be able to reconstruct the coarse wav from coarse transformer to soundstream decoder

        cond_scale = 3.,

        filter_thres = 0.9,

        temperature = 1.,

        reshape_output = True

        reshape_output = True,

        reconstruct_wave = False

    ):

        batch, device = semantic_token_ids.shape[0], self.device

        output = mask_out_after_eos_id(output, self.eos_id, include_eos = False)

        if reshape_output:

        if reshape_output or reconstruct_wave:

            output = rearrange(output, 'b (n q) -> b n q', q = self.num_coarse_quantizers)

        if reconstruct_wave:

            assert exists(self.soundstream)

            wav = self.soundstream.decode_from_codebook_indices(output)

            wav = rearrange(wav, 'b 1 n -> b n')

            return wav

        return output

    def forward(

import torch

from torch import nn, einsum

import torch.nn.functional as F

from einops import rearrange

from einops import rearrange, reduce

from vector_quantize_pytorch import ResidualVQ

        self.adversarial_loss_weight = adversarial_loss_weight

        self.feature_loss_weight = feature_loss_weight

    def decode_from_codebook_indices(self, quantized_indices):

        codes = self.rq.get_codes_from_indices(quantized_indices)

        x = reduce(codes, 'q ... -> ...', 'sum')

        x = rearrange(x, 'b n c -> b c n')

        return self.decoder(x)

    def load(self, path):

        path = Path(path)

        assert path.exists()

setup(

  name = 'audiolm-pytorch',

  packages = find_packages(exclude=[]),

  version = '0.0.40',

  version = '0.0.41',

  license='MIT',

  description = 'AudioLM - Language Modeling Approach to Audio Generation from Google Research - Pytorch',

  author = 'Phil Wang',

    'torchaudio',

    'transformers',

    'typeguard',

    'vector-quantize-pytorch>=0.10.10'

    'vector-quantize-pytorch>=0.10.11'

  ],

  classifiers=[

    'Development Status :: 4 - Beta',