README.md

metadata

language:
  - en
license: apache-2.0
tags:
  - sentence-transformers
  - sentence-similarity
  - feature-extraction
  - dense
  - generated_from_trainer
  - dataset_size:900
  - loss:MatryoshkaLoss
  - loss:MultipleNegativesRankingLoss
base_model: microsoft/codebert-base
widget:
  - source_sentence: Best practices for _invocation_params
    sentences:
      - >-
        def after_model(self, state: StateT, runtime: Runtime[ContextT]) ->
        dict[str, Any] | None:
                """Logic to run after the model is called.

                Args:
                    state: The current agent state.
                    runtime: The runtime context.

                Returns:
                    Agent state updates to apply after model call.
                """
      - |-
        def _get_trace_callbacks(
            project_name: str | None = None,
            example_id: str | UUID | None = None,
            callback_manager: CallbackManager | AsyncCallbackManager | None = None,
        ) -> Callbacks:
            if _tracing_v2_is_enabled():
                project_name_ = project_name or _get_tracer_project()
                tracer = tracing_v2_callback_var.get() or LangChainTracer(
                    project_name=project_name_,
                    example_id=example_id,
                )
                if callback_manager is None:
                    cb = cast("Callbacks", [tracer])
                else:
                    if not any(
                        isinstance(handler, LangChainTracer)
                        for handler in callback_manager.handlers
                    ):
                        callback_manager.add_handler(tracer)
                        # If it already has a LangChainTracer, we don't need to add another one.
                        # this would likely mess up the trace hierarchy.
                    cb = callback_manager
            else:
                cb = None
            return cb
      - |-
        def _invocation_params(self) -> dict[str, Any]:
                params: dict = {"model": self.model, **self.model_kwargs}
                if self.dimensions is not None:
                    params["dimensions"] = self.dimensions
                return params
  - source_sentence: How does _approximate_token_counter work in Python?
    sentences:
      - |-
        def _approximate_token_counter(messages: Sequence[BaseMessage]) -> int:
            """Wrapper for `count_tokens_approximately` that matches expected signature."""
            return count_tokens_approximately(messages)
      - |-
        def remove_request_headers(request: Any) -> Any:
            for k in request.headers:
                request.headers[k] = "**REDACTED**"
            return request
      - |-
        def get_format_instructions(self) -> str:
                """Returns formatting instructions for the given output parser."""
                return self.format_instructions
  - source_sentence: How to implement _create_thread_and_run?
    sentences:
      - |-
        async def on_retriever_end(
                self, documents: Sequence[Document], **kwargs: Any
            ) -> None:
                """Run when the retriever ends running.

                Args:
                    documents: The retrieved documents.
                    **kwargs: Additional keyword arguments.

                """
                if not self.handlers:
                    return
                await ahandle_event(
                    self.handlers,
                    "on_retriever_end",
                    "ignore_retriever",
                    documents,
                    run_id=self.run_id,
                    parent_run_id=self.parent_run_id,
                    tags=self.tags,
                    **kwargs,
                )
      - |-
        def _create_thread_and_run(self, input_dict: dict, thread: dict) -> Any:
                params = {
                    k: v
                    for k, v in input_dict.items()
                    if k
                    in (
                        "instructions",
                        "model",
                        "tools",
                        "parallel_tool_calls",
                        "top_p",
                        "temperature",
                        "max_completion_tokens",
                        "max_prompt_tokens",
                        "run_metadata",
                    )
                }
                return self.client.beta.threads.create_and_run(
                    assistant_id=self.assistant_id,
                    thread=thread,
                    **params,
                )
      - |-
        def test_pandas_output_parser_col_no_array() -> None:
            with pytest.raises(OutputParserException):
                parser.parse("column:num_legs")
  - source_sentence: Explain the get_token_ids logic
    sentences:
      - |-
        def _runnable(inputs: dict[str, Any]) -> str:
            if inputs["text"] == "foo":
                return "first"
            if "exception" not in inputs:
                msg = "missing exception"
                raise ValueError(msg)
            if inputs["text"] == "bar":
                return "second"
            if isinstance(inputs["exception"], ValueError):
                raise RuntimeError  # noqa: TRY004
            return "third"
      - |-
        def validate_params(cls, values: dict) -> dict:
                """Validate similarity parameters."""
                if values["k"] is None and values["similarity_threshold"] is None:
                    msg = "Must specify one of `k` or `similarity_threshold`."
                    raise ValueError(msg)
                return values
      - |-
        def get_token_ids(self, text: str) -> list[int]:
                """Return the ordered IDs of the tokens in a text.

                Args:
                    text: The string input to tokenize.

                Returns:
                    A list of IDs corresponding to the tokens in the text, in order they occur
                        in the text.
                """
                if self.custom_get_token_ids is not None:
                    return self.custom_get_token_ids(text)
                return _get_token_ids_default_method(text)
  - source_sentence: How does __init__ work in Python?
    sentences:
      - |-
        def test_loading_few_shot_prompt_from_json() -> None:
            """Test loading few shot prompt from json."""
            with change_directory(EXAMPLE_DIR):
                prompt = load_prompt("few_shot_prompt.json")
                expected_prompt = FewShotPromptTemplate(
                    input_variables=["adjective"],
                    prefix="Write antonyms for the following words.",
                    example_prompt=PromptTemplate(
                        input_variables=["input", "output"],
                        template="Input: {input}\nOutput: {output}",
                    ),
                    examples=[
                        {"input": "happy", "output": "sad"},
                        {"input": "tall", "output": "short"},
                    ],
                    suffix="Input: {adjective}\nOutput:",
                )
                assert prompt == expected_prompt
      - |-
        def __init__(
                self,
                encoding_name: str = "gpt2",
                model_name: str | None = None,
                allowed_special: Literal["all"] | AbstractSet[str] = set(),
                disallowed_special: Literal["all"] | Collection[str] = "all",
                **kwargs: Any,
            ) -> None:
                """Create a new `TextSplitter`.

                Args:
                    encoding_name: The name of the tiktoken encoding to use.
                    model_name: The name of the model to use. If provided, this will
                        override the `encoding_name`.
                    allowed_special: Special tokens that are allowed during encoding.
                    disallowed_special: Special tokens that are disallowed during encoding.

                Raises:
                    ImportError: If the tiktoken package is not installed.
                """
                super().__init__(**kwargs)
                if not _HAS_TIKTOKEN:
                    msg = (
                        "Could not import tiktoken python package. "
                        "This is needed in order to for TokenTextSplitter. "
                        "Please install it with `pip install tiktoken`."
                    )
                    raise ImportError(msg)

                if model_name is not None:
                    enc = tiktoken.encoding_for_model(model_name)
                else:
                    enc = tiktoken.get_encoding(encoding_name)
                self._tokenizer = enc
                self._allowed_special = allowed_special
                self._disallowed_special = disallowed_special
      - |-
        def test_fixed_message_response_when_docs_found() -> None:
            fixed_resp = "I don't know"
            answer = "I know the answer!"
            llm = FakeListLLM(responses=[answer])
            retriever = SequentialRetriever(
                sequential_responses=[[Document(page_content=answer)]],
            )
            memory = ConversationBufferMemory(
                k=1,
                output_key="answer",
                memory_key="chat_history",
                return_messages=True,
            )
            qa_chain = ConversationalRetrievalChain.from_llm(
                llm=llm,
                memory=memory,
                retriever=retriever,
                return_source_documents=True,
                rephrase_question=False,
                response_if_no_docs_found=fixed_resp,
                verbose=True,
            )
            got = qa_chain("What is the answer?")
            assert got["chat_history"][1].content == answer
            assert got["answer"] == answer
pipeline_tag: sentence-similarity
library_name: sentence-transformers
metrics:
  - cosine_accuracy@1
  - cosine_accuracy@3
  - cosine_accuracy@5
  - cosine_accuracy@10
  - cosine_precision@1
  - cosine_precision@3
  - cosine_precision@5
  - cosine_precision@10
  - cosine_recall@1
  - cosine_recall@3
  - cosine_recall@5
  - cosine_recall@10
  - cosine_ndcg@10
  - cosine_mrr@10
  - cosine_map@100
model-index:
  - name: codeBert Base
    results:
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 768
          type: dim_768
        metrics:
          - type: cosine_accuracy@1
            value: 0.83
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.85
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.86
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.94
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.83
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.83
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.83
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.45299999999999996
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.16599999999999998
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.498
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.83
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.9059999999999999
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.8712089918828809
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.8532738095238095
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.861635686929646
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 512
          type: dim_512
        metrics:
          - type: cosine_accuracy@1
            value: 0.85
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.86
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.87
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.95
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.85
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.84
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.8419999999999999
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.45299999999999996
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.16999999999999996
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.504
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.8419999999999999
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.9059999999999999
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.8775797199885595
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.8699404761904762
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.8692738075020783
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 256
          type: dim_256
        metrics:
          - type: cosine_accuracy@1
            value: 0.86
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.89
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.9
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.93
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.86
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.85
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.85
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.45
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.17199999999999996
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.51
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.85
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.9
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.8789938349894767
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.8805952380952381
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.8726611807317667
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 128
          type: dim_128
        metrics:
          - type: cosine_accuracy@1
            value: 0.84
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.87
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.88
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.93
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.84
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.8366666666666667
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.8419999999999999
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.455
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.16799999999999998
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.502
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.8419999999999999
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.91
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.8777095006932575
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.8630000000000001
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.8661619081282643
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 64
          type: dim_64
        metrics:
          - type: cosine_accuracy@1
            value: 0.78
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.81
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.81
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.93
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.78
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.7866666666666667
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.7859999999999999
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.44799999999999995
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.15599999999999997
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.472
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.7859999999999999
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.8959999999999999
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.8445404597381452
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.8120634920634922
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.8308457034802883
            name: Cosine Map@100

codeBert Base

This is a sentence-transformers model finetuned from microsoft/codebert-base. 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: microsoft/codebert-base
• Maximum Sequence Length: 512 tokens
• Output Dimensionality: 768 dimensions
• Similarity Function: Cosine Similarity
• Language: en
• License: apache-2.0

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': False, 'architecture': 'RobertaModel'})
  (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("killdollar/codebert-embed-base-dense-retriever")
# Run inference
sentences = [
    'How does __init__ work in Python?',
    'def __init__(\n        self,\n        encoding_name: str = "gpt2",\n        model_name: str | None = None,\n        allowed_special: Literal["all"] | AbstractSet[str] = set(),\n        disallowed_special: Literal["all"] | Collection[str] = "all",\n        **kwargs: Any,\n    ) -> None:\n        """Create a new `TextSplitter`.\n\n        Args:\n            encoding_name: The name of the tiktoken encoding to use.\n            model_name: The name of the model to use. If provided, this will\n                override the `encoding_name`.\n            allowed_special: Special tokens that are allowed during encoding.\n            disallowed_special: Special tokens that are disallowed during encoding.\n\n        Raises:\n            ImportError: If the tiktoken package is not installed.\n        """\n        super().__init__(**kwargs)\n        if not _HAS_TIKTOKEN:\n            msg = (\n                "Could not import tiktoken python package. "\n                "This is needed in order to for TokenTextSplitter. "\n                "Please install it with `pip install tiktoken`."\n            )\n            raise ImportError(msg)\n\n        if model_name is not None:\n            enc = tiktoken.encoding_for_model(model_name)\n        else:\n            enc = tiktoken.get_encoding(encoding_name)\n        self._tokenizer = enc\n        self._allowed_special = allowed_special\n        self._disallowed_special = disallowed_special',
    'def test_fixed_message_response_when_docs_found() -> None:\n    fixed_resp = "I don\'t know"\n    answer = "I know the answer!"\n    llm = FakeListLLM(responses=[answer])\n    retriever = SequentialRetriever(\n        sequential_responses=[[Document(page_content=answer)]],\n    )\n    memory = ConversationBufferMemory(\n        k=1,\n        output_key="answer",\n        memory_key="chat_history",\n        return_messages=True,\n    )\n    qa_chain = ConversationalRetrievalChain.from_llm(\n        llm=llm,\n        memory=memory,\n        retriever=retriever,\n        return_source_documents=True,\n        rephrase_question=False,\n        response_if_no_docs_found=fixed_resp,\n        verbose=True,\n    )\n    got = qa_chain("What is the answer?")\n    assert got["chat_history"][1].content == answer\n    assert got["answer"] == answer',
]
embeddings = model.encode(sentences)
print(embeddings.shape)
# [3, 768]

# Get the similarity scores for the embeddings
similarities = model.similarity(embeddings, embeddings)
print(similarities)
# tensor([[1.0000, 0.7336, 0.0979],
#         [0.7336, 1.0000, 0.1742],
#         [0.0979, 0.1742, 1.0000]])

Evaluation

Metrics

Information Retrieval

• Dataset: dim_768
• Evaluated with InformationRetrievalEvaluator with these parameters:

  {
      "truncate_dim": 768
  }
  

Metric	Value
cosine_accuracy@1	0.83
cosine_accuracy@3	0.85
cosine_accuracy@5	0.86
cosine_accuracy@10	0.94
cosine_precision@1	0.83
cosine_precision@3	0.83
cosine_precision@5	0.83
cosine_precision@10	0.453
cosine_recall@1	0.166
cosine_recall@3	0.498
cosine_recall@5	0.83
cosine_recall@10	0.906
cosine_ndcg@10	0.8712
cosine_mrr@10	0.8533
cosine_map@100	0.8616

Information Retrieval

• Dataset: dim_512
• Evaluated with InformationRetrievalEvaluator with these parameters:

  {
      "truncate_dim": 512
  }
  

Metric	Value
cosine_accuracy@1	0.85
cosine_accuracy@3	0.86
cosine_accuracy@5	0.87
cosine_accuracy@10	0.95
cosine_precision@1	0.85
cosine_precision@3	0.84
cosine_precision@5	0.842
cosine_precision@10	0.453
cosine_recall@1	0.17
cosine_recall@3	0.504
cosine_recall@5	0.842
cosine_recall@10	0.906
cosine_ndcg@10	0.8776
cosine_mrr@10	0.8699
cosine_map@100	0.8693

Information Retrieval

• Dataset: dim_256
• Evaluated with InformationRetrievalEvaluator with these parameters:

  {
      "truncate_dim": 256
  }
  

Metric	Value
cosine_accuracy@1	0.86
cosine_accuracy@3	0.89
cosine_accuracy@5	0.9
cosine_accuracy@10	0.93
cosine_precision@1	0.86
cosine_precision@3	0.85
cosine_precision@5	0.85
cosine_precision@10	0.45
cosine_recall@1	0.172
cosine_recall@3	0.51
cosine_recall@5	0.85
cosine_recall@10	0.9
cosine_ndcg@10	0.879
cosine_mrr@10	0.8806
cosine_map@100	0.8727

Information Retrieval

• Dataset: dim_128
• Evaluated with InformationRetrievalEvaluator with these parameters:

  {
      "truncate_dim": 128
  }
  

Metric	Value
cosine_accuracy@1	0.84
cosine_accuracy@3	0.87
cosine_accuracy@5	0.88
cosine_accuracy@10	0.93
cosine_precision@1	0.84
cosine_precision@3	0.8367
cosine_precision@5	0.842
cosine_precision@10	0.455
cosine_recall@1	0.168
cosine_recall@3	0.502
cosine_recall@5	0.842
cosine_recall@10	0.91
cosine_ndcg@10	0.8777
cosine_mrr@10	0.863
cosine_map@100	0.8662

Information Retrieval

• Dataset: dim_64
• Evaluated with InformationRetrievalEvaluator with these parameters:

  {
      "truncate_dim": 64
  }
  

Metric	Value
cosine_accuracy@1	0.78
cosine_accuracy@3	0.81
cosine_accuracy@5	0.81
cosine_accuracy@10	0.93
cosine_precision@1	0.78
cosine_precision@3	0.7867
cosine_precision@5	0.786
cosine_precision@10	0.448
cosine_recall@1	0.156
cosine_recall@3	0.472
cosine_recall@5	0.786
cosine_recall@10	0.896
cosine_ndcg@10	0.8445
cosine_mrr@10	0.8121
cosine_map@100	0.8308

Training Details

Training Dataset

Unnamed Dataset

• Size: 900 training samples
• Columns: anchor and positive
• Approximate statistics based on the first 900 samples:

  	anchor	positive
  type	string	string
  details	min: 6 tokens mean: 13.15 tokens max: 42 tokens	min: 25 tokens mean: 239.87 tokens max: 512 tokens

• Samples:

  anchor	positive
  Explain the test_qdrant_similarity_search_with_relevance_scores logic	def test_qdrant_similarity_search_with_relevance_scores( batch_size: int, content_payload_key: str, metadata_payload_key: str, vector_name: str | None, ) -> None: """Test end to end construction and search.""" texts = ["foo", "bar", "baz"] docsearch = Qdrant.from_texts( texts, ConsistentFakeEmbeddings(), location=":memory:", content_payload_key=content_payload_key, metadata_payload_key=metadata_payload_key, batch_size=batch_size, vector_name=vector_name, ) output = docsearch.similarity_search_with_relevance_scores("foo", k=3) assert all( (score <= 1 or np.isclose(score, 1)) and score >= 0 for _, score in output )
  How to implement LangChainPendingDeprecationWarning?	class LangChainPendingDeprecationWarning(PendingDeprecationWarning): """A class for issuing deprecation warnings for LangChain users."""
  Example usage of random_name	def random_name() -> str: """Generate a random name.""" adjective = random.choice(adjectives) # noqa: S311 noun = random.choice(nouns) # noqa: S311 number = random.randint(1, 100) # noqa: S311 return f"{adjective}-{noun}-{number}"

• Loss: MatryoshkaLoss with these parameters:

  {
      "loss": "MultipleNegativesRankingLoss",
      "matryoshka_dims": [
          768,
          512,
          256,
          128,
          64
      ],
      "matryoshka_weights": [
          1,
          1,
          1,
          1,
          1
      ],
      "n_dims_per_step": -1
  }
  

Training Hyperparameters

Non-Default Hyperparameters

• eval_strategy: epoch
• per_device_train_batch_size: 4
• per_device_eval_batch_size: 4
• gradient_accumulation_steps: 16
• learning_rate: 2e-05
• num_train_epochs: 4
• lr_scheduler_type: cosine
• warmup_ratio: 0.1
• fp16: True
• load_best_model_at_end: True
• optim: adamw_torch
• batch_sampler: no_duplicates

All Hyperparameters

Click to expand

• overwrite_output_dir: False
• do_predict: False
• eval_strategy: epoch
• prediction_loss_only: True
• per_device_train_batch_size: 4
• per_device_eval_batch_size: 4
• per_gpu_train_batch_size: None
• per_gpu_eval_batch_size: None
• gradient_accumulation_steps: 16
• 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: 4
• max_steps: -1
• lr_scheduler_type: cosine
• 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
• bf16: False
• fp16: True
• 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: True
• 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}
• 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}
• parallelism_config: None
• deepspeed: None
• label_smoothing_factor: 0.0
• optim: adamw_torch
• optim_args: None
• adafactor: False
• 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
• 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
• hub_revision: None
• 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: 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
• prompts: None
• batch_sampler: no_duplicates
• multi_dataset_batch_sampler: proportional
• router_mapping: {}
• learning_rate_mapping: {}

Training Logs

Epoch	Step	Training Loss	dim_768_cosine_ndcg@10	dim_512_cosine_ndcg@10	dim_256_cosine_ndcg@10	dim_128_cosine_ndcg@10	dim_64_cosine_ndcg@10
0.7111	10	6.8447	-	-	-	-	-
1.0	15	-	0.1025	0.0367	0.0548	0.0502	0.1185
0.7111	10	4.8545	-	-	-	-	-
1.0	15	-	0.2250	0.3047	0.2895	0.2892	0.3178
0.7111	10	1.9011	-	-	-	-	-
1.0	15	-	0.6530	0.6393	0.6269	0.6631	0.6658
1.3556	20	0.6349	-	-	-	-	-
2.0	30	0.1887	0.8480	0.8643	0.8641	0.8532	0.7974
2.7111	40	0.0959	-	-	-	-	-
3.0	45	-	0.8688	0.8774	0.8754	0.8725	0.8457
3.3556	50	0.0359	-	-	-	-	-
4.0	60	0.0515	0.8712	0.8776	0.879	0.8777	0.8445

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.12.12
• Sentence Transformers: 5.2.0
• Transformers: 4.57.3
• PyTorch: 2.9.0+cu126
• Accelerate: 1.12.0
• Datasets: 4.0.0
• 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",
}

MatryoshkaLoss

@misc{kusupati2024matryoshka,
    title={Matryoshka Representation Learning},
    author={Aditya Kusupati and Gantavya Bhatt and Aniket Rege and Matthew Wallingford and Aditya Sinha and Vivek Ramanujan and William Howard-Snyder and Kaifeng Chen and Sham Kakade and Prateek Jain and Ali Farhadi},
    year={2024},
    eprint={2205.13147},
    archivePrefix={arXiv},
    primaryClass={cs.LG}
}

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}
}