)0.6448 (+0.1442)0.5568 (+0.1014)
0.092417948.457-----
0.094418338.3209-----
0.096418728.1242-----
0.098419118.1453-----
0.100419508.05617.66340.6806 (+0.1401)0.3843 (+0.0592)0.6495 (+0.1489)0.5715 (+0.1161)
0.102419897.855-----
0.104420287.7573-----
0.106420677.5976-----
0.108521067.4529-----
0.110521457.3727.07980.6667 (+0.1263)0.3952 (+0.0702)0.6386 (+0.1379)0.5668 (+0.1115)
0.112521847.1673-----
0.114522237.2075-----
0.116522627.1261-----
0.118523017.0716-----
0.120523407.01826.75600.6557 (+0.1153)0.3876 (+0.0626)0.6544 (+0.1537)0.5659 (+0.1105)
0.122523796.861-----
0.124524186.8366-----
0.126524576.6857-----
0.128524966.665-----
0.130525356.65646.28240.6796 (+0.1392)0.3834 (+0.0584)0.6637 (+0.1630)0.5756 (+0.1202)
0.132625746.5436-----
0.134626136.4091-----
0.136626526.2974-----
0.138626916.3251-----
0.140627306.27995.98520.6719 (+0.1314)0.3829 (+0.0579)0.6711 (+0.1704)0.5753 (+0.1199)
0.142627696.2637-----
0.144628086.1648-----
0.146628476.1382-----
0.148628866.1291-----
0.150629255.94135.84020.6665 (+0.1261)0.3950 (+0.0700)0.6448 (+0.1441)0.5688 (+0.1134)
0.152629645.9748-----
0.154630035.8973-----
0.156730425.8654-----
0.158730815.6654-----
0.160731205.85495.69180.6765 (+0.1361)0.3948 (+0.0697)0.6639 (+0.1632)0.5784 (+0.1230)
0.162731595.6934-----
0.164731985.7285-----
0.166732375.5589-----
0.168732765.6317-----
0.170733155.57415.45210.6545 (+0.1141)0.3988 (+0.0738)0.6863 (+0.1856)0.5799 (+0.1245)
0.172733545.4948-----
0.174733935.4782-----
0.176734325.5595-----
0.178734715.417-----
0.180835105.43395.27300.6224 (+0.0820)0.3970 (+0.0720)0.6666 (+0.1660)0.5620 (+0.1066)
0.182835495.3723-----
0.184835885.2479-----
0.186836275.2665-----
0.188836665.2302-----
0.190837055.18635.14190.6791 (+0.1387)0.4037 (+0.0786)0.6941 (+0.1934)0.5923 (+0.1369)
0.192837445.1855-----
0.194837835.1529-----
0.196838225.2058-----
0.198838615.098-----
0.200839005.01764.98690.6590 (+0.1186)0.3849 (+0.0599)0.6844 (+0.1837)0.5761 (+0.1207)
0.202839395.0708-----
0.204939785.0215-----
0.206940174.974-----
0.208940564.9687-----
0.210940954.96894.85060.6734 (+0.1330)0.3995 (+0.0745)0.7023 (+0.2016)0.5917 (+0.1364)
0.212941344.8809-----
0.214941734.9176-----
0.216942124.7451-----
0.218942514.7807-----
0.220942904.81574.71500.6269 (+0.0865)0.3948 (+0.0698)0.6938 (+0.1932)0.5719 (+0.1165)
0.222943294.7986-----
0.224943684.7942-----
0.226944074.7008-----
0.229044464.7572-----
0.231044854.76164.66570.6577 (+0.1172)0.4022 (+0.0772)0.7019 (+0.2013)0.5873 (+0.1319)
0.233045244.7014-----
0.235045634.6512-----
0.237046024.6997-----
0.239046414.5655-----
0.241046804.57274.53670.6826 (+0.1422)0.3937 (+0.0687)0.6958 (+0.1951)0.5907 (+0.1353)
0.243047194.5258-----
0.245047584.6012-----
0.247047974.5785-----
0.249048364.5415-----
0.251048754.49214.44620.6689 (+0.1285)0.3977 (+0.0727)0.7103 (+0.2096)0.5923 (+0.1369)
0.253149144.4911-----
0.255149534.4795-----
0.257149924.4027-----
0.259150314.3652-----
0.261150704.38684.39090.6535 (+0.1130)0.3868 (+0.0617)0.6815 (+0.1809)0.5739 (+0.1185)
0.263151094.4055-----
0.265151484.3968-----
0.267151874.3333-----
0.269152264.3369-----
0.271152654.30794.33550.6550 (+0.1145)0.3895 (+0.0644)0.6957 (+0.1950)0.5800 (+0.1247)
0.273153044.3211-----
0.275153434.2841-----
0.277253824.2753-----
0.279254214.221-----
0.281254604.21464.20560.6590 (+0.1186)0.3796 (+0.0545)0.7028 (+0.2021)0.5804 (+0.1251)
0.283254994.2692-----
0.285255384.2236-----
0.287255774.1555-----
0.289256164.1684-----
0.291256554.17314.13040.6734 (+0.1329)0.3865 (+0.0615)0.6950 (+0.1944)0.5850 (+0.1296)
0.293256944.1562-----
0.295257334.1689-----
0.297257724.1617-----
0.299258114.1256-----
0.301358504.05924.07230.6694 (+0.1290)0.3870 (+0.0619)0.7079 (+0.2073)0.5881 (+0.1327)
0.303358894.0894-----
0.305359284.103-----
0.307359674.0083-----
0.309360064.03-----
0.311360453.99314.00580.6695 (+0.1290)0.3914 (+0.0664)0.7024 (+0.2018)0.5878 (+0.1324)
0.313360844.0186-----
0.315361233.9312-----
0.317361624.0398-----
0.319362013.9672-----
0.321362403.98793.93220.6696 (+0.1292)0.3932 (+0.0681)0.7075 (+0.2068)0.5901 (+0.1347)
0.323362793.879-----
0.325463183.9123-----
0.327463573.9144-----
0.329463963.903-----
0.331464353.94473.90700.6440 (+0.1036)0.3849 (+0.0599)0.7099 (+0.2092)0.5796 (+0.1242)
0.333464743.9082-----
0.335465133.8405-----
0.337465523.8633-----
0.339465913.8301-----
0.341466303.81883.85240.6611 (+0.1206)0.3804 (+0.0554)0.6935 (+0.1929)0.5783 (+0.1230)
0.343466693.8292-----
0.345467083.8502-----
0.347467473.8649-----
0.349567863.7942-----
0.351568253.73243.79580.6732 (+0.1328)0.3816 (+0.0565)0.7011 (+0.2005)0.5853 (+0.1299)
0.353568643.7916-----
0.355569033.7299-----
0.357569423.8087-----
0.359569813.7821-----
0.361570203.73953.74380.6585 (+0.1181)0.3875 (+0.0625)0.7004 (+0.1998)0.5821 (+0.1268)
0.363570593.7345-----
0.365570983.7472-----
0.367571373.7277-----
0.369571763.6535-----
0.371572153.65863.69820.6616 (+0.1212)0.3823 (+0.0573)0.7002 (+0.1996)0.5814 (+0.1260)
0.373672543.6653-----
0.375672933.7074-----
0.377673323.6542-----
0.379673713.5972-----
0.381674103.64993.62830.6569 (+0.1165)0.3796 (+0.0545)0.7079 (+0.2073)0.5815 (+0.1261)
0.383674493.6373-----
0.385674883.6253-----
0.387675273.6441-----
0.389675663.6278-----
0.391676053.62913.60080.6695 (+0.1291)0.3928 (+0.0677)0.7187 (+0.2180)0.5936 (+0.1383)
0.393676443.5957-----
0.395676833.6031-----
0.397777223.5544-----
0.399777613.5823-----
0.401778003.54263.60260.6732 (+0.1328)0.3851 (+0.0600)0.7170 (+0.2163)0.5918 (+0.1364)
0.403778393.5943-----
0.405778783.4955-----
0.407779173.5072-----
0.409779563.5529-----
0.411779953.55233.47590.6666 (+0.1262)0.3826 (+0.0576)0.7048 (+0.2041)0.5847 (+0.1293)
0.413780343.4947-----
0.415780733.4753-----
0.417781123.431-----
0.419781513.4871-----
0.421881903.50723.48410.6703 (+0.1299)0.3880 (+0.0630)0.7216 (+0.2210)0.5933 (+0.1380)
0.423882293.4812-----
0.425882683.4408-----
0.427883073.4781-----
0.429883463.4667-----
0.431883853.46183.45750.6709 (+0.1305)0.3785 (+0.0535)0.7062 (+0.2055)0.5852 (+0.1298)
0.433884243.4711-----
0.435884633.4586-----
0.437885023.4246-----
0.439885413.4322-----
0.441885803.39013.42890.6657 (+0.1253)0.3780 (+0.0530)0.7118 (+0.2111)0.5852 (+0.1298)
0.443886193.3776-----
0.445986583.3819-----
0.447986973.393-----
0.449987363.3559-----
0.451987753.40043.29810.6748 (+0.1344)0.3870 (+0.0620)0.7121 (+0.2114)0.5913 (+0.1359)
0.453988143.426-----
0.455988533.3938-----
0.457988923.3304-----
0.459989313.3455-----
0.461989703.33143.35310.6700 (+0.1296)0.3775 (+0.0524)0.7173 (+0.2166)0.5882 (+0.1329)
0.463990093.3455-----
0.465990483.3363-----
0.467990873.3492-----
0.470091263.3272-----
0.472091653.32533.28390.6963 (+0.1558)0.3831 (+0.0581)0.7113 (+0.2107)0.5969 (+0.1415)
0.474092043.3214-----
0.476092433.2383-----
0.478092823.3198-----
0.480093213.2898-----
0.482093603.26413.22870.6832 (+0.1428)0.3880 (+0.0630)0.7206 (+0.2200)0.5973 (+0.1419)
0.484093993.2974-----
0.486094383.3147-----
0.488094773.2911-----
0.490095163.2745-----
0.492095553.21683.27290.6730 (+0.1326)0.3920 (+0.0669)0.7173 (+0.2167)0.5941 (+0.1387)
0.494195943.2036-----
0.496196333.2618-----
0.498196723.2786-----
0.500197113.2214-----
0.502197503.17593.21060.6764 (+0.1360)0.3852 (+0.0602)0.7226 (+0.2220)0.5948 (+0.1394)
0.504197893.2159-----
0.506198283.1856-----
0.508198673.209-----
0.510199063.2239-----
0.512199453.21423.21590.6697 (+0.1292)0.3913 (+0.0662)0.7009 (+0.2003)0.5873 (+0.1319)
0.514199843.2156-----
0.5161100233.1568-----
0.5182100623.1344-----
0.5202101013.1754-----
0.5222101403.14143.20090.6913 (+0.1509)0.3913 (+0.0663)0.7180 (+0.2173)0.6002 (+0.1448)
0.5242101793.2359-----
0.5262102183.149-----
0.5282102573.1741-----
0.5302102963.081-----
0.5322103353.14573.19390.6685 (+0.1281)0.3858 (+0.0608)0.7304 (+0.2297)0.5949 (+0.1395)
0.5342103743.1322-----
0.5362104133.1307-----
0.5382104523.1212-----
0.5402104913.105-----
0.5423105303.13393.13240.6693 (+0.1289)0.3942 (+0.0692)0.7101 (+0.2094)0.5912 (+0.1358)
0.5443105693.111-----
0.5463106083.1045-----
0.5483106473.0923-----
0.5503106863.075-----
0.5523107253.0623.13630.6931 (+0.1527)0.4079 (+0.0828)0.7233 (+0.2227)0.6081 (+0.1527)
0.5543107643.1167-----
0.5563108033.0903-----
0.5583108423.0604-----
0.5603108813.0724-----
0.5623109203.10523.11810.6856 (+0.1452)0.4003 (+0.0753)0.7249 (+0.2242)0.6036 (+0.1482)
0.5643109593.0984-----
0.5664109983.0779-----
0.5684110372.9911-----
0.5704110763.0431-----
0.5724111153.07933.11060.6709 (+0.1305)0.3998 (+0.0748)0.7186 (+0.2180)0.5965 (+0.1411)
0.5744111543.0504-----
0.5764111933.0118-----
0.5784112323.0704-----
0.5804112713.0498-----
0.5824113103.03933.04750.6806 (+0.1401)0.4091 (+0.0841)0.7122 (+0.2116)0.6006 (+0.1453)
0.5844113492.9919-----
0.5864113883.0291-----
0.5884114273.0119-----
0.5905114663.0201-----
0.5925115053.03193.00640.6852 (+0.1448)0.4061 (+0.0811)0.7163 (+0.2156)0.6025 (+0.1472)
0.5945115443.0309-----
0.5965115833.0097-----
0.5985116222.9759-----
0.6005116612.9937-----
0.6025117002.98853.02620.6729 (+0.1325)0.3985 (+0.0735)0.7227 (+0.2221)0.5980 (+0.1427)
0.6045117393.0092-----
0.6065117782.9569-----
0.6085118172.9665-----
0.6105118562.9774-----
0.6125118952.99152.98420.6857 (+0.1452)0.4055 (+0.0804)0.7379 (+0.2372)0.6097 (+0.1543)
0.6146119342.9555-----
0.6166119732.9833-----
0.6186120122.9858-----
0.6206120512.9743-----
0.6226120902.96862.95230.6821 (+0.1417)0.4044 (+0.0794)0.7238 (+0.2232)0.6035 (+0.1481)
0.6246121292.9867-----
0.6266121682.9548-----
0.6286122072.9557-----
0.6306122462.9506-----
0.6326122852.96922.93230.6862 (+0.1458)0.4063 (+0.0813)0.7257 (+0.2251)0.6061 (+0.1507)
0.6346123242.9458-----
0.6366123632.9131-----
0.6387124022.9118-----
0.6407124412.8914-----
0.6427124802.88452.98320.6796 (+0.1392)0.4009 (+0.0759)0.7250 (+0.2244)0.6019 (+0.1465)
0.6447125192.9312-----
0.6467125582.9344-----
0.6487125972.9033-----
0.6507126362.892-----
0.6527126752.89442.92860.6874 (+0.1469)0.4015 (+0.0764)0.7247 (+0.2241)0.6045 (+0.1491)
0.6547127142.9269-----
0.6567127532.8988-----
0.6587127922.9167-----
0.6607128312.8703-----
0.6628128702.86192.87240.6829 (+0.1425)0.4005 (+0.0754)0.7176 (+0.2170)0.6003 (+0.1450)
0.6648129092.868-----
0.6668129482.8775-----
0.6688129872.866-----
0.6708130262.8877-----
0.6728130652.8962.91170.6846 (+0.1442)0.4023 (+0.0773)0.7198 (+0.2192)0.6022 (+0.1469)
0.6748131042.8351-----
0.6768131432.8679-----
0.6788131822.9197-----
0.6808132212.822-----
0.6828132602.84432.90100.6775 (+0.1371)0.4039 (+0.0789)0.7116 (+0.2109)0.5977 (+0.1423)
0.6848132992.8646-----
0.6869133382.8645-----
0.6889133772.8659-----
0.6909134162.8264-----
0.6929134552.82222.86320.6947 (+0.1543)0.4057 (+0.0807)0.7314 (+0.2307)0.6106 (+0.1552)
0.6949134942.8565-----
0.6969135332.8385-----
0.6989135722.8305-----
0.7009136112.8368-----
0.7029136502.84162.83090.6784 (+0.1380)0.4113 (+0.0863)0.7270 (+0.2264)0.6056 (+0.1502)
0.7049136892.8007-----
0.7069137282.8565-----
0.7089137672.8893-----
0.7110138062.844-----
0.7130138452.82932.84230.6693 (+0.1289)0.4053 (+0.0803)0.7244 (+0.2238)0.5997 (+0.1443)
0.7150138842.8424-----
0.7170139232.7951-----
0.7190139622.8004-----
0.7210140012.7833-----
0.7230140402.81332.80210.6788 (+0.1383)0.4013 (+0.0763)0.7285 (+0.2278)0.6028 (+0.1475)
0.7250140792.8245-----
0.7270141182.7995-----
0.7290141572.7859-----
0.7310141962.8067-----
0.7330142352.76062.80980.6750 (+0.1346)0.4012 (+0.0762)0.7381 (+0.2375)0.6048 (+0.1494)
0.7351142742.7662-----
0.7371143132.8081-----
0.7391143522.8159-----
0.7411143912.7604-----
0.7431144302.77212.79120.6759 (+0.1355)0.4040 (+0.0790)0.7283 (+0.2276)0.6027 (+0.1474)
0.7451144692.7772-----
0.7471145082.8183-----
0.7491145472.7821-----
0.7511145862.7434-----
0.7531146252.8122.79400.6673 (+0.1268)0.4034 (+0.0783)0.7277 (+0.2271)0.5995 (+0.1441)
0.7551146642.7847-----
0.7571147032.7604-----
0.7592147422.7271-----
0.7612147812.7663-----
0.7632148202.77312.74890.6694 (+0.1290)0.3963 (+0.0713)0.7380 (+0.2373)0.6012 (+0.1459)
0.7652148592.8013-----
0.7672148982.762-----
0.7692149372.7646-----
0.7712149762.762-----
0.7732150152.772.73670.6815 (+0.1411)0.4066 (+0.0816)0.7262 (+0.2256)0.6048 (+0.1494)
0.7752150542.7827-----
0.7772150932.7027-----
0.7792151322.7395-----
0.7812151712.7425-----
0.7833152102.77572.73800.6720 (+0.1315)0.4057 (+0.0807)0.7374 (+0.2367)0.6050 (+0.1497)
0.7853152492.7185-----
0.7873152882.7287-----
0.7893153272.7311-----
0.7913153662.7281-----
0.7933154052.70392.71420.6854 (+0.1450)0.4034 (+0.0784)0.7364 (+0.2358)0.6084 (+0.1531)
0.7953154442.7028-----
0.7973154832.7182-----
0.7993155222.688-----
0.8013155612.7562-----
0.8033156002.69682.74130.6669 (+0.1264)0.4118 (+0.0867)0.7269 (+0.2263)0.6018 (+0.1465)
0.8053156392.6802-----
0.8074156782.6936-----
0.8094157172.7059-----
0.8114157562.7052-----
0.8134157952.71782.69680.6738 (+0.1334)0.4064 (+0.0813)0.7283 (+0.2277)0.6028 (+0.1475)
0.8154158342.669-----
0.8174158732.6881-----
0.8194159122.6973-----
0.8214159512.6861-----
0.8234159902.66952.71100.6821 (+0.1417)0.4104 (+0.0853)0.7304 (+0.2298)0.6076 (+0.1523)
0.8254160292.6974-----
0.8274160682.7043-----
0.8294161072.6929-----
0.8315161462.667-----
0.8335161852.70352.67460.6891 (+0.1487)0.4084 (+0.0834)0.7262 (+0.2256)0.6079 (+0.1526)
0.8355162242.6472-----
0.8375162632.6956-----
0.8395163022.6823-----
0.8415163412.695-----
0.8435163802.64392.68410.6835 (+0.1430)0.4015 (+0.0764)0.7288 (+0.2282)0.6046 (+0.1492)
0.8455164192.6767-----
0.8475164582.6556-----
0.8495164972.6431-----
0.8515165362.6594-----
0.8535165752.64032.66570.6709 (+0.1304)0.4044 (+0.0794)0.7323 (+0.2316)0.6025 (+0.1471)
0.8556166142.6687-----
0.8576166532.6887-----
0.8596166922.6761-----
0.8616167312.6371-----
0.8636167702.63682.67910.6768 (+0.1364)0.4068 (+0.0818)0.7314 (+0.2308)0.6050 (+0.1497)
0.8656168092.6325-----
0.8676168482.641-----
0.8696168872.6614-----
0.8716169262.7-----
0.8736169652.6782.63540.6938 (+0.1533)0.4108 (+0.0857)0.7274 (+0.2268)0.6107 (+0.1553)
0.8756170042.6619-----
0.8776170432.681-----
0.8797170822.6465-----
0.8817171212.6977-----
0.8837171602.64762.63240.6969 (+0.1565)0.4118 (+0.0868)0.7293 (+0.2287)0.6127 (+0.1573)
0.8857171992.6405-----
0.8877172382.6645-----
0.8897172772.6635-----
0.8917173162.6038-----
0.8937173552.68072.6420.6951 (+0.1547)0.4138 (+0.0888)0.7294 (+0.2287)0.6128 (+0.1574)
0.8957173942.6679-----
0.8977174332.6063-----
0.8997174722.6215-----
0.9017175112.6141-----
0.9038175502.63242.63300.6963 (+0.1559)0.4117 (+0.0866)0.7287 (+0.2280)0.6122 (+0.1568)
0.9058175892.6405-----
0.9078176282.619-----
0.9098176672.6153-----
0.9118177062.6323-----
0.9138177452.58912.62800.6886 (+0.1482)0.4110 (+0.0860)0.7275 (+0.2268)0.6090 (+0.1537)
0.9158177842.6385-----
0.9178178232.6138-----
0.9198178622.6181-----
0.9218179012.6414-----
0.9238179402.63622.62240.6907 (+0.1502)0.4085 (+0.0835)0.7360 (+0.2354)0.6117 (+0.1564)
0.9258179792.6156-----
0.9279180182.597-----
0.9299180572.6254-----
0.9319180962.6434-----
0.9339181352.64742.60140.6884 (+0.1479)0.4133 (+0.0882)0.7248 (+0.2242)0.6088 (+0.1534)
0.9359181742.6214-----
0.9379182132.6145-----
0.9399182522.617-----
0.9419182912.6209-----
0.9439183302.59762.60570.6871 (+0.1467)0.4072 (+0.0821)0.7290 (+0.2284)0.6078 (+0.1524)
0.9459183692.61-----
0.9479184082.614-----
0.9499184472.6187-----
0.9520184862.6004-----
0.9540185252.65152.60520.6871 (+0.1467)0.4091 (+0.0841)0.7270 (+0.2263)0.6077 (+0.1524)
0.9560185642.6141-----
0.9580186032.6016-----
0.9600186422.5918-----
0.9620186812.5684-----
0.9640187202.60872.59410.6797 (+0.1393)0.4105 (+0.0854)0.7284 (+0.2277)0.6062 (+0.1508)
0.9660187592.5961-----
0.9680187982.6121-----
0.9700188372.5896-----
0.9720188762.6101-----
0.9740189152.61062.59210.6856 (+0.1452)0.4093 (+0.0843)0.7284 (+0.2277)0.6078 (+0.1524)
0.9761189542.6046-----
0.9781189932.6155-----
0.9801190322.6166-----
0.9821190712.5866-----
0.9841191102.63692.59430.6838 (+0.1433)0.4117 (+0.0867)0.7284 (+0.2277)0.6079 (+0.1526)
0.9861191492.6319-----
0.9881191882.6035-----
0.9901192272.565-----
0.9921192662.6071-----
0.9941193052.59082.58660.6838 (+0.1433)0.4116 (+0.0866)0.7284 (+0.2277)0.6079 (+0.1525)
0.9961193442.5983-----
0.9981193832.5995-----
-1-1--0.6951 (+0.1547)0.4138 (+0.0888)0.7294 (+0.2287)0.6128 (+0.1574)
  • The bold row denotes the saved checkpoint.

Environmental Impact

Carbon emissions were measured using CodeCarbon.

  • Energy Consumed: 6.535 kWh
  • Carbon Emitted: 2.412 kg of CO2
  • Hours Used: 1.932 hours

Training Hardware

  • On Cloud: No
  • GPU Model: 8 x NVIDIA H100 80GB HBM3
  • CPU Model: AMD EPYC 7R13 Processor
  • RAM Size: 1999.99 GB

Framework Versions

  • Python: 3.10.14
  • Sentence Transformers: 5.1.2
  • Transformers: 4.57.1
  • PyTorch: 2.9.1+cu126
  • Accelerate: 1.12.0
  • Datasets: 4.4.1
  • Tokenizers: 0.22.1

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",
}

MarginMSELoss

@misc{hofstätter2021improving,
    title={Improving Efficient Neural Ranking Models with Cross-Architecture Knowledge Distillation},
    author={Sebastian Hofstätter and Sophia Althammer and Michael Schröder and Mete Sertkan and Allan Hanbury},
    year={2021},
    eprint={2010.02666},
    archivePrefix={arXiv},
    primaryClass={cs.IR}
}

Как запустить

На своей GPU в ailist. Цена — по текущим предложениям провайдеров

GPU в ailist

Универсальный рантайм · sentence-transformers

Можно запустить

Подбираем GPU…

Класс архитектуры
ModernBertForSequenceClassification
Ревизия весов
06f5d0b

0 моделей