mini-lm-stage1_only_etf

SentenceTransformer based on sentence-transformers/all-MiniLM-L12-v2

This is a sentence-transformers model finetuned from sentence-transformers/all-MiniLM-L12-v2 on the stage1_v1 dataset. 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-L12-v2 <!-- at revision c004d8e3e901237d8fa7e9fff12774962e391ce5 -->
• Maximum Sequence Length: 128 tokens
• Output Dimensionality: 384 dimensions
• Similarity Function: Cosine Similarity
• Training Dataset:
  • stage1_v1

<!-- - **Language:** Unknown --> <!-- - **License:** Unknown -->

Model Sources

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

Full Model Architecture

SentenceTransformer(
  (0): Transformer({'max_seq_length': 128, 'do_lower_case': False}) with Transformer model: 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("sentence_transformers_model_id")
# Run inference
sentences = [
    'The KraneShares Emerging Markets Consumer Technology ETF (KEMQ) aims to track the Solactive Emerging Market Consumer Technology Index, investing at least 80% of its net assets in instruments within or similar to its underlying index. This index comprises the equity securities of the 50 largest companies by market capitalization, primarily from emerging and frontier markets, focusing on the consumer and technology sectors. KEMQ offers concentrated exposure to emerging market tech companies, selected by a committee and tier-weighted based on market cap. The largest 10 securities are weighted at 3.5% each, the next 20 at 2.5% each, and the remaining 20 at 0.75% each. The index is reviewed and adjusted quarterly to ensure it reflects the most relevant market opportunities.',
    'The First Trust Consumer Discretionary AlphaDEX® ETF (FXD) is designed to outperform the US consumer discretionary sector by tracking the StrataQuant® Consumer Discretionary Index. This index is a modified equal-dollar weighted benchmark that selects stocks from the Russell 1000® using the innovative AlphaDEX® methodology. This approach incorporates both value and growth criteria to identify stocks with the potential for positive alpha. FXD strategically invests at least 90% of its net assets in these selected securities, resulting in notable mid-cap exposure and distinct industry tilts that differentiate it from traditional sector-focused investments. The fund employs a quasi-active selection process, reconstituted and rebalanced on a quarterly basis, making it an appealing choice for investors seeking higher returns rather than mere sector replication.',
    'The SPDR S&P Global Infrastructure ETF (GII) employs a strategic management approach aimed at closely tracking the S&P Global Infrastructure Index. To achieve this, the ETF allocates a minimum of 80% of its assets to the securities included in the index and their related depositary receipts. The index comprises 75 of the largest publicly listed infrastructure companies worldwide, selected based on stringent investability criteria. GII specifically targets firms within the energy, transportation, and utility sectors, maintaining a diversified portfolio with a composition of 30 transportation companies, 30 utility companies, and 15 energy companies. To enhance diversification and mitigate concentration risk, sector weights are capped at 40% for transportation and utilities, and 20% for energy. Furthermore, the fund limits the weight of any single security to a maximum of 5%. Within each sector, stocks are weighted according to market capitalization. GII undergoes substantial adjustments during its semi-annual rebalancing, ensuring alignment with the evolving market landscape while adhering to its investment strategy.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 384]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities.shape)
# [3, 3]

<!-- ### Direct Usage (Transformers) <details><summary>Click to see the direct usage in Transformers</summary> </details> --> <!-- ### Downstream Usage (Sentence Transformers) You can finetune this model on your own dataset. <details><summary>Click to expand</summary> </details> --> <!-- ### Out-of-Scope Use *List how the model may foreseeably be misused and address what users ought not to do with the model.* --> <!-- ## Bias, Risks and Limitations *What are the known or foreseeable issues stemming from this model? You could also flag here known failure cases or weaknesses of the model.* --> <!-- ### Recommendations *What are recommendations with respect to the foreseeable issues? For example, filtering explicit content.* -->

Training Details

Training Dataset

stage1_v1

• Dataset: stage1_v1 at 9be9e9c
• Size: 2,752 training samples
• Columns: <code>query</code>, <code>positive</code>, and <code>negative</code>
• Approximate statistics based on the first 1000 samples:

  	query	positive	negative
  type	string	string	string
  details	<ul><li>min: 123 tokens</li><li>mean: 128.0 tokens</li><li>max: 128 tokens</li></ul>	<ul><li>min: 123 tokens</li><li>mean: 128.0 tokens</li><li>max: 128 tokens</li></ul>	<ul><li>min: 128 tokens</li><li>mean: 128.0 tokens</li><li>max: 128 tokens</li></ul>

• Samples:

  query	positive	negative
  <code>The Global X Aging Population ETF (AGNG) is a fund designed to invest in companies that benefit from the growing number of older people in the world. It focuses on businesses in developed countries that help improve and extend the lives of seniors. This includes companies that work in areas like biotechnology, medical devices, pharmaceuticals, senior living facilities, and healthcare services. The fund aims to support the aging population trend by investing over 80% of its money in these sectors. <br><br>AGNG uses a special method to choose its investments, looking at a variety of businesses, including those in insurance and consumer products. The fund is updated once a year to make sure it stays balanced and diverse, meaning it spreads its investments across different kinds of companies. Before April 2021, it was called the Global X Longevity Thematic ETF and went by the ticker LNGR. This ETF is a way for investors to tap into the growing market of services and products for seniors.</code>	<code>The Amplify High Income ETF (YYY) is a fund of funds that aims to replicate the performance of the ISE High Income™ Index by investing at least 80% of its net assets in securities of the index. This index comprises the top 60 U.S. exchange-listed closed-end funds (CEFs), selected and weighted based on yield, discount to NAV, and trading volume. YYY typically holds about 30 CEFs, with a maximum weight of 4.25% per fund at rebalance, and can include funds across major asset classes. The ETF's strategy focuses on acquiring discounted CEFs with high yields and sufficient liquidity to minimize trading costs. YYY's fee structure includes the expenses of its constituent funds. The fund was reorganized under Amplify ETFs in 2019, maintaining its investment objectives and index.</code>	<code>The iShares Copper and Metals Mining ETF (ICOP) is strategically designed to replicate the performance of the STOXX Global Copper and Metals Mining Index, concentrating on equities from both U.S. and international companies primarily involved in copper and metal ore extraction. The fund commits at least 80% of its assets to the index's component securities, allowing for up to 20% allocation to derivatives such as futures, options, and swaps, as well as cash and equivalents. ICOP employs a market-capitalization weighted strategy, categorizing companies into three tiers based on their revenue exposure to copper mining: Tier 1 encompasses firms with over 50% revenue from copper, Tier 2 includes those with 25-50%, and Tier 3 comprises companies determined by market share. The index undergoes quarterly rebalancing, implementing caps of 8% on individual holdings and limiting those exceeding 4.5% to a combined weight of 45%. This non-diversified fund provides concentrated exposure specificall...</code>
  <code>The Global X Aging Population ETF (AGNG) seeks to track the performance of the Indxx Aging Population Thematic Index, investing over 80% of its assets in securities from developed markets that support the demographic trend of longer life spans. The fund targets companies involved in biotechnology, medical devices, pharmaceuticals, senior living facilities, and specialized healthcare services, focusing on enhancing and extending the lives of senior citizens. AGNG employs a proprietary research and analysis process, crossing traditional sector lines to include diverse businesses such as insurance and consumer products. The ETF is reconstituted and rebalanced annually, using a modified market-cap weighting with specific caps and floors to ensure diversification. Prior to April 2021, it was known as the Global X Longevity Thematic ETF under the ticker LNGR.</code>	<code>The iShares Biotechnology ETF (IBB) aims to track the performance of the NYSE Biotechnology Index, which comprises U.S.-listed biotechnology companies. These companies are involved in the research and development of therapeutic treatments and the production of tools or systems for biotechnology processes, excluding those focused on mass pharmaceutical production. IBB invests at least 80% of its assets in the index's component securities and up to 20% in futures, options, swap contracts, cash, and equivalents. The fund employs a modified market-cap-weighted methodology, capping the five largest constituents at 8% and others at 4%. It is non-diversified, rebalances quarterly, and fully reconstitutes annually in December. Prior to June 21, 2021, it was known as the iShares Nasdaq Biotechnology ETF.</code>	<code>The Invesco Global Clean Energy ETF (PBD) is designed to track the WilderHill New Energy Global Innovation Index, dedicating a minimum of 90% of its assets to securities within this index, which includes American Depositary Receipts (ADRs) and Global Depositary Receipts (GDRs). The index predominantly features companies committed to clean energy technologies, conservation, efficiency, and the advancement of renewable energy. While PBD is passively managed, it employs a strategy akin to active management by focusing on companies with significant capital appreciation potential, particularly emphasizing pure-play small- and mid-cap firms. The fund boasts a global diversification, with approximately half of its assets allocated internationally, while maintaining a limit of 5% on its largest holdings. The index undergoes quarterly rebalancing and reconstitution, ensuring a dynamic and varied portfolio that reflects the evolving landscape of the clean energy s...</code>
  <code>The Global X Aging Population ETF (AGNG) is strategically designed to track the performance of the Indxx Aging Population Thematic Index, focusing on the investment potential arising from the global demographic shift towards longer life spans. The ETF allocates over 80% of its assets to securities primarily in developed markets that are aligned with this trend. Target sectors include biotechnology, medical devices, pharmaceuticals, senior living facilities, and specialized healthcare services, all aimed at improving the quality of life for senior citizens. Additionally, AGNG incorporates a broader investment approach by including companies from diverse sectors such as insurance and consumer products, which are relevant to aging populations. The fund employs a proprietary research and analysis methodology that transcends traditional sector boundaries. It is reconstituted and rebalanced annually, utilizing a modified market-cap weighting approach that includes specific caps and floors to...</code>	<code>The iShares U.S. Health Care Providers ETF (IHF) employs a strategy aimed at closely tracking the performance of the Dow Jones U.S. Select Health Care Providers Index. This ETF is managed by investing at least 80% of its assets in the securities of companies that constitute the index, which primarily includes U.S. firms operating within the healthcare services sector. The remaining 20% of the fund's assets may be allocated to various financial instruments such as futures, options, swaps, cash, and cash equivalents to enhance liquidity and manage risk. IHF strategically targets key sectors within the healthcare provider landscape, focusing on managed healthcare, healthcare facilities, and health insurance companies, while deliberately excluding pharmaceutical firms. This approach allows IHF to offer cap-weighted exposure tailored to the healthcare provider space, providing investors with a concentrated yet comprehensive investment vehicle that captures the dynamics of health insurance a...</code>	<code>The First Trust Indxx NextG ETF (NXTG) seeks to replicate the performance of the Indxx 5G & NextG Thematic Index by investing at least 90% of its net assets in the index's securities. This index tracks global equities of companies that are significantly investing in the research, development, and application of fifth generation (5G) and next generation digital cellular technologies. NXTG includes mid- and large-cap companies from two main sub-themes: 5G infrastructure & hardware, which encompasses data center REITs, cell tower REITs, equipment manufacturers, network testing and validation equipment, and mobile phone manufacturers; and telecommunication service providers operating cellular and wireless communication networks with 5G access. Prior to May 29, 2019, NXTG was known as the First Trust NASDAQ Smartphone Index Fund (ticker FONE), focusing more broadly on the cellular phone industry.</code>

• Loss: <code>TripletLoss</code> with these parameters:

  {
      "distance_metric": "TripletDistanceMetric.COSINE",
      "triplet_margin": 0.5
  }
  

Evaluation Dataset

stage1_v1

• Dataset: stage1_v1 at 9be9e9c
• Size: 688 evaluation samples
• Columns: <code>query</code>, <code>positive</code>, and <code>negative</code>
• Approximate statistics based on the first 688 samples:

  	query	positive	negative
  type	string	string	string
  details	<ul><li>min: 123 tokens</li><li>mean: 127.99 tokens</li><li>max: 128 tokens</li></ul>	<ul><li>min: 123 tokens</li><li>mean: 127.99 tokens</li><li>max: 128 tokens</li></ul>	<ul><li>min: 120 tokens</li><li>mean: 127.99 tokens</li><li>max: 128 tokens</li></ul>

• Samples:

  query	positive	negative
  <code>The Global X Aging Population ETF (AGNG) aims to replicate the performance of the Indxx Aging Population Thematic Index by investing over 80% of its assets in securities from developed markets that capitalize on the trend of increasing life expectancies. The fund primarily focuses on companies engaged in biotechnology, medical devices, pharmaceuticals, senior living facilities, and specialized healthcare services, all aimed at enhancing and extending the quality of life for senior citizens. AGNG employs a proprietary research methodology that transcends traditional sector boundaries, incorporating a diverse range of industries, including insurance and consumer products. The ETF is reconstituted and rebalanced annually, utilizing a modified market-cap weighting approach with specific caps and floors to maintain diversification. Previously known as the Global X Longevity Thematic ETF under the ticker LNGR until April 2021, AGNG continues to align its investments with key demographic shif...</code>	<code>The SPDR S&P Biotech ETF (XBI) employs a strategic management approach aimed at closely tracking the performance of the S&P Biotechnology Select Industry Index through a sampling strategy. By investing a minimum of 80% of its total assets in the securities of this index, XBI focuses specifically on the biotechnology sector, which is a subset of the broader S&P Total Market Index. The ETF is distinguished by its equal-weighted methodology, which ensures diversified exposure across U.S. biotech stocks, particularly emphasizing small- and micro-cap companies. This approach mitigates single-name risk by reducing the influence of larger companies, resulting in a lower weighted-average market capitalization relative to its competitors. Additionally, the ETF's structure limits overlap with the pharmaceutical industry, allowing for a more concentrated investment in innovative biotech firms. The index undergoes quarterly rebalancing, which supports its commitment to maintaining a focused invest...</code>	<code>The VanEck Mortgage REIT Income ETF (MORT) employs a strategic approach to replicate the performance of the MVIS® US Mortgage REITs Index, focusing on a diverse range of mortgage real estate investment trusts (REITs). By allocating at least 80% of its total assets to securities within this benchmark, MORT targets companies across various market capitalizations, including small-, medium-, and large-cap mortgage REITs. The ETF is managed with a market-cap-weighted strategy, ensuring that larger companies have a more significant influence on its performance. While MORT features a lower expense ratio compared to its peer, the iShares Mortgage Real Estate Capped ETF (REM), it does experience challenges with liquidity. The fund maintains a concentrated portfolio, heavily aligned with its top holdings, which allows for targeted exposure to the mortgage REIT sector. This management strategy positions MORT as a compelling choice for investors seeking specialized investments in the mortgage REIT...</code>
  <code>The Global X Aging Population ETF (AGNG) aims to replicate the performance of the Indxx Aging Population Thematic Index by investing over 80% of its assets in securities from developed markets that capitalize on the trend of increasing life expectancies. The fund primarily focuses on companies engaged in biotechnology, medical devices, pharmaceuticals, senior living facilities, and specialized healthcare services, all aimed at enhancing and extending the quality of life for senior citizens. AGNG employs a proprietary research methodology that transcends traditional sector boundaries, incorporating a diverse range of industries, including insurance and consumer products. The ETF is reconstituted and rebalanced annually, utilizing a modified market-cap weighting approach with specific caps and floors to maintain diversification. Previously known as the Global X Longevity Thematic ETF under the ticker LNGR until April 2021, AGNG continues to align its investments with key demographic shif...</code>	<code>The Range Cancer Therapeutics ETF (CNCR) is designed to track the Range Oncology Therapeutics Index, targeting U.S. exchange-listed pharmaceutical and biotechnology stocks, as well as American Depository Receipts (ADRs) with market capitalizations exceeding $250 million. Launched in 2023 by Range Fund Holdings, CNCR strategically allocates a minimum of 80% of its assets to the securities within the index. This ETF provides equal-weighted exposure to companies engaged in the research, development, and commercialization of oncology drugs, placing a spotlight on smaller firms with significant growth potential. CNCR is particularly appealing to investors focused on the cancer therapeutics sector. The ETF, formerly known as the Loncar Cancer Immunotherapy ETF, broadened its investment scope in October 2023 by merging with the Loncar China BioPharma ETF, thereby enhancing its exposure to promising opportunities in the Chinese markets.</code>	<code>The Invesco S&P 500 Equal Weight Energy ETF (RSPG) is designed to replicate the performance of the S&P 500® Equal Weight Energy Index by investing a minimum of 90% of its total assets in securities that compose this index. This index includes all companies within the S&P 500® Energy Index that fall under the energy sector, as defined by the Global Industry Classification Standard (GICS). As a large-cap sector fund, RSPG offers equal-weight exposure to a diverse array of U.S. energy companies across various sub-industries, enhancing portfolio diversification. The fund is rebalanced quarterly to ensure a minimum inclusion of 22 companies, and it may also incorporate leading firms from the S&P MidCap 400 Index if necessary to maintain this threshold. Notably, prior to June 7, 2023, RSPG was traded under the ticker RYE.</code>
  <code>The First Trust RBA American Industrial Renaissance ETF (AIRR) is designed to closely track the performance of the Richard Bernstein Advisors American Industrial Renaissance® Index. This passively managed fund allocates a minimum of 90% of its net assets to equity securities within the index, emphasizing small and mid-cap U.S. companies primarily in the industrial and community banking sectors. Key industries targeted include Commercial Services & Supplies, Construction & Engineering, Electrical Equipment, Machinery, and Banks. The index utilizes a multifactor selection approach, systematically excluding firms with more than 25% of sales from outside the U.S. and community banks situated outside traditional Midwestern manufacturing regions. A proprietary optimization method is applied for weighting, limiting the banking sector to 10% and individual issuers to 4%. The index undergoes quarterly reconstitution and rebalancing, maintaining a focus on companies with a favorable 12-month for...</code>	<code>The Invesco Global Water ETF (PIO) aims to track the investment results of the NASDAQ OMX Global Water Index, investing at least 90% of its assets in securities within the index, including ADRs and GDRs. This index comprises global exchange-listed companies from the U.S., developed, and emerging markets that produce water conservation and purification products for homes, businesses, and industries. PIO employs a liquidity-weighted strategy, resulting in a concentrated portfolio dominated by large- to mid-cap firms. Eligible companies must participate in the Green Economy, as determined by SustainableBusiness.com LLC. The fund uses full replication to track its index, with quarterly rebalancing and annual reconstitution, while maintaining country and issuer diversification limits.</code>	<code>The Jacob Funds Inc. Jacob Forward ETF (JFWD) is actively managed with a focus on achieving long-term capital growth by investing in equity securities of U.S. companies engaged in innovative and disruptive technologies. The fund primarily holds common stocks but may also include other equity securities like preferred stocks, rights, or warrants. It targets companies of all sizes, with a significant emphasis on those in the early stages of development, particularly within the healthcare and information technology sectors. JFWD employs a forward-looking investment strategy, selecting securities based on a qualitative and quantitative assessment of companies' potential for above-average growth. The fund may also gain up to 25% foreign market exposure through global operations of U.S. companies. Notably, JFWD is non-diversified and will be delisted, with its last trading day on December 23, 2024.</code>

• Loss: <code>TripletLoss</code> with these parameters:

  {
      "distance_metric": "TripletDistanceMetric.COSINE",
      "triplet_margin": 0.5
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: steps
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• num_train_epochs: 10
• warmup_ratio: 0.1
• bf16: True
• dataloader_drop_last: True
• load_best_model_at_end: True
• batch_sampler: no_duplicates

All Hyperparameters

<details><summary>Click to expand</summary>

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: steps
• prediction_loss_only: True
• per_device_train_batch_size: 16
• per_device_eval_batch_size: 16
• 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: 10
• 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: 42
• 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: 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}
• tp_size: 0
• 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}
• deepspeed: None
• label_smoothing_factor: 0.0
• optim: adamw_torch
• 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
• 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
• eval_use_gather_object: False
• average_tokens_across_devices: False
• prompts: None
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional

</details>

Training Logs

Framework Versions

• Python: 3.10.12
• Sentence Transformers: 4.1.0
• Transformers: 4.51.3
• PyTorch: 2.1.0+cu118
• Accelerate: 1.6.0
• Datasets: 3.5.0
• Tokenizers: 0.21.1

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",
}

TripletLoss

@misc{hermans2017defense,
    title={In Defense of the Triplet Loss for Person Re-Identification},
    author={Alexander Hermans and Lucas Beyer and Bastian Leibe},
    year={2017},
    eprint={1703.07737},
    archivePrefix={arXiv},
    primaryClass={cs.CV}
}

<!-- ## Glossary *Clearly define terms in order to be accessible across audiences.* --> <!-- ## Model Card Authors *Lists the people who create the model card, providing recognition and accountability for the detailed work that goes into its construction.* --> <!-- ## Model Card Contact *Provides a way for people who have updates to the Model Card, suggestions, or questions, to contact the Model Card authors.* -->