SentenceTransformer based on sentence-transformers/all-MiniLM-L6-v2

This is a sentence-transformers model finetuned from sentence-transformers/all-MiniLM-L6-v2. It maps sentences & paragraphs to a 384-dimensional dense vector space and can be used for semantic textual similarity, semantic search, paraphrase mining, text classification, clustering, and more.

Model Details

Model Description

  • Model Type: Sentence Transformer
  • Base model: sentence-transformers/all-MiniLM-L6-v2
  • Maximum Sequence Length: 256 tokens
  • Output Dimensionality: 384 dimensions
  • Similarity Function: Cosine Similarity

Model Sources

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 256, 'do_lower_case': False, 'architecture': 'BertModel'})
  (1): Pooling({'word_embedding_dimension': 384, 'pooling_mode_cls_token': False, 'pooling_mode_mean_tokens': True, '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()
)

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("zhenqingli/all-MiniLM-L6-v2-ft-home-depot-triplet")
# Run inference
sentences = [
    'cabinet locks 1/2',
    '{"product_title": "First Watch Security Polished Brass Keyed Alike Cabinet and Drawer Lock", "product_description": "First Watch Security has been designing, manufacturing, and delivering home security products since 1959. Today, First Watch Security is one of the most recognized brands in the market. Our Brand is found on products that include home, office, and door security offerings. Throughout the home and office there are drawers, cabinets, doors, and mailboxes that require privacy and protection.Max cabinet or drawer thickness 3/4 in.Polished brass finishVarious materialsKeyed alikeRequires 7/8 in. mounting hole", "Builders Hardware Product Type": "Accessory", "Bullet01": "Max cabinet or drawer thickness 3/4 in.", "Bullet02": "Polished brass finish", "Bullet03": "Various materials", "Bullet04": "Keyed alike", "Bullet05": "Requires 7/8 in. mounting hole", "MFG Brand Name": "First Watch Security", "Product Depth (in.)": "1.04", "Product Height (in.)": "1.7", "Product Width (in.)": "1.96", "Safety & Security Product Type": "Lock & Door Reinforcer"}',
    '{"product_title": "Prime-Line 1-1/8 in. Brass Drawer and Cabinet Keyed Cam Lock", "product_description": "This Defender Security 1-1/8 in. Brass Drawer and Cabinet Keyed Cam Lock features die-cast constructed housing and steel cams. It can be mounted to a metal or wood door or drawer 13/16 in. thick. It includes 2 keys.Keyed locking for cabinet or drawer securityPlated die-cast housingSteel camsBrass finishCan be mounted on a 13/16 in. thick wood or metal door or drawerIncludes 1 lock, 2 cams, 1 trim collar, 1 spur washer, 1 rotation washer, 1 lock nut and 2 keys", "Builders Hardware Product Type": "Other Cabinet Hardware", "Bullet01": "Keyed locking for cabinet or drawer security", "Bullet02": "Plated die-cast housing", "Bullet03": "Steel cams", "Bullet04": "Brass finish", "Bullet05": "Can be mounted on a 1/2 in. to 1 in. thick wood or metal door or drawer", "Bullet06": "Includes 1 lock, 2 cams, 1 trim collar, 1 spur washer, 1 rotation washer, 1 lock nut and 2 keys", "Finish": "Brass", "Hardware Finish Family": "Brass", "Material": "Metal", "MFG Brand Name": "Prime-Line", "Package Quantity": "1", "Product Depth (in.)": "1.125", "Product Height (in.)": "5.438", "Product Width (in.)": "3.813"}',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.6769, 0.5457],
#         [0.6769, 1.0000, 0.6671],
#         [0.5457, 0.6671, 1.0000]])

Evaluation

Metrics

Triplet

Metric test eval
cosine_accuracy 0.9371 0.947

Training Details

Training Dataset

Unnamed Dataset

  • Size: 9,328 training samples
  • Columns: anchor, positive, and negative
  • Approximate statistics based on the first 1000 samples:
    anchor positive negative
    type string string string
    details
    • min: 3 tokens
    • mean: 6.17 tokens
    • max: 17 tokens
    • min: 222 tokens
    • mean: 255.85 tokens
    • max: 256 tokens
    • min: 228 tokens
    • mean: 255.92 tokens
    • max: 256 tokens
  • Samples:
    anchor positive negative
    range hoodu {"product_title": "NuTone Allure I Series 30 in. Convertible Range Hood in Stainless Steel", "product_description": "The NuTone Allure I line of range hoods feature quiet blowers, exceptional ventilation performance and bright lighting of work surfaces. They are 50% quieter than the average range hood. The design features a two speed contoured rocker switch control, smooth surfaces, a fully enclosed underside and 2 large aluminum dishwasher-safe filters that provide the easiest clean-up of any range hood. The dual halogen light system (bulbs not included) is designed to flood your cooking surface with a bright light, the low light setting provides the perfect night light. The axial blade system eliminates smoke and moisture more quietly and effectively than the average hood or over-the-range microwave. Alluresimply performs better than any other hood.Convertible range hood can be installed as 3-1/4 in. x 10 in. (horizontal and vertical) ducted, 7 in. round (vertical) ducted and non-duc... {"product_title": "180 CFM Through-the-Wall Exhaust Fan with On/Off Switch", "product_description": "The Broan Direct Discharge 180 CFM Utility Through-Wall Fan can be used in rooms up to 160 sq. ft. for effective ventilation. The 180 CFM fan helps improve airflow with a powerful 6-1/2 sones motor. The fan is designed with an integrated on/off rotary switch to prevent the need for wall switch wiring. Installation is easy when you combine the 8 in. through-wall design with walls 4-1/2 in. to 9-1/2 in. thick.180 CFM helps improve ventilationMotor features 6-1/2 sones for efficiencyFor use in rooms up to 160 sq. ft. for ideal effectivenessOn/off rotary switch is integrated to prevent the need for wall switch wiring8 in. through-wall design can be installed in walls 4-1/2 in. to 9-1/2 in. thickDirect dischargeHome Depot Protection Plan:", "Air Volume (CFM)": "180", "Bullet01": "180 CFM helps improve ventilation", "Bullet02": "Motor features 6-1/2 sones for efficiency", "Bullet03": "For use...
    navy and brown floor lamp {"product_title": "ORE International 60 in. Dark Brown Floor Lamp", "product_description": "This well made floor lamp adds richness to any home decor. The simplistic features of this lamp bring out a clean and fresh sensation. The dark brown is stylized to compliment any room you place it in.Body is made of wood and steel, with a painted dark wooden finishCan be cleaned with a damp clothUses a 3-way 100 Watt type A bulbStylized linen shade compliments dark wood finished of base", "Adjustable Lamp Head": "No", "Base Color/Finish Family": "Wood", "Bulb Type Included": "Incandescent,No Bulbs Included", "Bullet01": "Body is made of wood and steel, with a painted dark wooden finish", "Bullet02": "Can be cleaned with a damp cloth", "Bullet03": "Uses a 3-way 100 Watt type A bulb", "Bullet04": "Stylized linen shade compliments dark wood finished of base", "California Title 20 Compliant": "No", "Certifications and Listings": "1-UL Listed", "Collection Name": "N/A", "Connection Type": "Plug-in",... {"product_title": "Elegant Lighting Sydney 63 in. Mocha Brown Floor Lamp with Golden Teak Smoky Crystal", "product_description": "Crystal bars in clear or rich and vibrant hues of gold or silver exploding with prisms layered in circular or rectangular shapes. The Sydney will satisfy your desire to hold the past and embrace the future at the same time.Industrial styleHigh quality materials used, built to last and offers clean stylish lookUL and cUL listed6 ft. cord included", "Adjustable Lamp Head": "No", "Base Color/Finish Family": "Brown", "Bulb Type Included": "Incandescent,No Bulbs Included", "Bullet01": "Industrial style", "Bullet02": "High quality materials used, built to last and offers clean stylish look", "Bullet03": "UL and cUL listed", "Bullet04": "6 ft. cord included", "California Title 20 Compliant": "No", "Certifications and Listings": "1-UL Listed,cUL Listed", "Connection Type": "Plug-in", "ENERGY STAR Certified": "No", "Fixture Color/Finish": "Mocha Brown", "Harp Style":...
    martha stewart bathroom vanity set {"product_title": "Martha Stewart Living Seal Harbor 30 in. Vanity in Sharkey Gray with Vanity Top in Alpine White", "product_description": "Cottage styled with "X" detailing and a furniture door design, the Martha Stewart Living Seal Harbor Vanity in Sharkey Gray with Vanity Top in Alpine White provides plenty of storage with three full-extension side drawers, a spacious interior cabinet with a pullout drawer and an open bottom shelf. Featuring a premium and easy-to-maintain ABS composite vanity top that is pre-drilled for an 8 in. widespread faucet, the vanity top also includes an integral sink with a stylish flat top edge. Durably made of solid wood and laminated engineered wood in a soothing gray finish, this vanity is ideal for a wide range of bathroom styles and includes polished chrome hardware for a sleek look. Pair with the matching Seal Harbor collection mirror for added functionality and polish.30.4 in. W x 21.2 in. D is ideal for small to medium sized bath or powder rooms... {"product_title": "Martha Stewart Living Wayland Collection 3-Light Brushed Nickel Plated Vanity Light", "product_description": "Update your light fixture with the Martha Stewart Living Wayland Collection 3-Light Brushed Nickel Wall Vanity with Clear Reeded Glass Shades. This wall vanity fixture features clear reeded glass shades for an attractive appearance.Hand-polished cast fittingsBrushed nickel-plated finishClear, reeded glass shadesUses three 60-Watt A19 bulbs (not included)Fixture can be installed with glass shades in an up or down positionHardware included for easy installation", "Bulb Type Included": "No Bulbs Included", "Bullet01": "Hand-polished cast fittings", "Bullet02": "Brushed nickel-plated finish", "Bullet03": "Clear, reeded glass shades", "Bullet04": "Uses three 60-Watt A19 bulbs (not included)", "Bullet05": "Fixture can be installed with glass shades in an up or down position", "Bullet06": "Hardware included for easy installation", "Certifications and Listings": "ETL...
  • Loss: MultipleNegativesRankingLoss with these parameters:
    {
        "scale": 20.0,
        "similarity_fct": "cos_sim",
        "gather_across_devices": false
    }
    

Evaluation Dataset

Unnamed Dataset

  • Size: 1,037 evaluation samples
  • Columns: anchor, positive, and negative
  • Approximate statistics based on the first 1000 samples:
    anchor positive negative
    type string string string
    details
    • min: 3 tokens
    • mean: 6.32 tokens
    • max: 15 tokens
    • min: 195 tokens
    • mean: 255.8 tokens
    • max: 256 tokens
    • min: 196 tokens
    • mean: 255.74 tokens
    • max: 256 tokens
  • Samples:
    anchor positive negative
    fluorescent fixture {"product_title": "Lithonia Lighting 4-Light High Output White Fluorescent Strip Light Fixture", "product_description": "The Lithonia Lighting 4-Light High Output Fluorescent Strip Light Fixture features a slim, compact design ideal for closets, attics and garages. This light fixture can be wall or ceiling mounted and uses 4 T5 fluorescent bulbs (not included). The slim channel makes these easy for handling in unit or row installation.Slim, compact designUses four 54-Watt T5 high output fluorescent light bulbs (not included)Can be wall or ceiling mountedIdeal for closets, garages and atticsUL listedHardware included", "Actual Color Temperature (K)": "0", "Bulb Type Included": "No Bulbs Included", "Bullet01": "Slim, compact design", "Bullet02": "Uses four 54-Watt T5 high output fluorescent light bulbs (not included)", "Bullet03": "Can be wall or ceiling mounted", "Bullet04": "Ideal for closets, garages and attics", "Bullet05": "UL listed", "Bullet06": "Hardware included", "Certification... {"product_title": "Philips 4 ft. T8 32-Watt Daylight (6500K) ALTO Linear Fluorescent Light Bulb (10-Pack)", "product_description": "Create a cool and refreshing environment with the Philips 32-Watt 4 ft. T8 Daylight Linear Alto Fluorescent Light Bulbs (10-Pack). These light bulbs cast an arctic white light making them ideal for bathrooms, laundry rooms or any room. The long-lasting bulbs can shine up to 30,000 hours using 2,850 Lumens. These bulbs feature Alto low-mercury technology for environmentally friendly use.Brightness: 2850 LumensEstimated yearly energy cost: $3.85 (based on 3 hours/day 11/kWh. cost depend on rates and use)Life: 27.4 years (based on 3 hours/day)Light appearance: 6500K (daylight)Energy used: 32-WattLumens per Watt: 89.06Great for indoor, commercial and residential useCasts an arctic white light perfect for commercial or residential general purpose use in bathrooms, laundry rooms or kitchensT8, bi-pinFluorescent lamp with alto technology has reduced mercury conte...
    plumber wrench {"product_title": "HDX Basin Wrench", "product_description": "The HDX Steel Basin Wrench is made of heavy-duty steel and has a reversible head for working in tight spaces. It has a spring-loaded jaw that securely grips and easily removes even the toughest frozen-on faucet nuts. All products have been reviewed and tested by a third party testing lab for durability, function ability and product longevity.Heavy-duty steel construction and solid-steel handle for maximum leverageReversible head for tight spacesSpring-loaded jaw prevents slippageDurable solid steel handle for greater torque leverage", "Bullet01": "Heavy-duty steel construction and solid-steel handle for maximum leverage", "Bullet02": "Reversible head for tight spaces", "Bullet03": "Spring-loaded jaw prevents slippage", "Bullet04": "Durable solid steel handle for greater torque leverage", "Maximum pipe size capacity (in.)": "1", "MFG Brand Name": "HDX", "Product Depth (in.)": "0.58", "Product Height (in.)": "11.7", "Product W... {"product_title": "Husky 2 in. Universal Faucet Nut Wrench", "product_description": "The unique slots in the tool grab locknut wings without breaking them off. This compact tool fits in hard to reach areas behind the sink. The universal design fits metal lock nuts, coupling nuts, PVC locknuts, most metal and plastic toilet supply nuts.Cutaway design fits around existing supply lines to remove coupling nutOnce coupling nuts is remove, cutaway fits around the faucet shank (1/2 in.) to remove metal and plastic nutsUse with 1/4 in. or 3/8 in. ratchet drive, adapter includedPatented design in the US and Canada", "Bullet01": "Cutaway design fits around existing supply lines to remove coupling nut", "Bullet02": "Once coupling nuts is remove, cutaway fits around the faucet shank (1/2 in.) to remove metal and plastic nuts", "Bullet03": "Use with 1/4 in. or 3/8 in. ratchet drive, adapter included", "Bullet04": "Patented design in the US and Canada", "Maximum pipe size capacity (in.)": "1.5", "MF...
    glidden gl ext sg base 1 {"product_title": "Glidden Premium 1-gal. Semi-Gloss Latex Exterior Paint", "product_description": "The Glidden Premium 1-gal. semi-gloss Latex Exterior Paint is ideal for application on all siding, trim, doors, shutters and fences. The 100% acrylic latex base resists fading and other damage caused by UV rays and severe weather conditions. The mildew-resistant paint can be applied in temperatures as low as 35F, allowing you an extended painting season.California residents: see Proposition 65 informationIdeal for application on all siding, trim, doors, shutters and fencesCan be applied in temperatures as low as 35F100% acrylic latex base provides a durable, chip-resistant finishFormulated for an easy cleanup with soap and waterQuick-drying formula is dry to the touch in as little as 45 minutesResists fading and other damage caused by UV rays and severe weather conditions for a long-lasting freshly painted look with minimal maintenanceResists mildew growth to maintain a pristine app... {"product_title": "Glidden Premium 1-qt. Semi-Gloss Latex Exterior Paint", "product_description": "The Glidden Premium 1-qt. Semi-Gloss Latex Exterior Paint is ideal for application on all siding, trim, doors, shutters and fences. The 100% acrylic latex formula resists fading and other damage caused by UV rays and severe weather conditions. It can be applied in temperatures as low as 35F, allowing you an extended painting season.California residents: see Proposition 65 informationIdeal for application on all siding, trim, doors, shutters and fencesCan be applied in temperatures as low as 35 degrees Fahrenheit100% acrylic latex base provides exceptional adhesion and coverageFormulated for an easy cleanup with soap and waterQuick-drying formula is dry to the touch in as little as 45 minutesDurable, chip-resistant finishResists fading and other damage caused by UV rays and severe weather conditions for a long-lasting freshly painted appearance with minimal maintenanceResists mildew g...
  • Loss: MultipleNegativesRankingLoss with these parameters:
    {
        "scale": 20.0,
        "similarity_fct": "cos_sim",
        "gather_across_devices": false
    }
    

Training Hyperparameters

Non-Default Hyperparameters

  • eval_strategy: epoch
  • num_train_epochs: 5
  • load_best_model_at_end: True
  • batch_sampler: no_duplicates

All Hyperparameters

Click to expand
  • overwrite_output_dir: False
  • do_predict: False
  • eval_strategy: epoch
  • prediction_loss_only: True
  • per_device_train_batch_size: 8
  • 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: 5e-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: 5
  • max_steps: -1
  • lr_scheduler_type: linear
  • lr_scheduler_kwargs: {}
  • warmup_ratio: 0.0
  • 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: 42
  • data_seed: None
  • jit_mode_eval: False
  • use_ipex: False
  • bf16: False
  • 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}
  • 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 Validation Loss test_cosine_accuracy eval_cosine_accuracy
-1 -1 - - 0.6292 -
0.0858 100 0.8689 - - -
0.1715 200 0.7896 - - -
0.2573 300 0.696 - - -
0.3431 400 0.6568 - - -
0.4288 500 0.6342 - - -
0.5146 600 0.5934 - - -
0.6003 700 0.6081 - - -
0.6861 800 0.5093 - - -
0.7719 900 0.5112 - - -
0.8576 1000 0.548 - - -
0.9434 1100 0.4435 - - -
1.0 1166 - 0.3647 - 0.8611
1.0292 1200 0.387 - - -
1.1149 1300 0.2302 - - -
1.2007 1400 0.2607 - - -
1.2864 1500 0.2399 - - -
1.3722 1600 0.2216 - - -
1.4580 1700 0.2027 - - -
1.5437 1800 0.2286 - - -
1.6295 1900 0.2349 - - -
1.7153 2000 0.1967 - - -
1.8010 2100 0.2032 - - -
1.8868 2200 0.1951 - - -
1.9726 2300 0.1875 - - -
2.0 2332 - 0.2562 - 0.9142
2.0583 2400 0.142 - - -
2.1441 2500 0.1001 - - -
2.2298 2600 0.0937 - - -
2.3156 2700 0.1041 - - -
2.4014 2800 0.1048 - - -
2.4871 2900 0.0897 - - -
2.5729 3000 0.112 - - -
2.6587 3100 0.1397 - - -
2.7444 3200 0.0991 - - -
2.8302 3300 0.0985 - - -
2.9160 3400 0.0988 - - -
3.0 3498 - 0.2049 - 0.9325
3.0017 3500 0.0963 - - -
3.0875 3600 0.0535 - - -
3.1732 3700 0.0637 - - -
3.2590 3800 0.0615 - - -
3.3448 3900 0.0649 - - -
3.4305 4000 0.0537 - - -
3.5163 4100 0.0642 - - -
3.6021 4200 0.0579 - - -
3.6878 4300 0.0411 - - -
3.7736 4400 0.0779 - - -
3.8593 4500 0.0607 - - -
3.9451 4600 0.0654 - - -
4.0 4664 - 0.1888 - 0.9441
4.0309 4700 0.0541 - - -
4.1166 4800 0.0479 - - -
4.2024 4900 0.0413 - - -
4.2882 5000 0.0492 - - -
4.3739 5100 0.0446 - - -
4.4597 5200 0.043 - - -
4.5455 5300 0.052 - - -
4.6312 5400 0.0395 - - -
4.7170 5500 0.0325 - - -
4.8027 5600 0.0375 - - -
4.8885 5700 0.0416 - - -
4.9743 5800 0.0403 - - -
5.0 5830 - 0.1843 - 0.947
-1 -1 - - 0.9371 -
  • The bold row denotes the saved checkpoint.

Framework Versions

  • Python: 3.11.5
  • Sentence Transformers: 5.1.0
  • Transformers: 4.56.1
  • PyTorch: 2.8.0+cu128
  • Accelerate: 1.10.1
  • Datasets: 4.0.0
  • Tokenizers: 0.22.0

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{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}
}
Downloads last month
2
Safetensors
Model size
22.7M params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for zhenqingli/all-MiniLM-L6-v2-ft-home-depot-triplet

Finetuned
(779)
this model

Papers for zhenqingli/all-MiniLM-L6-v2-ft-home-depot-triplet

Evaluation results