EmiliaLee/k_model

SentenceTransformer based on BAAI/bge-m3

This is a sentence-transformers model finetuned from BAAI/bge-m3. It maps sentences & paragraphs to a 1024-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, classification, clustering, and more.

Model Details

Model Description

• Model Type: Sentence Transformer
• Base model: BAAI/bge-m3
• Maximum Sequence Length: 8192 tokens
• Output Dimensionality: 1024 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': 'XLMRobertaModel'})
  (1): Pooling({'embedding_dimension': 1024, 'pooling_mode': 'cls', 'include_prompt': True})
  (2): Normalize({})
)

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
sentences = [
    'compact garage storage shed under 3 feet wide',
    'outdoor 2 ft. 9 in . w x 2 ft. d plastic vertical tool shed Sheds with its tall profile , this 22 cubic foot vertical shed is great for storing ladders , long-handled tools , garden accessories , and more when space is at a premium . the durable , all-weather resin construction provides water resistance and uv protection , while the multi-wall panels are engineere interiorwidth-sidetoside:27|interiorheight-toptobottom:69|compatibleshelfpartnumber : xa1182|windrating:65|overallwidth-sidetoside:33|countryoforigin : united states|interiordepth-fronttoback:20.25|levelofassembly : full assembly needed|countryoforigin-additionaldetails : made in usa|doorwidth-sidet',
    'garden plastic/metal storage shed Sheds easily add storage space and value to your property with an outdoor garden shed . the metal shed has distinct advantages over traditional wooden models that make it a much more attractive choice for purchase . a metal shed is more affordable and being pre-fabricated significantly cuts down on expens dsmetallic : steel|shedtype : storage shed|color : grey|construction : traditional|overallheight-toptobottom:75.5|countryoforigin-additionaldetails : made in china|warrantylength:1 year|levelofassembly : full assembly needed|overallwidth-sidetoside:126.75|overalldepth-fronttoback:109|overallproductw',
]
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, 0.5560, 0.5587],
#         [0.5560, 1.0000, 0.5424],
#         [0.5587, 0.5424, 1.0000]])

Training Details

Training Dataset

Unnamed Dataset

• Size: 284,573 training samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 100 samples:

  	anchor	positive	negative
  type	string	string	string
  modality	text	text	text
  details	min: 7 tokens mean: 10.2 tokens max: 16 tokens	min: 115 tokens mean: 173.42 tokens max: 196 tokens	min: 97 tokens mean: 168.43 tokens max: 205 tokens

• Samples:

  anchor	positive	negative
  modern 10 inch semi-flush mount light for entryway	bucholz 1 -light 10 '' semi flush mount Flush Mount Lighting this 1-light semi-flush mount has a sleek , streamlined shape that gives a modern look to your hallway , entryway , or over your dining room table . it has a metal frame and downrod , and an egg-shaped glass shade with an opening at the bottom for light to shine through . the downrod retracts and ro shadematerial : glass|dry , damporwetlocationlisted : dry|whatisdrydamporwetlocationlisted : this indicates whether the fixture is safe to use in dry locations , damp locations ( moist environments ) , or wet locations ( direct exposure to water ) .|estimatedtimetosetup:5|bodyheight-toptobottom:10|w	weatherford 1 - light 10 '' semi flush mount Flush Mount Lighting marrying modern minimalism and factory flair , this understated one-light flush mount brings a splash of industrial style as it shines a light down on your space . crafted of metal , this fixture features a round canopy and a conical shade with its upper portion left open to provide a peek at the so canopywidth-sidetoside:4.875|recommendedbulbshapecode : a19|warrantylength:1 year|canopydepth-fronttoback:0.75|powersource : hardwired|dssecondaryproductstyle : nautical|canopyheight-toptobottom:4.875|recommendedbulbshape : standard|producttype : flush mount|whatisullisted : the underwriters laborat
  sleek glass pendant light for dining room	bucholz 1 -light 10 '' semi flush mount Flush Mount Lighting this 1-light semi-flush mount has a sleek , streamlined shape that gives a modern look to your hallway , entryway , or over your dining room table . it has a metal frame and downrod , and an egg-shaped glass shade with an opening at the bottom for light to shine through . the downrod retracts and ro shadematerial : glass|dry , damporwetlocationlisted : dry|whatisdrydamporwetlocationlisted : this indicates whether the fixture is safe to use in dry locations , damp locations ( moist environments ) , or wet locations ( direct exposure to water ) .|estimatedtimetosetup:5|bodyheight-toptobottom:10|w	6 '' h glass sphere pendant shade ( screw on ) in smoke color : smoke|shape : sphere|overallproductweight:0.67|overalldepth-fronttoback:6|theme : no theme|attachmenttype : screw on|style : modern & contemporary|overallwidth-sidetoside:6|fittersize:3.25|producttype : pendant shade|primarymaterial : glass|overallheight-toptobottom:6| : no|fitterincluded :
  small metal frame flush mount light with glass shade	bucholz 1 -light 10 '' semi flush mount Flush Mount Lighting this 1-light semi-flush mount has a sleek , streamlined shape that gives a modern look to your hallway , entryway , or over your dining room table . it has a metal frame and downrod , and an egg-shaped glass shade with an opening at the bottom for light to shine through . the downrod retracts and ro shadematerial : glass|dry , damporwetlocationlisted : dry|whatisdrydamporwetlocationlisted : this indicates whether the fixture is safe to use in dry locations , damp locations ( moist environments ) , or wet locations ( direct exposure to water ) .|estimatedtimetosetup:5|bodyheight-toptobottom:10|w	3 - light semi flush mount Flush Mount Lighting|Commercial Ceiling Lighting it is perfect to transform your dining room or entryway with this charming transitional flush mount ceiling light . bodydepth-fronttoback:16.1|producttype : semi flush mount|recommendedbulbshape : candle|bodydepth-fronttoback:12.2|lightdirection : up|warrantylength:2 years|dssecondaryproductstyle : classic glam|dsprimaryproductstyle : glam|fixturedesign : unique/statement|craftsmanshiptype : handcrafted / hand-fo

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim",
      "gather_across_devices": false,
      "directions": [
          "query_to_doc"
      ],
      "partition_mode": "joint",
      "hardness_mode": null,
      "hardness_strength": 0.0
  }
  

Evaluation Dataset

Unnamed Dataset

• Size: 14,978 evaluation samples
• Columns: anchor, positive, and negative
• Approximate statistics based on the first 100 samples:

  	anchor	positive	negative
  type	string	string	string
  modality	text	text	text
  details	min: 8 tokens mean: 11.03 tokens max: 18 tokens	min: 110 tokens mean: 159.21 tokens max: 193 tokens	min: 90 tokens mean: 168.18 tokens max: 201 tokens

• Samples:

  anchor	positive	negative
  quirky rooster shelf sitters	4 piece rooster shelf sitter figurine set Decorative Objects add these colorful rooster sitters to shelves , windowsills , and more throughout your home . the set includes four hand-painted roosters with delightful floral accents . each cute sitter has dangling legs and plenty of charm . overallwidth-sidetoside:1.63|geometricshapes : no geometric & shapes|producttype : figurine|subject : animals|nature : no nature|spiritualreligious : no spiritual & religious|structuresbuildings : no structures & buildings|wordstext : no words & text|overallproductweight:0.25|quantity:4|entertainmen	dunellen farmhouse rooster 2 piece figurine set Decorative Objects be the early bird in grabbing these farmhouse silver metal and wood rooster sculpture by the head and complete your farmhouse set up now . this 2-piece set features two small and medium standing rooster sculptures with a gray finish on the body , and silver on the rest . this will look great as a ce overallwidth-sidetoside:13|color : brown , silver|purposefuldistressingtype : distressed metal|style : farmhouse / country|overallheight-toptobottom:5.1|fantasysci-fi : no fantasy & sci-fi|overallheight-toptobottom:16|nature : no nature|geometricshapes : no geometric & shapes|spiritualreligious : no
  decorative rooster figurines with floral accents	4 piece rooster shelf sitter figurine set Decorative Objects add these colorful rooster sitters to shelves , windowsills , and more throughout your home . the set includes four hand-painted roosters with delightful floral accents . each cute sitter has dangling legs and plenty of charm . overallwidth-sidetoside:1.63|geometricshapes : no geometric & shapes|producttype : figurine|subject : animals|nature : no nature|spiritualreligious : no spiritual & religious|structuresbuildings : no structures & buildings|wordstext : no words & text|overallproductweight:0.25|quantity:4|entertainmen	2 piece metal figurine set Decorative Objects the antique style of these two metal rooster decors will bring a fascinating appeal to any rustic , vintage , or antique accent table . with its iron construction , these rooster sculptures will continue to enhance the look of your home interior for years to come . featuring a decorative rustic bron geometricshapes : no geometric & shapes|spiritualreligious : no spiritual & religious|overallwidth-sidetoside:20.47|producttype : figurine|nature : no nature|foodbeverage : no food & beverage|fantasysci-fi : no fantasy & sci-fi|people : no people|agegroup : adult|animals : rooster|overallproductweig
  comfortable anti-fatigue mat for kitchen counter	catrisha soothing foam comfort anti-fatigue mat Kitchen Mats this anti-fatigue mat provides a full inch of soothing foam comfort for your tired feet . perfect in front of a kitchen counter and has a skid-resistant backing . color : red|pattern : solid color|backingmaterial : polyester|color : brown|shape : rectangle|supplierintendedandapproveduse : non residential use|productcare : machine washable|construction : machine made|color : blue|dssecondaryproductstyle : transitional modern|overalllength-endtoend:30|color : g	grano hello standing anti-fatigue mat Kitchen Mats use this mat in any part of your home . whether it be the kitchen , laundry room , bathroom , or doorway this elegant mat will bring luscious texture , enduring material , and comfortable steps for your convenience . this perfectly crafted pvc mat is joined with a rubber backing for long-lasting usa pattern : no pattern and not solid color|overalllength-endtoend:30|overalllength-endtoend:47|overallproductweight:2.09|holidayoccasion : no holiday|dssecondaryproductstyle : modern farmhouse|overallheight-toptobottom:0.47|material : synthetics|theme : text|materialdetails:100 % pvc|overallwidth-side

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim",
      "gather_across_devices": false,
      "directions": [
          "query_to_doc"
      ],
      "partition_mode": "joint",
      "hardness_mode": null,
      "hardness_strength": 0.0
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 64
• num_train_epochs: 1
• learning_rate: 2e-05
• warmup_steps: 0.1
• bf16: True
• per_device_eval_batch_size: 64
• dataloader_num_workers: 4

All Hyperparameters

Click to expand

• per_device_train_batch_size: 64
• num_train_epochs: 1
• max_steps: -1
• learning_rate: 2e-05
• lr_scheduler_type: linear
• lr_scheduler_kwargs: None
• warmup_steps: 0.1
• optim: adamw_torch_fused
• optim_args: None
• weight_decay: 0.0
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• optim_target_modules: None
• gradient_accumulation_steps: 1
• average_tokens_across_devices: True
• max_grad_norm: 1.0
• label_smoothing_factor: 0.0
• bf16: True
• fp16: False
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: None
• gradient_checkpointing: False
• gradient_checkpointing_kwargs: None
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• use_liger_kernel: False
• liger_kernel_config: None
• use_cache: False
• neftune_noise_alpha: None
• torch_empty_cache_steps: None
• auto_find_batch_size: False
• log_on_each_node: True
• logging_nan_inf_filter: True
• include_num_input_tokens_seen: no
• log_level: passive
• log_level_replica: warning
• disable_tqdm: False
• project: huggingface
• trackio_space_id: None
• trackio_bucket_id: None
• trackio_static_space_id: None
• per_device_eval_batch_size: 64
• prediction_loss_only: True
• eval_on_start: False
• eval_do_concat_batches: True
• eval_use_gather_object: False
• eval_accumulation_steps: None
• include_for_metrics: []
• batch_eval_metrics: False
• save_only_model: False
• save_on_each_node: False
• enable_jit_checkpoint: False
• push_to_hub: False
• hub_private_repo: None
• hub_model_id: None
• hub_strategy: every_save
• hub_always_push: False
• hub_revision: None
• load_best_model_at_end: False
• ignore_data_skip: False
• restore_callback_states_from_checkpoint: False
• full_determinism: False
• seed: 42
• data_seed: None
• use_cpu: False
• 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
• dataloader_drop_last: False
• dataloader_num_workers: 4
• dataloader_pin_memory: True
• dataloader_persistent_workers: False
• dataloader_prefetch_factor: None
• remove_unused_columns: True
• label_names: None
• train_sampling_strategy: random
• length_column_name: length
• ddp_find_unused_parameters: None
• ddp_bucket_cap_mb: None
• ddp_broadcast_buffers: False
• ddp_static_graph: None
• ddp_backend: None
• ddp_timeout: 1800
• fsdp: None
• fsdp_config: None
• deepspeed: None
• debug: []
• skip_memory_metrics: True
• do_predict: False
• resume_from_checkpoint: None
• warmup_ratio: None
• local_rank: -1
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	Validation Loss
0.0225	100	1.1596	-
0.0450	200	0.6243	-
0.0675	300	0.5484	-
0.0899	400	0.4963	-
0.1124	500	0.4665	0.4796
0.1349	600	0.4426	-
0.1574	700	0.4353	-
0.1799	800	0.4284	-
0.2024	900	0.4111	-
0.2249	1000	0.3946	0.4274
0.2474	1100	0.3883	-
0.2698	1200	0.3806	-
0.2923	1300	0.3590	-
0.3148	1400	0.3661	-
0.3373	1500	0.3632	0.4086
0.3598	1600	0.3717	-
0.3823	1700	0.3517	-
0.4048	1800	0.3428	-
0.4273	1900	0.3574	-
0.4497	2000	0.3412	0.3882
0.4722	2100	0.3340	-
0.4947	2200	0.3356	-
0.5172	2300	0.3261	-
0.5397	2400	0.3358	-
0.5622	2500	0.3208	0.3833
0.5847	2600	0.3206	-
0.6072	2700	0.3254	-
0.6296	2800	0.3215	-
0.6521	2900	0.3082	-
0.6746	3000	0.3099	0.3684
0.6971	3100	0.3060	-
0.7196	3200	0.3009	-
0.7421	3300	0.3021	-
0.7646	3400	0.3163	-
0.7870	3500	0.3020	0.3652
0.8095	3600	0.3012	-
0.8320	3700	0.2927	-
0.8545	3800	0.2878	-
0.8770	3900	0.2974	-
0.8995	4000	0.3125	0.3584
0.9220	4100	0.2938	-
0.9445	4200	0.2997	-
0.9669	4300	0.2915	-
0.9894	4400	0.2923	-
1.0	4447	-	0.3572

Training Time

• Training: 26.3 minutes
• Evaluation: 3.2 minutes
• Total: 29.5 minutes

Framework Versions

• Python: 3.12.13
• Sentence Transformers: 5.7.0
• Transformers: 5.14.1
• PyTorch: 2.13.0+cu130
• Accelerate: 1.14.0
• Datasets: 5.0.1
• Tokenizers: 0.22.2

Additional Resources

• Training and Finetuning Embedding Models with Sentence Transformers: the end-to-end guide for training or finetuning Sentence Transformer models.
• Introduction to Matryoshka Embedding Models: variable-size embeddings that can be truncated with minimal quality loss.
• Binary and Scalar Embedding Quantization for Significantly Faster & Cheaper Retrieval: post-training compression of embedding vectors.
• Multimodal Embedding & Reranker Models with Sentence Transformers: use text, image, audio, and video models through the same API.
• Training and Finetuning Multimodal Embedding & Reranker Models with Sentence Transformers: train multimodal embedding models, with a Visual Document Retrieval walkthrough.

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",
}

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},
}