Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks
Paper
•
1908.10084
•
Published
•
12
This is a sentence-transformers model finetuned from BAAI/bge-large-zh-v1.5. It maps sentences & paragraphs to a 1024-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.
SentenceTransformer(
(0): Transformer({'max_seq_length': 256, 'do_lower_case': True, 'architecture': '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): Normalize()
)
First install the Sentence Transformers library:
pip install -U sentence-transformers
Then you can load this model and run inference.
from sentence_transformers import SentenceTransformer
# Download from the 🤗 Hub
model = SentenceTransformer("sentence_transformers_model_id")
# Run inference
sentences = [
'为这个句子生成表示以用于检索相关文章:血缘关系变更姓氏的材料是非必要的话,还需要提供复印件吗?',
'xjzw 填报须知:\n纸质复印件材 材料类型: dt/rest/atta\n料份数: 原件 ch/openAtt 要求填报的材\n0 材料形式: ach?client= 料依据:\n纸质和电子\n纸质材料规格:\n原件\n3 婚姻关系 纸质原件材料 材料必要性 https://zwf 来源渠道:\n证件(因 份数: 非必要 w.xinjiang.g 政府部门核发\n涉外婚姻 1 其他要求: ov.cn/xjzw 填报须知:\n关系变更 纸质复印件材 材料类型: dt/rest/atta\n姓氏的需 料份数: 原件 ch/openAtt 要求填报的材\n提交) 0 材料形式: ach?client= 料依据:\n纸质和电子\n纸质材料规格:\nA4\n4 关系凭证 纸质原件材料 材料必要性 https://zwf 来源渠道:\n(因血缘 份数: 非必要 w.xinjiang.g 政府部门核发\n关系在其 1 其他要求: ov.cn/xjzw 填报须知:\n直系长辈 纸质复印件材 材料类型: dt/rest/atta\n血亲之间 料份数: 原件 ch/openAtt 要求填报的材\n变更姓氏 0 材料形式: ach?client',
'百三十七条\n条款内容:第一百三十二条申请变更姓名,应当提交居民户口簿居民身份证,并按照第一百三十一规定提交书面申请,向户口所在地公安派出所申请。\n未满18周岁的须由亲生父母协商一致并同时到户口所在地公安派出所申请,已满8周岁的还应当征得本人签字同意;父母离婚后,双方未取得一\n致意见申请未满18岁子女姓名变更的,不予受理;父母一方死亡的,由另一方吃注销原因为死亡的《户口注销证明》,到户口所在地公安派出所办\n理。\n第一百三十七条因收养等关系变化或者重新确认,按照第七十五条规定申请变更。\n常见问题',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 1024]
# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.7892, 0.2984],
# [0.7892, 1.0000, 0.2375],
# [0.2984, 0.2375, 1.0000]])
dev_evalInformationRetrievalEvaluator| Metric | Value |
|---|---|
| cosine_accuracy@1 | 0.2727 |
| cosine_accuracy@3 | 0.3636 |
| cosine_accuracy@5 | 0.6364 |
| cosine_accuracy@10 | 0.7273 |
| cosine_precision@1 | 0.2727 |
| cosine_precision@3 | 0.1212 |
| cosine_precision@5 | 0.1273 |
| cosine_precision@10 | 0.0727 |
| cosine_recall@1 | 0.2727 |
| cosine_recall@3 | 0.3636 |
| cosine_recall@5 | 0.6364 |
| cosine_recall@10 | 0.7273 |
| cosine_ndcg@10 | 0.4643 |
| cosine_mrr@10 | 0.3841 |
| cosine_map@100 | 0.3906 |
sentence_0, sentence_1, sentence_2, and sentence_3| sentence_0 | sentence_1 | sentence_2 | sentence_3 | |
|---|---|---|---|---|
| type | string | string | string | string |
| details |
|
|
|
|
| sentence_0 | sentence_1 | sentence_2 | sentence_3 |
|---|---|---|---|
为这个句子生成表示以用于检索相关文章:办婚姻登记要花钱吗? |
书》; |
百三十七条 |
纸质 cf2c66b5-2 |
为这个句子生成表示以用于检索相关文章:边境管理区通行证怎么办理? |
边境管理区通行证(深圳、珠海经济特区除外)核 |
进行初步审 |
委托 通办范围 跨县 |
为这个句子生成表示以用于检索相关文章:怎么申请户口簿表证? |
.申请:申领人通过窗口或新疆政务服务网(微警务)申请,提交申请材料。 |
= 料依据: |
人 |
MultipleNegativesRankingLoss with these parameters:{
"scale": 20.0,
"similarity_fct": "cos_sim",
"gather_across_devices": false
}
eval_strategy: stepsper_device_train_batch_size: 1per_device_eval_batch_size: 1fp16: Truemulti_dataset_batch_sampler: round_robinoverwrite_output_dir: Falsedo_predict: Falseeval_strategy: stepsprediction_loss_only: Trueper_device_train_batch_size: 1per_device_eval_batch_size: 1per_gpu_train_batch_size: Noneper_gpu_eval_batch_size: Nonegradient_accumulation_steps: 1eval_accumulation_steps: Nonetorch_empty_cache_steps: Nonelearning_rate: 5e-05weight_decay: 0.0adam_beta1: 0.9adam_beta2: 0.999adam_epsilon: 1e-08max_grad_norm: 1num_train_epochs: 3max_steps: -1lr_scheduler_type: linearlr_scheduler_kwargs: {}warmup_ratio: 0.0warmup_steps: 0log_level: passivelog_level_replica: warninglog_on_each_node: Truelogging_nan_inf_filter: Truesave_safetensors: Truesave_on_each_node: Falsesave_only_model: Falserestore_callback_states_from_checkpoint: Falseno_cuda: Falseuse_cpu: Falseuse_mps_device: Falseseed: 42data_seed: Nonejit_mode_eval: Falsebf16: Falsefp16: Truefp16_opt_level: O1half_precision_backend: autobf16_full_eval: Falsefp16_full_eval: Falsetf32: Nonelocal_rank: 0ddp_backend: Nonetpu_num_cores: Nonetpu_metrics_debug: Falsedebug: []dataloader_drop_last: Falsedataloader_num_workers: 0dataloader_prefetch_factor: Nonepast_index: -1disable_tqdm: Falseremove_unused_columns: Truelabel_names: Noneload_best_model_at_end: Falseignore_data_skip: Falsefsdp: []fsdp_min_num_params: 0fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': False}fsdp_transformer_layer_cls_to_wrap: Noneaccelerator_config: {'split_batches': False, 'dispatch_batches': None, 'even_batches': True, 'use_seedable_sampler': True, 'non_blocking': False, 'gradient_accumulation_kwargs': None}parallelism_config: Nonedeepspeed: Nonelabel_smoothing_factor: 0.0optim: adamw_torch_fusedoptim_args: Noneadafactor: Falsegroup_by_length: Falselength_column_name: lengthproject: huggingfacetrackio_space_id: trackioddp_find_unused_parameters: Noneddp_bucket_cap_mb: Noneddp_broadcast_buffers: Falsedataloader_pin_memory: Truedataloader_persistent_workers: Falseskip_memory_metrics: Trueuse_legacy_prediction_loop: Falsepush_to_hub: Falseresume_from_checkpoint: Nonehub_model_id: Nonehub_strategy: every_savehub_private_repo: Nonehub_always_push: Falsehub_revision: Nonegradient_checkpointing: Falsegradient_checkpointing_kwargs: Noneinclude_inputs_for_metrics: Falseinclude_for_metrics: []eval_do_concat_batches: Truefp16_backend: autopush_to_hub_model_id: Nonepush_to_hub_organization: Nonemp_parameters: auto_find_batch_size: Falsefull_determinism: Falsetorchdynamo: Noneray_scope: lastddp_timeout: 1800torch_compile: Falsetorch_compile_backend: Nonetorch_compile_mode: Noneinclude_tokens_per_second: Falseinclude_num_input_tokens_seen: noneftune_noise_alpha: Noneoptim_target_modules: Nonebatch_eval_metrics: Falseeval_on_start: Falseuse_liger_kernel: Falseliger_kernel_config: Noneeval_use_gather_object: Falseaverage_tokens_across_devices: Trueprompts: Nonebatch_sampler: batch_samplermulti_dataset_batch_sampler: round_robinrouter_mapping: {}learning_rate_mapping: {}| Epoch | Step | dev_eval_cosine_ndcg@10 |
|---|---|---|
| 0.4948 | 48 | 0.4643 |
@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",
}
@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}
}
Base model
BAAI/bge-large-zh-v1.5