***

layout: overview
slug: nemo-curator/nemo\_curator/models/nsfw
title: nemo\_curator.models.nsfw
--------------------------------

## Module Contents

### Classes

| Name                                                       | Description                                            |
| ---------------------------------------------------------- | ------------------------------------------------------ |
| [`NSFWModel`](#nemo_curator-models-nsfw-NSFWModel)         | NSFW detection model.                                  |
| [`NSFWScorer`](#nemo_curator-models-nsfw-NSFWScorer)       | Public interface for NSFW scoring of image embeddings. |
| [`Normalization`](#nemo_curator-models-nsfw-Normalization) | Normalization layer for NSFW model.                    |

### Data

[`_NSFW_MODEL_ID`](#nemo_curator-models-nsfw-_NSFW_MODEL_ID)

[`_URL_MAPPING`](#nemo_curator-models-nsfw-_URL_MAPPING)

### API

<Anchor id="nemo_curator-models-nsfw-NSFWModel">
  <CodeBlock showLineNumbers={false} wordWrap={true}>
    ```python
    class nemo_curator.models.nsfw.NSFWModel()
    ```
  </CodeBlock>
</Anchor>

<Indent>
  **Bases:** `Module`

  NSFW detection model.

  A neural network that processes CLIP embeddings to predict NSFW scores.
  Based on LAION's CLIP-based-NSFW-Detector.

  <ParamField path="act" type="= nn.ReLU()" />

  <ParamField path="act_out" type="= nn.Sigmoid()" />

  <ParamField path="linear_1" type="= nn.Linear(768, 64)" />

  <ParamField path="linear_2" type="= nn.Linear(64, 512)" />

  <ParamField path="linear_3" type="= nn.Linear(512, 256)" />

  <ParamField path="linear_4" type="= nn.Linear(256, 1)" />

  <ParamField path="norm" type="= Normalization([768])" />

  <Anchor id="nemo_curator-models-nsfw-NSFWModel-forward">
    <CodeBlock showLineNumbers={false} wordWrap={true}>
      ```python
      nemo_curator.models.nsfw.NSFWModel.forward(
          x: torch.Tensor
      ) -> torch.Tensor
      ```
    </CodeBlock>
  </Anchor>

  <Indent>
    Forward pass through the NSFW model.

    **Parameters:**

    <ParamField path="x" type="torch.Tensor">
      Input embeddings tensor.
    </ParamField>

    **Returns:** `torch.Tensor`

    NSFW probability scores.
  </Indent>
</Indent>

<Anchor id="nemo_curator-models-nsfw-NSFWScorer">
  <CodeBlock showLineNumbers={false} wordWrap={true}>
    ```python
    class nemo_curator.models.nsfw.NSFWScorer(
        model_dir: str
    )
    ```
  </CodeBlock>
</Anchor>

<Indent>
  **Bases:** [ModelInterface](/nemo-curator/nemo_curator/models/base#nemo_curator-models-base-ModelInterface)

  Public interface for NSFW scoring of image embeddings.

  This class provides a standardized interface for scoring the likelihood
  of images containing sexually explicit material using a pre-trained model.

  <ParamField path="conda_env_name" type="str">
    Get the name of the conda environment required for this model.
  </ParamField>

  <ParamField path="device" type="= 'cuda' if torch.cuda.is_available() else 'cpu'" />

  <ParamField path="dtype" type="= torch.float32" />

  <ParamField path="model_id_names" type="list[str]">
    Get the model ID names associated with this NSFW scorer.
  </ParamField>

  <Anchor id="nemo_curator-models-nsfw-NSFWScorer-__call__">
    <CodeBlock showLineNumbers={false} wordWrap={true}>
      ```python
      nemo_curator.models.nsfw.NSFWScorer.__call__(
          embeddings: torch.Tensor | numpy.typing.NDArray[numpy.float32]
      ) -> torch.Tensor
      ```
    </CodeBlock>
  </Anchor>

  <Indent>
    Score the NSFW likelihood of input embeddings.

    **Parameters:**

    <ParamField path="embeddings" type="torch.Tensor | npt.NDArray[np.float32]">
      Input embeddings as either a torch tensor or numpy array.
    </ParamField>

    **Returns:** `torch.Tensor`

    NSFW probability scores for each input embedding.
  </Indent>

  <Anchor id="nemo_curator-models-nsfw-NSFWScorer-download_weights_on_node">
    <CodeBlock showLineNumbers={false} wordWrap={true}>
      ```python
      nemo_curator.models.nsfw.NSFWScorer.download_weights_on_node(
          model_dir: str
      ) -> None
      ```
    </CodeBlock>
  </Anchor>

  <Indent>
    <Badge>
      classmethod
    </Badge>

    Download NSFW model weights from LAION repository.

    **Parameters:**

    <ParamField path="model_dir" type="str">
      Directory to download the weights to.
    </ParamField>
  </Indent>

  <Anchor id="nemo_curator-models-nsfw-NSFWScorer-setup">
    <CodeBlock showLineNumbers={false} wordWrap={true}>
      ```python
      nemo_curator.models.nsfw.NSFWScorer.setup() -> None
      ```
    </CodeBlock>
  </Anchor>

  <Indent>
    Set up the NSFW scoring model by loading weights.
  </Indent>
</Indent>

<Anchor id="nemo_curator-models-nsfw-Normalization">
  <CodeBlock showLineNumbers={false} wordWrap={true}>
    ```python
    class nemo_curator.models.nsfw.Normalization(
        shape: list[int]
    )
    ```
  </CodeBlock>
</Anchor>

<Indent>
  **Bases:** `Module`

  Normalization layer for NSFW model.

  Applies normalization to input tensors using pre-computed mean and variance.

  <Anchor id="nemo_curator-models-nsfw-Normalization-forward">
    <CodeBlock showLineNumbers={false} wordWrap={true}>
      ```python
      nemo_curator.models.nsfw.Normalization.forward(
          x: torch.Tensor
      ) -> torch.Tensor
      ```
    </CodeBlock>
  </Anchor>

  <Indent>
    Apply normalization to input tensor.

    **Parameters:**

    <ParamField path="x" type="torch.Tensor">
      Input tensor to normalize.
    </ParamField>

    **Returns:** `torch.Tensor`

    Normalized tensor.
  </Indent>
</Indent>

<Anchor id="nemo_curator-models-nsfw-_NSFW_MODEL_ID">
  <CodeBlock showLineNumbers={false} wordWrap={true}>
    ```python
    nemo_curator.models.nsfw._NSFW_MODEL_ID = 'laion/clip-autokeras-binary-nsfw'
    ```
  </CodeBlock>
</Anchor>

<Anchor id="nemo_curator-models-nsfw-_URL_MAPPING">
  <CodeBlock showLineNumbers={false} wordWrap={true}>
    ```python
    nemo_curator.models.nsfw._URL_MAPPING = {'laion/clip-autokeras-binary-nsfw': 'https://github.com/LAION-AI/CLIP-based-NSF...
    ```
  </CodeBlock>
</Anchor>