SentenceTransformer based on BAAI/bge-base-en-v1.5

This is a sentence-transformers model finetuned from BAAI/bge-base-en-v1.5 on the marvelousjaguar_embedding dataset. It maps sentences & paragraphs to a 768-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: BAAI/bge-base-en-v1.5
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 dimensions
• Similarity Function: Cosine Similarity
• Training Dataset:
  • marvelousjaguar_embedding

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': 512, 'do_lower_case': True}) with Transformer model: BertModel 
  (1): Pooling({'word_embedding_dimension': 768, '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()
)

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 = [
    ' stated in its report which is included \nherein.\nChanges in Internal Control Over Financial Reporting\nThere have been no changes in our internal control over financial reporting during the quarter ended January\xa026, 2025  \nthat have materially affected, or are reasonably likely to materially affect, our internal control over financial reporting. In \nfiscal year 2022, we began an upgrade of our ERP system, which will update much of our existing core financial systems. \nThe ERP system is designed to accurately maintain our financial records used to report operating results. The upgrade will \noccur in phases. We will continue to evaluate each quarter whether there are changes that materially affect our internal \ncontrol over financial reporting.\nInherent Limitations on Effectiveness of Controls\nOur management, including our Chief Executive Officer and Chief Financial Officer, does not expect that our disclosure \ncontrols and procedures or our internal controls, will prevent all error and all fraud. A control system, no matter how well \nconceived and operated, can provide only reasonable, not absolute, assurance that the objectives of the control system \nare met. Further, the design of a control system must reflect the fact that there are resource constraints, and the \nbenefits of controls must be considered relative to their costs. Because of the inherent limitations in all control systems, \nno evaluation of controls can provide absolute assurance that all control issues and instances of fraud, if any, within \nNVIDIA have been detected.\n46',
    "Compare the impacts of 2022 export controls on NVIDIA's operations versus its competitive position and licensing challenges.",
    'Compare how cryptocurrency market volatility and Ethereum 2.0 merge differently influence GPU demand and pricing.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

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

Training Details

Training Dataset

marvelousjaguar_embedding

• Dataset: marvelousjaguar_embedding
• Size: 455 training samples
• Columns: anchor and positive
• Approximate statistics based on the first 455 samples:

  	anchor	positive
  type	string	string
  details	min: 5 tokensmean: 308.11 tokensmax: 505 tokens	min: 16 tokensmean: 25.4 tokensmax: 51 tokens

• Samples:

  anchor	positive
  (5) The PSU was earned based on achievement of a performance goal for the 1-year period that ended on January 26, 2025. The PSU vested as to 25% of the shares on March 19, 2025, and vests as to 6.25% approximately every three months thereafter over the next three years such that the PSU will be fully vested on March 15, 2028.(6) Represents shares that could be earned upon achievement of Stretch Compensation Plan goals, based on our TSR relative to the S&P 500 from January 30, 2023 through January 25, 2026. If the performance goal is achieved, 100% of the shares earned will vest on March 18, 2026. If the Threshold performance goal is achieved, 126,220 shares will be earned by Mr. Huang, 6,190 shares will be earned by Ms. Kress, 5,960 shares will be earned by Mr. Puri, 5,100 shares will be earned by Ms. Shoquist, and 5,100 shares will be earned by Mr. Teter. If the Base Compensation Plan performance goal is achieved, 504,910 shares will be earned by Mr. Huang, 24,780 shares wil...	Describe NVIDIA's shift from chip maker to AI factory infrastructure builder, and its AI impact.
  Israel and Regional ConflictsWe are monitoring the impact of the geopolitical conflict in and around Israel on our operations, including the health and safety of our approximately 4,700 employees in the region who primarily support the research and development, operations, and sales and marketing of our networking products. Our global supply chain for our networking products has not experienced any significant impact. Some of our employees in the region have been on active military duty for an extended period and may continue to be absent, which may cause disruption to our product development or operations. We have not experienced significant impact or expense to our business; however, if the conflict is further extended or expanded, it could impact future product development, operations, and revenue or create other uncertainty for our business.Fiscal Year 2025 Summary  Year Ended  Jan 26, 2025 Jan 28, 2024 Change($ in millions, except per share data)Revenue $ 130,497 $ 60...	Describe NVIDIA's shift from chip maker to AI factory infrastructure builder, and its AI impact.
  , from training to reasoning at scale. With FP4 precision, long-context processing, and dynamic compute allocation, it makes test-time scaling real time as well as energy and cost efficient. Now in full production, Blackwell is NVIDIA’s fastest product ramp ever—powering AI infrastructure across hyperscalers, enterprises, and sovereign clouds.At its core is the NVIDIA GB200 NVL72: 72 GPUs, 36 Grace CPUs, unified by NVLink, Spectrum-X Ethernet, and liquid or air cooling. It delivers up to 30x faster inference for trillion-parameter models like GPT-4 or Grok-3—at the lowest total cost per watt, per rack, and per dollar invested.A Full-Stack Platform for Scaling IntelligenceBlackwell is more than silicon. Its supporting technologies are foundational to the next wave of AI factories: › NVLink 5 and NVSwitch connect 72 Blackwell GPUs into a massive compute resource enabling exascale performance for AI factories. › Spectrum-X extends networking capabilities to s...	Describe NVIDIA's shift from chip maker to AI factory infrastructure builder, and its AI impact.

• Loss: MultipleNegativesRankingLoss with these parameters:

  {
      "scale": 20.0,
      "similarity_fct": "cos_sim"
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• learning_rate: 2e-05
• warmup_ratio: 0.1

All Hyperparameters

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: no
• 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: 2e-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: 3
• 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: 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: 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}
• 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: batch_sampler
• multi_dataset_batch_sampler: proportional

Training Logs

Framework Versions

• Python: 3.11.13
• Sentence Transformers: 3.4.1
• Transformers: 4.51.3
• PyTorch: 2.7.0
• Accelerate: 1.3.0
• Datasets: 2.19.1
• 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",
}

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}
}