# Copyright 2021 Zilliz. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import numpy as np
import torch
import towhee

from PIL import Image as PILImage

from transformers import Data2VecAudioModel, Wav2Vec2Processor
from towhee.operator.base import NNOperator

class Data2VecAudio(NNOperator):
    def __init__(self, model_name = "facebook/data2vec-audio-base-960h"):
        self.model = Data2VecAudioModel.from_pretrained(model_name)
        self.processor = Wav2Vec2Processor.from_pretrained(model_name)

    def __call__(self, data):
        audio = np.hstack(data).reshape(1, -1)
        audio = audio.astype(np.float32, order='C') / 32768.0
        sampling_rate = data[0]._sample_rate
        inputs = self.processor(audio.flatten(), sampling_rate=sampling_rate, return_tensors="pt")
        with torch.no_grad():
            outputs = self.model(**inputs)
        last_hidden_states = outputs.last_hidden_state
        feat = last_hidden_states[:,-1,:].flatten().detach().cpu().numpy()
        return feat