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: How to implement __del__?
    sentences:
      - |-
        class SampleMultiCrewFlow(Flow[SimpleState]):
                @start()
                def first_crew(self):
                    """Run first crew."""
                    agent = Agent(
                        role="first agent",
                        goal="first task",
                        backstory="first agent",
                        llm=mock_llm_1,
                    )
                    task = Task(
                        description="First task",
                        expected_output="first result",
                        agent=agent,
                    )
                    crew = Crew(
                        agents=[agent],
                        tasks=[task],
                        share_crew=True,
                    )

                    result = crew.kickoff()

                    assert crew._execution_span is not None
                    return str(result.raw)

                @listen(first_crew)
                def second_crew(self, first_result: str):
                    """Run second crew."""
                    agent = Agent(
                        role="second agent",
                        goal="second task",
                        backstory="second agent",
                        llm=mock_llm_2,
                    )
                    task = Task(
                        description="Second task",
                        expected_output="second result",
                        agent=agent,
                    )
                    crew = Crew(
                        agents=[agent],
                        tasks=[task],
                        share_crew=True,
                    )

                    result = crew.kickoff()

                    assert crew._execution_span is not None

                    self.state.result = f"{first_result} + {result.raw}"
                    return self.state.result
      - |-
        async def test_anthropic_async_with_tools():
            """Test async call with tools."""
            llm = AnthropicCompletion(model="claude-sonnet-4-0")

            tools = [
                {
                    "type": "function",
                    "function": {
                        "name": "get_weather",
                        "description": "Get the current weather for a location",
                        "parameters": {
                            "type": "object",
                            "properties": {
                                "location": {
                                    "type": "string",
                                    "description": "The city and state, e.g. San Francisco, CA"
                                }
                            },
                            "required": ["location"]
                        }
                    }
                }
            ]

            result = await llm.acall(
                "What's the weather in San Francisco?",
                tools=tools
            )
            logging.debug("result: %s", result)

            assert result is not None
            assert isinstance(result, str)
      - |-
        def __del__(self):
                """Cleanup connections on deletion."""
                try:
                    if self._connection_pool:
                        for conn in self._connection_pool:
                            try:
                                conn.close()
                            except Exception:  # noqa: PERF203, S110
                                pass
                    if self._thread_pool:
                        self._thread_pool.shutdown()
                except Exception:  # noqa: S110
                    pass
  - source_sentence: How does route_to_cycle work in Python?
    sentences:
      - |-
        def route_to_cycle(self):
                    execution_log.append("router_initial")
                    return "loop"
      - >-
        def _register_system_event_handlers(self, event_bus: CrewAIEventsBus) ->
        None:
                """Register handlers for system signal events (SIGTERM, SIGINT, etc.)."""

                @on_signal
                def handle_signal(source: Any, event: SignalEvent) -> None:
                    """Flush trace batch on system signals to prevent data loss."""
                    if self.batch_manager.is_batch_initialized():
                        self.batch_manager.finalize_batch()
      - |-
        async def aadd(self) -> None:
                """Add JSON file content asynchronously."""
                content_str = (
                    str(self.content) if isinstance(self.content, dict) else self.content
                )
                new_chunks = self._chunk_text(content_str)
                self.chunks.extend(new_chunks)
                await self._asave_documents()
  - source_sentence: Explain the test_evaluate logic
    sentences:
      - |-
        def test_flow_copy_state_with_unpickleable_objects():
            """Test that _copy_state handles unpickleable objects like RLock.

            Regression test for issue #3828: Flow should not crash when state contains
            objects that cannot be deep copied (like threading.RLock).
            """

            class StateWithRLock(BaseModel):
                counter: int = 0
                lock: Optional[threading.RLock] = None

            class FlowWithRLock(Flow[StateWithRLock]):
                @start()
                def step_1(self):
                    self.state.counter += 1

                @listen(step_1)
                def step_2(self):
                    self.state.counter += 1

            flow = FlowWithRLock(initial_state=StateWithRLock())
            flow._state.lock = threading.RLock()

            copied_state = flow._copy_state()
            assert copied_state.counter == 0
            assert copied_state.lock is not None
      - |-
        def test_evaluate(self, crew_planner):
                task_output = TaskOutput(
                    description="Task 1", agent=str(crew_planner.crew.agents[0])
                )

                with mock.patch.object(Task, "execute_sync") as execute:
                    execute().pydantic = TaskEvaluationPydanticOutput(quality=9.5)
                    crew_planner.evaluate(task_output)
                    assert crew_planner.tasks_scores[0] == [9.5]
      - |-
        class SlowAsyncTool(BaseTool):
                    name: str = "slow_async"
                    description: str = "Simulates slow I/O"

                    def _run(self, task_id: int, delay: float) -> str:
                        return f"Task {task_id} done"

                    async def _arun(self, task_id: int, delay: float) -> str:
                        await asyncio.sleep(delay)
                        return f"Task {task_id} done"
  - source_sentence: Explain the test_clean_action_no_formatting logic
    sentences:
      - |-
        def test_task_interpolation_with_hyphens():
            agent = Agent(
                role="Researcher",
                goal="be an assistant that responds with {interpolation-with-hyphens}",
                backstory="You're an expert researcher, specialized in technology, software engineering, AI and startups. You work as a freelancer and is now working on doing research and analysis for a new customer.",
                allow_delegation=False,
            )
            task = Task(
                description="be an assistant that responds with {interpolation-with-hyphens}",
                expected_output="The response should be addressing: {interpolation-with-hyphens}",
                agent=agent,
            )
            crew = Crew(
                agents=[agent],
                tasks=[task],
                verbose=True,
            )
            result = crew.kickoff(inputs={"interpolation-with-hyphens": "say hello world"})
            assert "say hello world" in task.prompt()

            assert result.raw == "Hello, World!"
      - |-
        class LLMCallCompletedEvent(LLMEventBase):
            """Event emitted when a LLM call completes"""

            type: str = "llm_call_completed"
            messages: str | list[dict[str, Any]] | None = None
            response: Any
            call_type: LLMCallType
            model: str | None = None
      - |-
        def test_clean_action_no_formatting():
            action = "Ask question to senior researcher"
            cleaned_action = parser._clean_action(action)
            assert cleaned_action == "Ask question to senior researcher"
  - source_sentence: Example usage of test_status_code_and_content_type
    sentences:
      - |-
        class NavigateBackToolInput(BaseModel):
            """Input for NavigateBackTool."""

            thread_id: str = Field(
                default="default", description="Thread ID for the browser session"
            )
      - |-
        def test_status_code_and_content_type(self, mock_bs, mock_get):
                for status in [200, 201, 301]:
                    mock_get.return_value = self.setup_mock_response(
                        f"<html><body>Status {status}</body></html>", status_code=status
                    )
                    mock_bs.return_value = self.setup_mock_soup(f"Status {status}")
                    result = WebPageLoader().load(
                        SourceContent(f"https://example.com/{status}")
                    )
                    assert result.metadata["status_code"] == status

                for ctype in ["text/html", "text/plain", "application/xhtml+xml"]:
                    mock_get.return_value = self.setup_mock_response(
                        "<html><body>Content</body></html>", content_type=ctype
                    )
                    mock_bs.return_value = self.setup_mock_soup("Content")
                    result = WebPageLoader().load(SourceContent("https://example.com"))
                    assert result.metadata["content_type"] == ctype
      - |-
        def set_crew(self, crew: Any) -> Memory:
                """Set the crew for this memory instance."""
                self.crew = crew
                return self
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 Fine-tuned on CrewAI (LR=2e-05)
    results:
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 768
          type: dim_768
        metrics:
          - type: cosine_accuracy@1
            value: 0.04
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.04
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.04
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.06
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.04
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.04
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.04
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.03
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.008
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.024
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.04
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.06
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.050819890355577976
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.04333333333333334
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.06130275691848844
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 512
          type: dim_512
        metrics:
          - type: cosine_accuracy@1
            value: 0.01
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.01
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.01
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.01
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.01
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.01
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.01
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.005
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.002
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.006
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.01
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.01
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.01
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.01
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.019316331411936505
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 256
          type: dim_256
        metrics:
          - type: cosine_accuracy@1
            value: 0.01
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.01
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.01
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.03
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.01
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.01
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.01
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.015
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.002
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.006
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.01
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.03
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.020819890355577977
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.013333333333333334
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.028978936077832484
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 128
          type: dim_128
        metrics:
          - type: cosine_accuracy@1
            value: 0.01
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.01
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.01
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.01
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.01
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.01
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.01
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.005
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.002
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.006
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.01
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.01
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.01
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.01
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.027544667112101906
            name: Cosine Map@100
      - task:
          type: information-retrieval
          name: Information Retrieval
        dataset:
          name: dim 64
          type: dim_64
        metrics:
          - type: cosine_accuracy@1
            value: 0.05
            name: Cosine Accuracy@1
          - type: cosine_accuracy@3
            value: 0.05
            name: Cosine Accuracy@3
          - type: cosine_accuracy@5
            value: 0.05
            name: Cosine Accuracy@5
          - type: cosine_accuracy@10
            value: 0.07
            name: Cosine Accuracy@10
          - type: cosine_precision@1
            value: 0.05
            name: Cosine Precision@1
          - type: cosine_precision@3
            value: 0.05
            name: Cosine Precision@3
          - type: cosine_precision@5
            value: 0.05
            name: Cosine Precision@5
          - type: cosine_precision@10
            value: 0.035
            name: Cosine Precision@10
          - type: cosine_recall@1
            value: 0.01
            name: Cosine Recall@1
          - type: cosine_recall@3
            value: 0.03
            name: Cosine Recall@3
          - type: cosine_recall@5
            value: 0.05
            name: Cosine Recall@5
          - type: cosine_recall@10
            value: 0.07
            name: Cosine Recall@10
          - type: cosine_ndcg@10
            value: 0.06081989035557797
            name: Cosine Ndcg@10
          - type: cosine_mrr@10
            value: 0.05333333333333334
            name: Cosine Mrr@10
          - type: cosine_map@100
            value: 0.0838507480466874
            name: Cosine Map@100

CodeBERT Fine-tuned on CrewAI (LR=2e-05)

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("itsanan/codebert-finetuned-crewai-base")
# Run inference
sentences = [
    'Example usage of test_status_code_and_content_type',
    'def test_status_code_and_content_type(self, mock_bs, mock_get):\n        for status in [200, 201, 301]:\n            mock_get.return_value = self.setup_mock_response(\n                f"<html><body>Status {status}</body></html>", status_code=status\n            )\n            mock_bs.return_value = self.setup_mock_soup(f"Status {status}")\n            result = WebPageLoader().load(\n                SourceContent(f"https://example.com/{status}")\n            )\n            assert result.metadata["status_code"] == status\n\n        for ctype in ["text/html", "text/plain", "application/xhtml+xml"]:\n            mock_get.return_value = self.setup_mock_response(\n                "<html><body>Content</body></html>", content_type=ctype\n            )\n            mock_bs.return_value = self.setup_mock_soup("Content")\n            result = WebPageLoader().load(SourceContent("https://example.com"))\n            assert result.metadata["content_type"] == ctype',
    'def set_crew(self, crew: Any) -> Memory:\n        """Set the crew for this memory instance."""\n        self.crew = crew\n        return self',
]
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, 0.9009, 0.9087],
#         [0.9009, 1.0000, 0.9053],
#         [0.9087, 0.9053, 1.0000]])

Evaluation

Metrics

Information Retrieval

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

  {
      "truncate_dim": 768
  }
  

Metric	Value
cosine_accuracy@1	0.04
cosine_accuracy@3	0.04
cosine_accuracy@5	0.04
cosine_accuracy@10	0.06
cosine_precision@1	0.04
cosine_precision@3	0.04
cosine_precision@5	0.04
cosine_precision@10	0.03
cosine_recall@1	0.008
cosine_recall@3	0.024
cosine_recall@5	0.04
cosine_recall@10	0.06
cosine_ndcg@10	0.0508
cosine_mrr@10	0.0433
cosine_map@100	0.0613

Information Retrieval

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

  {
      "truncate_dim": 512
  }
  

Metric	Value
cosine_accuracy@1	0.01
cosine_accuracy@3	0.01
cosine_accuracy@5	0.01
cosine_accuracy@10	0.01
cosine_precision@1	0.01
cosine_precision@3	0.01
cosine_precision@5	0.01
cosine_precision@10	0.005
cosine_recall@1	0.002
cosine_recall@3	0.006
cosine_recall@5	0.01
cosine_recall@10	0.01
cosine_ndcg@10	0.01
cosine_mrr@10	0.01
cosine_map@100	0.0193

Information Retrieval

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

  {
      "truncate_dim": 256
  }
  

Metric	Value
cosine_accuracy@1	0.01
cosine_accuracy@3	0.01
cosine_accuracy@5	0.01
cosine_accuracy@10	0.03
cosine_precision@1	0.01
cosine_precision@3	0.01
cosine_precision@5	0.01
cosine_precision@10	0.015
cosine_recall@1	0.002
cosine_recall@3	0.006
cosine_recall@5	0.01
cosine_recall@10	0.03
cosine_ndcg@10	0.0208
cosine_mrr@10	0.0133
cosine_map@100	0.029

Information Retrieval

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

  {
      "truncate_dim": 128
  }
  

Metric	Value
cosine_accuracy@1	0.01
cosine_accuracy@3	0.01
cosine_accuracy@5	0.01
cosine_accuracy@10	0.01
cosine_precision@1	0.01
cosine_precision@3	0.01
cosine_precision@5	0.01
cosine_precision@10	0.005
cosine_recall@1	0.002
cosine_recall@3	0.006
cosine_recall@5	0.01
cosine_recall@10	0.01
cosine_ndcg@10	0.01
cosine_mrr@10	0.01
cosine_map@100	0.0275

Information Retrieval

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

  {
      "truncate_dim": 64
  }
  

Metric	Value
cosine_accuracy@1	0.05
cosine_accuracy@3	0.05
cosine_accuracy@5	0.05
cosine_accuracy@10	0.07
cosine_precision@1	0.05
cosine_precision@3	0.05
cosine_precision@5	0.05
cosine_precision@10	0.035
cosine_recall@1	0.01
cosine_recall@3	0.03
cosine_recall@5	0.05
cosine_recall@10	0.07
cosine_ndcg@10	0.0608
cosine_mrr@10	0.0533
cosine_map@100	0.0839

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: 13.86 tokens max: 141 tokens	min: 20 tokens mean: 253.07 tokens max: 512 tokens

• Samples:

  anchor	positive
  How to implement LLMCallCompletedEvent?	class LLMCallCompletedEvent(LLMEventBase): """Event emitted when a LLM call completes""" type: str = "llm_call_completed" messages: str | list[dict[str, Any]] | None = None response: Any call_type: LLMCallType model: str | None = None
  How does get_llm_response work in Python?	def get_llm_response( llm: LLM | BaseLLM, messages: list[LLMMessage], callbacks: list[TokenCalcHandler], printer: Printer, from_task: Task | None = None, from_agent: Agent | LiteAgent | None = None, response_model: type[BaseModel] | None = None, executor_context: CrewAgentExecutor | LiteAgent | None = None, ) -> str: """Call the LLM and return the response, handling any invalid responses. Args: llm: The LLM instance to call. messages: The messages to send to the LLM. callbacks: List of callbacks for the LLM call. printer: Printer instance for output. from_task: Optional task context for the LLM call. from_agent: Optional agent context for the LLM call. response_model: Optional Pydantic model for structured outputs. executor_context: Optional executor context for hook invocation. Returns: The response from the LLM as a string. Raises: Exception: If an error ...
  Example usage of _run	def _run( self, **kwargs: Any, ) -> Any: website_url: str | None = kwargs.get("website_url", self.website_url) if website_url is None: raise ValueError("Website URL must be provided.") page = requests.get( website_url, timeout=15, headers=self.headers, cookies=self.cookies if self.cookies else {}, ) page.encoding = page.apparent_encoding parsed = BeautifulSoup(page.text, "html.parser") text = "The following text is scraped website content:\n\n" text += parsed.get_text(" ") text = re.sub("[ \t]+", " ", text) return re.sub("\s+\n\s+", "\n", text)

• 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: steps
• per_device_train_batch_size: 4
• per_device_eval_batch_size: 4
• gradient_accumulation_steps: 32
• learning_rate: 2e-05
• weight_decay: 0.01
• num_train_epochs: 20
• 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: steps
• 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: 32
• eval_accumulation_steps: None
• torch_empty_cache_steps: None
• learning_rate: 2e-05
• weight_decay: 0.01
• adam_beta1: 0.9
• adam_beta2: 0.999
• adam_epsilon: 1e-08
• max_grad_norm: 1.0
• num_train_epochs: 20
• max_steps: -1
• lr_scheduler_type: cosine
• lr_scheduler_kwargs: None
• 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
0.9956	7	-	0.04	0.04	0.03	0.0262	0.0308
1.2844	10	7.098	-	-	-	-	-
1.8533	14	-	0.0362	0.02	0.0354	0.0154	0.0508
2.5689	20	6.5515	-	-	-	-	-
2.7111	21	-	0.0508	0.01	0.0208	0.01	0.0608

• The bold row denotes the saved checkpoint.

Framework Versions

• Python: 3.12.12
• Sentence Transformers: 5.2.2
• Transformers: 4.57.6
• 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}
}