SentenceTransformer based on capemox/ettin-encoder-17m-pretrain-50m

This is a sentence-transformers model finetuned from capemox/ettin-encoder-17m-pretrain-50m on the msmarco-embed-dense_on-triples-v3 dataset. It maps sentences & paragraphs to a 768-dimensional dense vector space and can be used for retrieval.

Model Details

Model Description

• Model Type: Sentence Transformer
• Base model: capemox/ettin-encoder-17m-pretrain-50m
• Maximum Sequence Length: 8192 tokens
• Output Dimensionality: 768 dimensions
• Similarity Function: Cosine Similarity
• Supported Modality: Text
• Training Dataset:
  • msmarco-embed-dense_on-triples-v3

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': 'ModernBertModel'})
  (1): Pooling({'embedding_dimension': 256, 'pooling_mode': 'cls', 'include_prompt': True})
  (2): Dense({'in_features': 256, 'out_features': 768, 'bias': False, 'activation_function': 'torch.nn.modules.linear.Identity', 'module_input_name': 'sentence_embedding', 'module_output_name': 'sentence_embedding'})
  (3): 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
queries = [
    'define reservedness',
]
documents = [
    'Definition of reservedness in the English dictionary The definition of reservedness in the dictionary is the quality or state of being cool or formal in manner or of being restrained, silent, or reticent. english words that begin like reservedness reserve tank',
    "Princeton's WordNet(0.00 / 0 votes)Rate this definition: 1  reserved(adj) set aside for the use of a particular person or party. 2  reserved(adj) marked by self-restraint and reticence.",
    'In a computer language, a reserved word is a word that cannot be used as an identifier, such as the name of a variable, function, or label â\x80\x93 it is reserved from use. This is a syntactic definition, and a reserved word may have no meaning. A closely related and often conflated notion is a keyword, which is a word with special meaning in a particular context. This is a semantic definition. By contrast, names in a standard library but not built into the language are not considered reserved ...',
]
query_embeddings = model.encode_query(queries)
document_embeddings = model.encode_document(documents)
print(query_embeddings.shape, document_embeddings.shape)
# [1, 768] [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(query_embeddings, document_embeddings)
print(similarities)
# tensor([[0.5320, 0.4636, 0.3707]])

Evaluation

Metrics

Information Retrieval

• Datasets: NanoClimateFEVER, NanoDBPedia, NanoFEVER, NanoFiQA2018, NanoHotpotQA, NanoMSMARCO, NanoNFCorpus, NanoNQ, NanoQuoraRetrieval, NanoSCIDOCS, NanoArguAna, NanoSciFact and NanoTouche2020
• Evaluated with InformationRetrievalEvaluator with these parameters:

  {
    "query_prompt": "query: ",
    "corpus_prompt": "document: "
  }
  

Nano BEIR

• Dataset: NanoBEIR_mean
• Evaluated with NanoBEIREvaluator with these parameters:

  {
    "dataset_names": [
      "climatefever",
      "dbpedia",
      "fever",
      "fiqa2018",
      "hotpotqa",
      "msmarco",
      "nfcorpus",
      "nq",
      "quoraretrieval",
      "scidocs",
      "arguana",
      "scifact",
      "touche2020"
    ],
    "dataset_id": "sentence-transformers/NanoBEIR-en",
    "query_prompts": {
      "climatefever": "query: ",
      "dbpedia": "query: ",
      "fever": "query: ",
      "fiqa2018": "query: ",
      "hotpotqa": "query: ",
      "msmarco": "query: ",
      "nfcorpus": "query: ",
      "nq": "query: ",
      "quoraretrieval": "query: ",
      "scidocs": "query: ",
      "arguana": "query: ",
      "scifact": "query: ",
      "touche2020": "query: "
    },
    "corpus_prompts": {
      "climatefever": "document: ",
      "dbpedia": "document: ",
      "fever": "document: ",
      "fiqa2018": "document: ",
      "hotpotqa": "document: ",
      "msmarco": "document: ",
      "nfcorpus": "document: ",
      "nq": "document: ",
      "quoraretrieval": "document: ",
      "scidocs": "document: ",
      "arguana": "document: ",
      "scifact": "document: ",
      "touche2020": "document: "
    }
  }
  

Training Details

Training Dataset

msmarco-embed-dense_on-triples-v3

• Dataset: msmarco-embed-dense_on-triples-v3
• Size: 483,497 training samples
• Columns: query, positive, negative_1, negative_2, negative_3, negative_4, negative_5, negative_6, negative_7, negative_8, and label |
• Loss: EmbedDistillLoss with these parameters:

  {
    "distance_metric": "l2",
    "projection_dim": null
  }
  

Evaluation Dataset

msmarco-embed-dense_on-triples-v3

• Dataset: msmarco-embed-dense_on-triples-v3

• Size: 10,000 evaluation samples

• Loss: EmbedDistillLoss with these parameters:

  {
    "distance_metric": "l2",
    "projection_dim": null
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size: 128
• num_train_epochs: 4
• max_steps: 15000
• learning_rate: 0.0001
• lr_scheduler_type: constant
• weight_decay: 0.01
• bf16: True
• per_device_eval_batch_size: 16
• seed: 12
• dataloader_num_workers: 2
• remove_unused_columns: False

All Hyperparameters

• per_device_train_batch_size: 128
• num_train_epochs: 4
• max_steps: 15000
• learning_rate: 0.0001
• lr_scheduler_type: constant
• lr_scheduler_kwargs: None
• warmup_steps: 0
• optim: adamw_torch_fused
• optim_args: None
• weight_decay: 0.01
• 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: 16
• 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: 12
• 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: 2
• dataloader_pin_memory: True
• dataloader_persistent_workers: False
• dataloader_prefetch_factor: None
• remove_unused_columns: False
• 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 Time

• Training: 3.3 hours
• Evaluation: 32.0 minutes
• Total: 3.9 hours

Framework Versions

• Python: 3.11.12
• Sentence Transformers: 5.6.1
• Transformers: 5.14.1
• PyTorch: 2.13.0+cu130
• Accelerate: 1.14.0
• Datasets: 5.0.1
• 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",
}

EmbedDistillLoss

@article{kim2023embeddistill,
    title={EmbedDistill: A Geometric Knowledge Distillation for Information Retrieval},
    author={Kim, Seungyeon and Rawat, Ankit Singh and Zaheer, Manzil and Jayasumana, Sadeep and Sadhanala, Veeranjaneyulu and Jitkrittum, Wittawat and Menon, Aditya Krishna and Fergus, Rob and Kumar, Sanjiv},
    year={2023},
    eprint={2301.12005},
    archivePrefix={arXiv},
    primaryClass={cs.IR}
}