Shuu12121 / CodeSearch-ModernBERT-Owl-ALT

SentenceTransformer based on Shuu12121/CodeModernBERT-Owl

This is a sentence-transformers model finetuned from Shuu12121/CodeModernBERT-Owl . 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
• Maximum Sequence Length: 2048 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': 2048, 'do_lower_case': False}) with Transformer model: 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 = [
    'Write combining tables to filesystem as python code.',
    'def _do_write(fname, variable, version, date, table):\n        \n        # pylint: disable=R0914\n        #         Too many local variables (19/15) (col 4)\n        print("writing {} ..".format(fname))\n        import unicodedata\n        import datetime\n        import string\n        utc_now = datetime.datetime.utcnow()\n        indent = 4\n        with open(fname, \'w\') as fout:\n            fout.write(\n                \'\\n\'\n                "# Generated: {iso_utc}\\n"\n                "# Source: {version}\\n"\n                "# Date: {date}\\n"\n                "{variable} = (".format(iso_utc=utc_now.isoformat(),\n                                        version=version,\n                                        date=date,\n                                        variable=variable,\n                                        variable_proper=variable.title()))\n            for start, end in table:\n                ucs_start, ucs_end = unichr(start), unichr(end)\n                hex_start, hex_end = (\'0x{0:04x}\'.format(start),\n                                      \'0x{0:04x}\'.format(end))\n                try:\n                    name_start = string.capwords(unicodedata.name(ucs_start))\n                except ValueError:\n                    name_start = u\'\'\n                try:\n                    name_end = string.capwords(unicodedata.name(ucs_end))\n                except ValueError:\n                    name_end = u\'\'\n                fout.write(\'\\n\' + (\' \' * indent))\n                fout.write(\'({0}, {1},),\'.format(hex_start, hex_end))\n                fout.write(\'  # {0:24s}..{1}\'.format(\n                    name_start[:24].rstrip() or \'(nil)\',\n                    name_end[:24].rstrip()))\n            fout.write(\'\\n)\\n\')\n        print("complete.")',
    'def RGB_to_HSV(cobj, *args, **kwargs):\n    \n    var_R = cobj.rgb_r\n    var_G = cobj.rgb_g\n    var_B = cobj.rgb_b\n\n    var_max = max(var_R, var_G, var_B)\n    var_min = min(var_R, var_G, var_B)\n\n    var_H = __RGB_to_Hue(var_R, var_G, var_B, var_min, var_max)\n\n    if var_max == 0:\n        var_S = 0\n    else:\n        var_S = 1.0 - (var_min / var_max)\n\n    var_V = var_max\n\n    return HSVColor(\n        var_H, var_S, var_V)',
]
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]

Evaluation

Metrics

Information Retrieval

• Dataset: code-docstring-retrieval
• Evaluated with InformationRetrievalEvaluator

Training Details

Training Dataset

Unnamed Dataset

• Size: 1,455,632 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: 5 tokens mean: 96.39 tokens max: 2048 tokens	min: 27 tokens mean: 159.11 tokens max: 889 tokens	min: 1.0 mean: 1.0 max: 1.0

• Samples:

  sentence_0	sentence_1	label
  Set the text for this element. Arguments: text (str): The text cls (str): The class of the text, defaults to current (leave this unless you know what you are doing). There may be only one text content element of each class associated with the element.	def settext(self, text, cls='current'): self.replace(TextContent, value=text, cls=cls)	1.0
  Associate a document with this element. Arguments: doc (:class: Document ): A document Each element must be associated with a FoLiA document.	def setdocument(self, doc): assert isinstance(doc, Document) if not self.doc: self.doc = doc if self.id: if self.id in doc: raise DuplicateIDError(self.id) else: self.doc.index[id] = self for e in self: #recursive for all children if isinstance(e,AbstractElement): e.setdocument(doc)	1.0
  Tests whether a new element of this class can be added to the parent. This method is mostly for internal use. This will use the OCCURRENCES property, but may be overidden by subclasses for more customised behaviour. Parameters: parent (:class: AbstractElement ): The element that is being added to set (str or None): The set raiseexceptions (bool): Raise an exception if the element can't be added? Returns: bool Raises: ValueError	def addable(Class, parent, set=None, raiseexceptions=True): if not parent. class .accepts(Class, raiseexceptions, parent): return False if Class.OCCURRENCES > 0: #check if the parent doesn't have too many already count = parent.count(Class,None,True,[True, AbstractStructureElement]) #never descend into embedded structure annotatioton if count >= Class.OCCURRENCES: if raiseexceptions: if parent.id: extra = ' (id=' + parent.id + ')' else: extra = '' raise DuplicateAnnotationError("Unable to add another object of type " + Class. name + " to " + parent. class . name + " " + extra + ". There are already " + str(count) + " instances of this class, which is the maximum.") else: return False if Class.OCCURRENCES_PER_SET > 0 and set and Class.R...	1.0

• Loss: MultipleNegativesRankingLoss with these parameters:

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

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy : steps
• per_device_train_batch_size : 24
• per_device_eval_batch_size : 24
• num_train_epochs : 5
• fp16 : True
• multi_dataset_batch_sampler : round_robin

All Hyperparameters

Training Logs

Framework Versions

• Python: 3.11.11
• Sentence Transformers: 3.4.1
• Transformers: 4.50.3
• PyTorch: 2.6.0+cu124
• Accelerate: 1.5.2
• 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",
}

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}
}