SentenceTransformer based on llm-semantic-router/mmbert-embed-32k-2d-matryoshka

This is a sentence-transformers model finetuned from llm-semantic-router/mmbert-embed-32k-2d-matryoshka. 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: llm-semantic-router/mmbert-embed-32k-2d-matryoshka
• Maximum Sequence Length: 32768 tokens
• Output Dimensionality: 768 dimensions
• Similarity Function: Cosine Similarity

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': 32768, 'do_lower_case': False, 'architecture': 'ModernBertModel'})
  (1): Pooling({'word_embedding_dimension': 768, '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})
)

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 = [
    'What is included in Item 8 of the document?',
    'What is included in Item 8 of the document?\n\nAnswer: Financial Statements and Supplementary Data',
    'What is the content of Item 8 in the document?\n\nAnswer: Item 8 of the document includes Financial Statements and Supplementary Data.',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0078, 0.9297, 0.5508],
#         [0.9297, 0.9922, 0.6680],
#         [0.5508, 0.6680, 1.0000]], dtype=torch.bfloat16)

Training Details

Training Dataset

Unnamed Dataset

• Size: 606 training samples
• Columns: sentence_0, sentence_1, and sentence_2
• Approximate statistics based on the first 606 samples:

  	sentence_0	sentence_1	sentence_2
  type	string	string	string
  details	min: 10 tokensmean: 22.46 tokensmax: 55 tokens	min: 18 tokensmean: 52.15 tokensmax: 229 tokens	min: 18 tokensmean: 49.13 tokensmax: 120 tokens

• Samples:

  sentence_0	sentence_1	sentence_2
  What led to the increase in operating income margin for the Dollar Tree segment in 2022?	What led to the increase in operating income margin for the Dollar Tree segment in 2022?Answer: The increase in operating income margin for the Dollar Tree segment in 2022 was primarily due to the gross profit margin increase and a decrease in the selling, general and administrative expense rate.	What was the increase in the gross profit margin for the fiscal year 2022 compared to the previous year?Answer: The increase in the gross profit margin for the fiscal year 2022 compared to the previous year was 2.1%.
  How much net cash was provided by operating activities in 2022?	How much net cash was provided by operating activities in 2022?Answer: $4.5 billion	What were the main components contributing to the net cash provided by operating activities in 2023?Answer: Operating income, depreciation and amortization, and various adjustments in operating assets and liabilities were the main contributors.
  How was the stock-based compensation expense of $254 million accounted for in the company's financial statements?	How was the stock-based compensation expense of $254 million accounted for in the company's financial statements?Answer: The stock-based compensation expense of $254 million was recorded in the line item selling, general, and administrative expenses in the company's consolidated statement of income.	How much did the cost of revenue increase in stock-based compensation expense from fiscal year 2021 to 2023?Answer: $2 million

• Loss: MultipleNegativesRankingLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

• num_train_epochs: 2
• multi_dataset_batch_sampler: round_robin

All Hyperparameters

• do_predict: False
• eval_strategy: no
• prediction_loss_only: True
• per_device_train_batch_size: 8
• per_device_eval_batch_size: 8
• 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
• num_train_epochs: 2
• max_steps: -1
• lr_scheduler_type: linear
• lr_scheduler_kwargs: None
• warmup_ratio: None
• warmup_steps: 0
• log_level: passive
• log_level_replica: warning
• log_on_each_node: True
• logging_nan_inf_filter: True
• enable_jit_checkpoint: False
• save_on_each_node: False
• save_only_model: False
• restore_callback_states_from_checkpoint: False
• use_cpu: False
• seed: 42
• data_seed: None
• bf16: False
• fp16: False
• bf16_full_eval: False
• fp16_full_eval: False
• tf32: None
• local_rank: -1
• ddp_backend: None
• debug: []
• dataloader_drop_last: False
• dataloader_num_workers: 0
• dataloader_prefetch_factor: None
• disable_tqdm: False
• remove_unused_columns: True
• label_names: None
• load_best_model_at_end: False
• ignore_data_skip: False
• fsdp: []
• fsdp_config: {'min_num_params': 0, 'xla': False, 'xla_fsdp_v2': False, 'xla_fsdp_grad_ckpt': 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
• deepspeed: None
• label_smoothing_factor: 0.0
• optim: adamw_torch
• optim_args: None
• group_by_length: False
• length_column_name: length
• project: huggingface
• trackio_space_id: trackio
• 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
• 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_for_metrics: []
• eval_do_concat_batches: True
• auto_find_batch_size: False
• full_determinism: False
• ddp_timeout: 1800
• torch_compile: False
• torch_compile_backend: None
• torch_compile_mode: None
• include_num_input_tokens_seen: no
• 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: True
• use_cache: False
• prompts: None
• batch_sampler: batch_sampler
• multi_dataset_batch_sampler: round_robin
• router_mapping: {}
• learning_rate_mapping: {}

Framework Versions

• Python: 3.12.10
• Sentence Transformers: 5.2.2
• Transformers: 5.0.0
• PyTorch: 2.7.0+cu128
• Accelerate: 1.12.0
• Datasets: 4.5.0
• 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",
}

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}
}