How to use _get_encoder_class method in autotest

Best Python code snippet using autotest_python

transformer_builders.py

Source:transformer_builders.py

...111    RecurrentTransformerEncoder.112    Inheriting classes should implement the following:113    - _get_attention_builder()114    - _get_attention_layer_class()115    - _get_encoder_class()116    - _get_encoder_layer_class()117    """118    def __init__(self):119        super(BaseTransformerEncoderBuilder, self).__init__()120        self._attention_builder = self._get_attention_builder()121        self._attention_type = "full"122    def _get_attention_builder(self):123        """Return an instance of the appropriate attention builder."""124        raise NotImplementedError()125    def _get_attention_layer_class(self):126        """Return the class for the layer that projects queries keys and127        values."""128        raise NotImplementedError()129    def _get_encoder_class(self):130        """Return the class for the transformer encoder."""131        raise NotImplementedError()132    def _get_encoder_layer_class(self):133        """Return the class for the transformer encoder layer."""134        raise NotImplementedError()135    @property136    def attention(self):137        """The attention builder instance."""138        return self._attention_builder139    @property140    def attention_type(self):141        """The attention implementation chosen."""142        return self._attention_type143    @attention_type.setter144    def attention_type(self, val):145        if not self._attention_builder.validate_attention_type(val):146            raise ValueError(("{!r} is not an available attention "147                              "type").format(val))148        self._attention_type = val149    def __setattr__(self, key, val):150        # "protected" attributes are settable (probably from withing the class)151        if key[0] == "_":152            return super().__setattr__(key, val)153        # Existing attributes are settable but they might also be attention154        # parameters so try that as well155        fail_on_exception = True156        if hasattr(self, key):157            super().__setattr__(key, val)158            fail_on_exception = False159        # Non-existing "public" attributes may be attention parameters160        try:161            setattr(self._attention_builder, key, val)162        except:163            if fail_on_exception:164                raise165    def get(self):166        """Build the transformer and return it."""167        # Set the event dispatcher to the attention builder168        self.attention.event_dispatcher = self.event_dispatcher169        # Extract into local variables the classes to be used170        Encoder = self._get_encoder_class()171        EncoderLayer = self._get_encoder_layer_class()172        Attention = self._get_attention_layer_class()173        model_dimensions = self.value_dimensions*self.n_heads174        return Encoder(175            [176                EncoderLayer(177                    Attention(178                        self.attention.get(self.attention_type),179                        model_dimensions,180                        self.n_heads,181                        d_keys=self.query_dimensions,182                        d_values=self.value_dimensions,183                        event_dispatcher=self.event_dispatcher184                    ),185                    model_dimensions,186                    self.feed_forward_dimensions,187                    self.dropout,188                    self.activation,189                    event_dispatcher=self.event_dispatcher190                )191                for _ in range(self.n_layers)192            ],193            (LayerNorm(model_dimensions) if self.final_normalization else None),194            event_dispatcher=self.event_dispatcher195        )196class TransformerEncoderBuilder(BaseTransformerEncoderBuilder):197    """Build a batch transformer encoder for training or processing of198    sequences all elements at a time.199    Example usage:200        builder = TransformerEncoderBuilder()201        builder.n_layers = 12202        builder.n_heads = 8203        builder.feed_forward_dimensions = 1024204        builder.query_dimensions = 64205        builder.value_dimensions = 64206        builder.dropout = 0.1207        builder.attention_dropout = 0.1208        builder.attention_type = "linear"209        transformer = builder.get()210    """211    def _get_attention_builder(self):212        """Return an instance of the appropriate attention builder."""213        return AttentionBuilder()214    def _get_attention_layer_class(self):215        """Return the class for the layer that projects queries keys and216        values."""217        return AttentionLayer218    def _get_encoder_class(self):219        """Return the class for the transformer encoder."""220        return TransformerEncoder221    def _get_encoder_layer_class(self):222        """Return the class for the transformer encoder layer."""223        return TransformerEncoderLayer224class RecurrentEncoderBuilder(BaseTransformerEncoderBuilder):225    """Build a transformer encoder for autoregressive processing of sequences.226    Example usage:227        builder = RecurrentEncoderBuilder()228        builder.n_layers = 12229        builder.n_heads = 8230        builder.feed_forward_dimensions = 1024231        builder.query_dimensions = 64232        builder.value_dimensions = 64233        builder.dropout = 0.1234        builder.attention_dropout = 0.1235        builder.attention_type = "linear"236        transformer = builder.get()237    """238    def _get_attention_builder(self):239        """Return an attention builder for recurrent attention."""240        return RecurrentAttentionBuilder()241    def _get_attention_layer_class(self):242        """Return the class for the recurrent layer that projects queries keys243        and values."""244        return RecurrentAttentionLayer245    def _get_encoder_class(self):246        """Return the class for the recurrent transformer encoder."""247        return RecurrentTransformerEncoder248    def _get_encoder_layer_class(self):249        """Return the class for the recurrent transformer encoder layer."""250        return RecurrentTransformerEncoderLayer251class BaseTransformerDecoderBuilder(BaseTransformerBuilder):252    """Similar to BaseTransformerEncoderBuilder implement the logic of253    building the transformer decoder without defining concrete layers.254    Inheriting classes should implement the following:255    - _get_self_attention_builder() and _get_cross_attention_builder()256    - _get_self_attention_layer_class() and _get_cross_attention_layer_class()257    - _get_decoder_class()258    - _get_decoder_layer_class()259    """...

Automation Testing Tutorials

Learn to execute automation testing from scratch with LambdaTest Learning Hub. Right from setting up the prerequisites to run your first automation test, to following best practices and diving deeper into advanced test scenarios. LambdaTest Learning Hubs compile a list of step-by-step guides to help you be proficient with different test automation frameworks i.e. Selenium, Cypress, TestNG etc.