Shuu12121 / CodeSearch-ModernBERT-Owl-2.2

SentenceTransformer based on Shuu12121/CodeModernBERT-Owl-2.2-Pre

This is a sentence-transformers model finetuned from Shuu12121/CodeModernBERT-Owl-2.2-Pre . 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: Shuu12121/CodeModernBERT-Owl-2.2-Pre
• Maximum Sequence Length: 1024 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': 1024, 'do_lower_case': False}) with Transformer model: ModernBertModel 
  (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})
)

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 = [
    '<pre>\nField extraction metadata on the property.\n</pre>\n\n<code>.google.cloud.documentai.v1beta3.FieldExtractionMetadata field_extraction_metadata = 9;\n</code>\n\n@return Whether the fieldExtractionMetadata field is set.',
    '@java.lang.Override\n  public boolean hasFieldExtractionMetadata() {\n    return ((bitField0_ & 0x00000001) != 0);\n  }',
    'pub fn poller(self) -> impl lro::Poller<(), crate::model::DeleteSitemapMetadata> {\n            type Operation =\n                lro::internal::Operation<wkt::Empty, crate::model::DeleteSitemapMetadata>;\n            let polling_error_policy = self.0.stub.get_polling_error_policy(&self.0.options);\n            let polling_backoff_policy = self.0.stub.get_polling_backoff_policy(&self.0.options);\n\n            let stub = self.0.stub.clone();\n            let mut options = self.0.options.clone();\n            options.set_retry_policy(gax::retry_policy::NeverRetry);\n            let query = move |name| {\n                let stub = stub.clone();\n                let options = options.clone();\n                async {\n                    let op = GetOperation::new(stub)\n                        .set_name(name)\n                        .with_options(options)\n                        .send()\n                        .await?;\n                    Ok(Operation::new(op))\n                }\n            };\n\n            let start = move || async {\n                let op = self.send().await?;\n                Ok(Operation::new(op))\n            };\n\n            lro::internal::new_unit_response_poller(\n                polling_error_policy,\n                polling_backoff_policy,\n                start,\n                query,\n            )\n        }',
]
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

Unnamed Dataset

• Size: 2,732,400 training samples
• Columns: sentence_0 , sentence_1 , and label
• Approximate statistics based on the first 1000 samples:

  	sentence_0	sentence_1	label
  type	string	string	float
  details	min: 8 tokens mean: 70.49 tokens max: 1024 tokens	min: 5 tokens mean: 138.25 tokens max: 1024 tokens	min: 1.0 mean: 1.0 max: 1.0

• Samples:

  sentence_0	sentence_1	label
  Prints the specified pkg . If is_main is not set, nested package notation is used.	pub fn print_package( &mut self, resolve: &Resolve, pkg: PackageId, is_main: bool, ) -> Result<()> { let pkg = &resolve.packages[pkg]; self.print_package_outer(pkg)?; if is_main { self.output.semicolon(); self.output.newline(); } else { self.output.indent_start(); } for (name, id) in pkg.interfaces.iter() { self.print_interface_outer(resolve, *id, name)?; self.output.indent_start(); self.print_interface(resolve, *id)?; self.output.indent_end(); if is_main { self.output.newline(); } } for (name, id) in pkg.worlds.iter() { self.print_docs(&resolve.worlds[*id].docs); self.print_stability(&resolve.worlds[*id].stability); self.output.keyword("world"); self.output.str(" "); self.print_name_type(name, TypeKind:...	1.0
  An alternative descriptive name for the user.	pub fn nick_name(mut self, input: impl ::std::convert::Into<::std::string::String>) -> Self { self.nick_name = ::std::option::Option::Some(input.into()); self }	1.0
  Indicates whether the match is case sensitive.	pub fn case_sensitive(mut self, input: bool) -> Self { self.case_sensitive = ::std::option::Option::Some(input); self }	1.0

• Loss: MultipleNegativesRankingLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

• per_device_train_batch_size : 150
• per_device_eval_batch_size : 150
• num_train_epochs : 5
• fp16 : True
• multi_dataset_batch_sampler : round_robin

All Hyperparameters

Training Logs

Framework Versions

• Python: 3.11.13
• Sentence Transformers: 4.1.0
• Transformers: 4.52.4
• PyTorch: 2.6.0+cu124
• Accelerate: 1.7.0
• Datasets: 3.6.0
• 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",
}

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}
}