steerrec/yuan-embedding-2.0-zh-query-note

SentenceTransformer based on IEITYuan/Yuan-embedding-2.0-zh

This is a sentence-transformers model finetuned from IEITYuan/Yuan-embedding-2.0-zh. It maps sentences & paragraphs to a 1792-dimensional dense vector space and can be used for retrieval.

Model Details

Model Description

• Model Type: Sentence Transformer
• Base model: IEITYuan/Yuan-embedding-2.0-zh
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 1792 dimensions
• Similarity Function: Cosine Similarity
• Supported Modality: Text

Model Sources

• Documentation: Sentence Transformers Documentation
• Repository: Sentence Transformers on GitHub
• Hugging Face: Sentence Transformers on Hugging Face

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'transformer_task': 'feature-extraction', 'modality_config': {'text': {'method': 'forward', 'method_output_name': 'last_hidden_state'}}, 'module_output_name': 'token_embeddings', 'architecture': 'BertModel'})
  (1): Pooling({'embedding_dimension': 1024, 'pooling_mode': 'mean', 'include_prompt': True})
  (2): Dense({'in_features': 1024, 'out_features': 1792, 'bias': True, 'activation_function': 'torch.nn.modules.linear.Identity', 'module_input_name': 'sentence_embedding', 'module_output_name': 'sentence_embedding'})
)

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 SentenceTransformer

# Download from the 🤗 Hub
model = SentenceTransformer("sentence_transformers_model_id")
# Run inference
queries = [
    '上学文具分享',
]
documents = [
    '准初三生的书包里有啥😉👉🏻💗\n都是一些很真实的东西哈哈哈 \n我就问有谁懂…？\n#笔袋[话题]# #笔袋介绍[话题]# #我的文具分享[话题]# \n#晒晒我的书桌[话题]# #我的日常[话题]# \n#whatsinmybag[话题]# #书包里面装什么[话题]# \n#书包[话题]# ',
    '下辈子我也要当仓鼠\n傻傻的胖胖的不知道悲伤……#珍藏的宠物照[话题]# #侏儒仓鼠[话题]# #鼠鼠教[话题]# #宠物[话题]# #宠物日常[话题]# ',
    '云南丽江～ 束河古镇风景（上）\n拍于2023年11.19哦～\n从白沙坐公交去的束河古镇\n进门的时候忘记拍牌楼了[笑哭R]\n刚进门不久走到了茶马古道博物馆但是是关闭状态的\n不走人多的地方风景很不错\n图上那只松鼠的表情真的笑死\n看起来比香格里拉和雨崩的可爱\n建议沿着水流走 风景不会太差\n#云南[话题]# #云南游[话题]# #云南丽江[话题]# #束河古镇[话题]# ',
]
query_embeddings = model.encode_query(queries)
document_embeddings = model.encode_document(documents)
print(query_embeddings.shape, document_embeddings.shape)
# [1, 1792] [3, 1792]

# Get the similarity scores for the embeddings
similarities = model.similarity(query_embeddings, document_embeddings)
print(similarities)
# tensor([[ 0.6374,  0.0472, -0.0356]])

Evaluation

Metrics

Information Retrieval

• Datasets: query_note_test and query_note_val
• Evaluated with InformationRetrievalEvaluator

Metric	query_note_test	query_note_val
cosine_accuracy@1	0.2839	0.3002
cosine_accuracy@3	0.5495	0.5971
cosine_accuracy@5	0.6792	0.7274
cosine_accuracy@10	0.8055	0.8523
cosine_precision@1	0.2839	0.3002
cosine_precision@3	0.2471	0.2702
cosine_precision@5	0.2212	0.2335
cosine_precision@10	0.1722	0.1784
cosine_recall@1	0.111	0.1289
cosine_recall@3	0.2748	0.3301
cosine_recall@5	0.3833	0.4495
cosine_recall@10	0.5335	0.6154
cosine_ndcg@10	0.4031	0.4517
cosine_ndcg@100	0.5231	0.5634
cosine_mrr@10	0.4463	0.475
cosine_mrr@100	0.4544	0.4818
cosine_map@10	0.2898	0.334
cosine_map@100	0.3372	0.3809

Information Retrieval

• Datasets: query_note_test_256 and query_note_val_256
• Evaluated with InformationRetrievalEvaluator with these parameters:

  {
      "truncate_dim": 256
  }
  

Metric	query_note_test_256	query_note_val_256
cosine_accuracy@1	0.2763	0.2939
cosine_accuracy@3	0.5444	0.5918
cosine_accuracy@5	0.6674	0.7245
cosine_accuracy@10	0.8007	0.8479
cosine_precision@1	0.2763	0.2939
cosine_precision@3	0.2432	0.2667
cosine_precision@5	0.2151	0.2339
cosine_precision@10	0.1695	0.1766
cosine_recall@1	0.1077	0.1255
cosine_recall@3	0.2702	0.3267
cosine_recall@5	0.3715	0.4489
cosine_recall@10	0.5229	0.607
cosine_ndcg@10	0.3951	0.4467
cosine_ndcg@100	0.5137	0.5581
cosine_mrr@10	0.4397	0.4712
cosine_mrr@100	0.4477	0.4781
cosine_map@10	0.2826	0.3295
cosine_map@100	0.3287	0.376

Training Details

Training Dataset

Unnamed Dataset

• Size: 164,633 training samples
• Columns: anchor and positive
• Approximate statistics based on the first 100 samples:

  	anchor	positive
  type	string	string
  modality	text	text
  details	min: 3 tokens mean: 8.78 tokens max: 17 tokens	min: 38 tokens mean: 308.37 tokens max: 512 tokens

• Samples:

  anchor	positive
  jeep冲锋衣属于档次	🧥吉普冲锋衣，秋季必备🍂 Jeep吉普官方正品，品质有保障👍。 卡其色/M码，适合90-110斤，款式时尚百搭。 防风防水设计，适合户外活动，保暖又舒适😊。 秋季穿搭必备，你值得拥有！ #小红书秋日焕新[话题]# #jeep[话题]# #冲锋衣[话题]# #秋季外套[话题]# #正品保证[话题]# #防风防水[话题]# #户外穿搭[话题]# #时尚百搭[话题]# #品质选择[话题]#
  喝酒文案微醺	秋冬时刻需要威士忌 | 见面吗？！ 🍽️ 属于我和先生的浪漫约会日常 便是放下孩子后的小酌怡情 刚好最近气温骤降，多了些凉意 想来一杯Whisky的心情达到上限 🍽️ 听闻格兰菲迪最近和MU·慕有出合作套餐 与先生的couple vacation时刻瞬间启动 在轻奢环境下更适合开一瓶格兰菲迪22年单一麦芽苏格兰威士忌 @格兰菲迪 Glenfiddich 作为在全球屡获殊荣的单一麦芽苏格兰威士忌品牌 自诞生起就从未停止探索的脚步 不断探究威士忌与餐食搭配的新风味 🍽️ 当杯中深红琥珀色的酒体 在灯光下衬得更加迷人时 黑巧克力以及葡萄果干的浓郁香气也同时散发 象拔蚌和波士顿龙虾的上场 将视觉的色与口感的香瞬间突出 薄如刀片的爽脆蚌肉和Q弹龙虾肉 在酒精的碰撞下太立体太美好 余韵的温润柑橘果香和馥郁水果蛋糕香 像是激情生活之后的源远流长 搭配新西兰羊排嫩多汁的奶香气息 像是被酒精注入新的灵感和层次 🤎 婚后的微醺节奏 同样能拉近彼此的距离 大家也可以打卡格兰菲迪合作餐厅哦 用一杯格兰菲迪杯找寻浪漫~找寻生活吧 #广州探店[话题]# #福鹿结伴山海寻鲜[话题]# #威士忌推荐[话题]# #格兰菲迪[话题]# #格兰菲迪单一麦芽威士忌[话题]#
  长沙酒吧	长沙一定要去的6️⃣大夜店！！！ 长沙夜生活 蹦迪💃酒吧 再多的攻略都不及去亲身体验一次！！ ✅长沙一定要去的夜店❗️❗️ 1️⃣X-sta：长沙最火爆的高空edm场 声光电体验感强 非常具有科技感 总投资1.8亿[赞R] 四五六楼是包厢 娱乐综合体 2️⃣one9：小厅bounce场 韩国江南风夜店 前身是火爆长沙的NBEK 喜欢小厅的朋友不要错过啦[派对R] 3️⃣FTF：新开业中型edm/bounce酒吧 位置在海信广场 以前的超级猴子[派对R] 4️⃣kok：网红爆品小厅 每晚都有男团演出与玩家们互动 人气也一直居高不下[自拍R] 5️⃣猩猩地堡：连锁品牌 打卡的人很多 长沙年轻人的聚集地 6️⃣试音：长沙hiphop吧天花板 嘉宾众多气氛火爆 值得打卡 🌿住宿推荐 建议住五一广场/开福区万达广场（去哪都方便） ⚠️温馨提示：想去长沙夜店玩的朋友记得提前做好攻略，提前找销售预约好门店避免排队！ 希望这些对您有所帮助！ 图片来源于网络 侵删！ #长沙酒吧[话题]# #长沙酒吧推荐[话题]# #长沙酒吧订台[话题]# #长沙旅游[话题]# #长沙蹦迪[话题]#

• Loss: MatryoshkaLoss with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          1024,
          256
      ],
      "matryoshka_weights": [
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 256
• learning_rate: 0.0001
• weight_decay: 0.01
• num_train_epochs: 1
• lr_scheduler_type: cosine
• warmup_ratio: 0.05
• seed: 3407
• bf16: True
• batch_sampler: no_duplicates

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• prediction_loss_only: True
• per_device_train_batch_size: 256
• per_device_eval_batch_size: 8
• 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: 0.0001
• weight_decay: 0.01
• 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: cosine
• lr_scheduler_kwargs: {}
• warmup_ratio: 0.05
• 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: 3407
• 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}
• 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
• 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: False
• 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: False
• prompts: None
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	query_note_test_cosine_ndcg@100	query_note_test_256_cosine_ndcg@100	query_note_val_cosine_ndcg@100	query_note_val_256_cosine_ndcg@100
-1	-1	-	0.4211	0.4014	-	-
0.0776	50	1.8314	-	-	-	-
0.1553	100	0.7712	-	-	-	-
0.2329	150	0.7237	-	-	-	-
0.3106	200	0.7116	-	-	-	-
0.3882	250	0.6376	-	-	-	-
0.4658	300	0.6284	-	-	-	-
0.5435	350	0.6425	-	-	-	-
0.6211	400	0.6198	-	-	-	-
0.6988	450	0.6409	-	-	-	-
0.7764	500	0.5954	-	-	-	-
0.8540	550	0.5891	-	-	-	-
0.9317	600	0.5833	-	-	-	-
1.0	644	-	-	-	0.5634	0.5581
-1	-1	-	0.5231	0.5137	-	-

Training Time

• Training: 54.3 minutes

Framework Versions

• Python: 3.12.3
• Sentence Transformers: 5.5.1
• Transformers: 4.56.2
• PyTorch: 2.11.0+cu128
• Accelerate: 1.13.0
• Datasets: 4.3.0
• Tokenizers: 0.22.2

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",
}

MatryoshkaLoss

@misc{kusupati2024matryoshka,
    title={Matryoshka Representation Learning},
    author={Aditya Kusupati and Gantavya Bhatt and Aniket Rege and Matthew Wallingford and Aditya Sinha and Vivek Ramanujan and William Howard-Snyder and Kaifeng Chen and Sham Kakade and Prateek Jain and Ali Farhadi},
    year={2024},
    eprint={2205.13147},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

MultipleNegativesRankingLoss

@misc{oord2019representationlearningcontrastivepredictive,
      title={Representation Learning with Contrastive Predictive Coding},
      author={Aaron van den Oord and Yazhe Li and Oriol Vinyals},
      year={2019},
      eprint={1807.03748},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/1807.03748},
}