Upload feature extractor

feature_extraction_wavlm_spkreg.py

@@ -0,0 +1,225 @@
+"""
+Feature extractor class for WavLM
+"""
+
+from typing import List, Optional, Union
+
+import numpy as np
+
+from transformers.feature_extraction_sequence_utils import SequenceFeatureExtractor
+from transformers.feature_extraction_utils import BatchFeature
+from transformers.utils import PaddingStrategy, TensorType, logging
+
+logger = logging.get_logger(__name__)
+
+
+class Wav2LMSpkRegFeatureExtractor(SequenceFeatureExtractor):
+    r"""
+    Constructs a WavLM feature extractor.
+
+    This feature extractor inherits from [`~feature_extraction_sequence_utils.SequenceFeatureExtractor`] which contains
+    most of the main methods. Users should refer to this superclass for more information regarding those methods.
+
+    Args:
+        feature_size (`int`, *optional*, defaults to 1):
+            The feature dimension of the extracted features.
+        sampling_rate (`int`, *optional*, defaults to 16000):
+            The sampling rate at which the audio files should be digitalized expressed in hertz (Hz).
+        padding_value (`float`, *optional*, defaults to 0.0):
+            The value that is used to fill the padding values.
+        do_normalize (`bool`, *optional*, defaults to `True`):
+            Whether or not to zero-mean unit-variance normalize the input. Normalizing can help to significantly
+            improve the performance for some models, *e.g.*,
+            [wav2vec2-lv60](https://huggingface.co/models?search=lv60).
+        return_attention_mask (`bool`, *optional*, defaults to `False`):
+            Whether or not [`~Wav2Vec2FeatureExtractor.__call__`] should return `attention_mask`.
+
+            <Tip>
+
+            Wav2Vec2 models that have set `config.feat_extract_norm == "group"`, such as
+            [wav2vec2-base](https://huggingface.co/facebook/wav2vec2-base-960h), have **not** been trained using
+            `attention_mask`. For such models, `input_values` should simply be padded with 0 and no `attention_mask`
+            should be passed.
+
+            For Wav2Vec2 models that have set `config.feat_extract_norm == "layer"`, such as
+            [wav2vec2-lv60](https://huggingface.co/facebook/wav2vec2-large-960h-lv60-self), `attention_mask` should be
+            passed for batched inference.
+
+            </Tip>"""
+
+    model_input_names = ["input_values", "attention_mask"]
+
+    def __init__(
+        self,
+        feature_size=1,
+        sampling_rate=16000,
+        padding_value=0.0,
+        return_attention_mask=False,
+        do_normalize=True,
+        **kwargs,
+    ):
+        super().__init__(feature_size=feature_size, sampling_rate=sampling_rate, padding_value=padding_value, **kwargs)
+        self.return_attention_mask = return_attention_mask
+        self.do_normalize = do_normalize
+
+    @staticmethod
+    def zero_mean_unit_var_norm(
+        input_values: List[np.ndarray], attention_mask: List[np.ndarray], padding_value: float = 0.0
+    ) -> List[np.ndarray]:
+        """
+        Every array in the list is normalized to have zero mean and unit variance
+        """
+        if attention_mask is not None:
+            attention_mask = np.array(attention_mask, np.int32)
+            normed_input_values = []
+
+            for vector, length in zip(input_values, attention_mask.sum(-1)):
+                normed_slice = (vector - vector[:length].mean()) / np.sqrt(vector[:length].var() + 1e-7)
+                if length < normed_slice.shape[0]:
+                    normed_slice[length:] = padding_value
+
+                normed_input_values.append(normed_slice)
+        else:
+            normed_input_values = [(x - x.mean()) / np.sqrt(x.var() + 1e-7) for x in input_values]
+
+        return normed_input_values
+
+    def __call__(
+        self,
+        raw_speech: Union[np.ndarray, List[float], List[np.ndarray], List[List[float]]],
+        padding: Union[bool, str, PaddingStrategy] = False,
+        max_length: Optional[int] = None,
+        truncation: bool = False,
+        pad_to_multiple_of: Optional[int] = None,
+        return_attention_mask: Optional[bool] = None,
+        return_tensors: Optional[Union[str, TensorType]] = None,
+        sampling_rate: Optional[int] = None,
+        **kwargs,
+    ) -> BatchFeature:
+        """
+        Main method to featurize and prepare for the model one or several sequence(s).
+
+        Args:
+            raw_speech (`np.ndarray`, `List[float]`, `List[np.ndarray]`, `List[List[float]]`):
+                The sequence or batch of sequences to be padded. Each sequence can be a numpy array, a list of float
+                values, a list of numpy arrays or a list of list of float values. Must be mono channel audio, not
+                stereo, i.e. single float per timestep.
+            padding (`bool`, `str` or [`~utils.PaddingStrategy`], *optional*, defaults to `False`):
+                Select a strategy to pad the returned sequences (according to the model's padding side and padding
+                index) among:
+
+                - `True` or `'longest'`: Pad to the longest sequence in the batch (or no padding if only a single
+                  sequence if provided).
+                - `'max_length'`: Pad to a maximum length specified with the argument `max_length` or to the maximum
+                  acceptable input length for the model if that argument is not provided.
+                - `False` or `'do_not_pad'` (default): No padding (i.e., can output a batch with sequences of different
+                  lengths).
+            max_length (`int`, *optional*):
+                Maximum length of the returned list and optionally padding length (see above).
+            truncation (`bool`):
+                Activates truncation to cut input sequences longer than *max_length* to *max_length*.
+            pad_to_multiple_of (`int`, *optional*):
+                If set will pad the sequence to a multiple of the provided value.
+
+                This is especially useful to enable the use of Tensor Cores on NVIDIA hardware with compute capability
+                `>= 7.5` (Volta), or on TPUs which benefit from having sequence lengths be a multiple of 128.
+            return_attention_mask (`bool`, *optional*):
+                Whether to return the attention mask. If left to the default, will return the attention mask according
+                to the specific feature_extractor's default.
+
+                [What are attention masks?](../glossary#attention-mask)
+
+                <Tip>
+
+                Wav2Vec2 models that have set `config.feat_extract_norm == "group"`, such as
+                [wav2vec2-base](https://huggingface.co/facebook/wav2vec2-base-960h), have **not** been trained using
+                `attention_mask`. For such models, `input_values` should simply be padded with 0 and no
+                `attention_mask` should be passed.
+
+                For Wav2Vec2 models that have set `config.feat_extract_norm == "layer"`, such as
+                [wav2vec2-lv60](https://huggingface.co/facebook/wav2vec2-large-960h-lv60-self), `attention_mask` should
+                be passed for batched inference.
+
+                </Tip>
+
+            return_tensors (`str` or [`~utils.TensorType`], *optional*):
+                If set, will return tensors instead of list of python integers. Acceptable values are:
+
+                - `'tf'`: Return TensorFlow `tf.constant` objects.
+                - `'pt'`: Return PyTorch `torch.Tensor` objects.
+                - `'np'`: Return Numpy `np.ndarray` objects.
+            sampling_rate (`int`, *optional*):
+                The sampling rate at which the `raw_speech` input was sampled. It is strongly recommended to pass
+                `sampling_rate` at the forward call to prevent silent errors.
+            padding_value (`float`, *optional*, defaults to 0.0):
+        """
+
+        if sampling_rate is not None:
+            if sampling_rate != self.sampling_rate:
+                raise ValueError(
+                    f"The model corresponding to this feature extractor: {self} was trained using a sampling rate of"
+                    f" {self.sampling_rate}. Please make sure that the provided `raw_speech` input was sampled with"
+                    f" {self.sampling_rate} and not {sampling_rate}."
+                )
+        else:
+            logger.warning(
+                "It is strongly recommended to pass the ``sampling_rate`` argument to this function. "
+                "Failing to do so can result in silent errors that might be hard to debug."
+            )
+
+        is_batched_numpy = isinstance(raw_speech, np.ndarray) and len(raw_speech.shape) > 1
+        if is_batched_numpy and len(raw_speech.shape) > 2:
+            raise ValueError(f"Only mono-channel audio is supported for input to {self}")
+        is_batched = is_batched_numpy or (
+            isinstance(raw_speech, (list, tuple)) and (isinstance(raw_speech[0], (np.ndarray, tuple, list)))
+        )
+
+        # always return batch
+        if not is_batched:
+            raw_speech = [raw_speech]
+
+        # convert into correct format for padding
+        encoded_inputs = BatchFeature({"input_values": raw_speech})
+
+        padded_inputs = self.pad(
+            encoded_inputs,
+            padding=padding,
+            max_length=max_length,
+            truncation=truncation,
+            pad_to_multiple_of=pad_to_multiple_of,
+            return_attention_mask=return_attention_mask,
+        )
+
+        # convert input values to correct format
+        input_values = padded_inputs["input_values"]
+        if not isinstance(input_values[0], np.ndarray):
+            padded_inputs["input_values"] = [np.asarray(array, dtype=np.float32) for array in input_values]
+        elif (
+            not isinstance(input_values, np.ndarray)
+            and isinstance(input_values[0], np.ndarray)
+            and input_values[0].dtype is np.dtype(np.float64)
+        ):
+            padded_inputs["input_values"] = [array.astype(np.float32) for array in input_values]
+        elif isinstance(input_values, np.ndarray) and input_values.dtype is np.dtype(np.float64):
+            padded_inputs["input_values"] = input_values.astype(np.float32)
+
+        # convert attention_mask to correct format
+        attention_mask = padded_inputs.get("attention_mask")
+        if attention_mask is not None:
+            padded_inputs["attention_mask"] = [np.asarray(array, dtype=np.int32) for array in attention_mask]
+
+        # zero-mean and unit-variance normalization
+        if self.do_normalize:
+            attention_mask = (
+                attention_mask
+                if self._get_padding_strategies(padding, max_length=max_length) is not PaddingStrategy.DO_NOT_PAD
+                else None
+            )
+            padded_inputs["input_values"] = self.zero_mean_unit_var_norm(
+                padded_inputs["input_values"], attention_mask=attention_mask, padding_value=self.padding_value
+            )
+
+        if return_tensors is not None:
+            padded_inputs = padded_inputs.convert_to_tensors(return_tensors)
+
+        return padded_inputs

preprocessor_config.json

@@ -0,0 +1,12 @@
+{
+  "auto_map": {
+    "AutoFeatureExtractor": "feature_extraction_wavlm_spkreg.Wav2LMSpkRegFeatureExtractor"
+  },
+  "do_normalize": false,
+  "feature_extractor_type": "Wav2LMSpkRegFeatureExtractor",
+  "feature_size": 1,
+  "padding_side": "right",
+  "padding_value": 0.0,
+  "return_attention_mask": true,
+  "sampling_rate": 16000
+}