forked from mrq/DL-Art-School
Go back to vanilla flavor of diffusion
This commit is contained in:
James Betker
parent
23da073037
commit
d016a2fbad
|
|
@ -141,12 +141,12 @@ class UnsupervisedAudioDataset(torch.utils.data.Dataset):
|
|||
if __name__ == '__main__':
|
||||
params = {
|
||||
'mode': 'unsupervised_audio',
|
||||
'path': ['Z:\\split\\cleaned\\books0', 'Z:\\split\\cleaned\\books2'],
|
||||
'cache_path': 'E:\\audio\\remote-cache.pth',
|
||||
'path': ['\\\\192.168.5.3\\rtx3080_audio_y\\split\\books2', '\\\\192.168.5.3\\rtx3080_audio\\split\\books1', '\\\\192.168.5.3\\rtx3080_audio\\split\\cleaned-2'],
|
||||
'cache_path': 'E:\\audio\\remote-cache2.pth',
|
||||
'sampling_rate': 22050,
|
||||
'pad_to_seconds': 5,
|
||||
'pad_to_samples': 40960,
|
||||
'phase': 'train',
|
||||
'n_workers': 4,
|
||||
'n_workers': 1,
|
||||
'batch_size': 16,
|
||||
'extra_samples': 4,
|
||||
}
|
||||
|
|
@ -156,8 +156,7 @@ if __name__ == '__main__':
|
|||
dl = create_dataloader(ds, params)
|
||||
i = 0
|
||||
for b in tqdm(dl):
|
||||
for b_ in range(16):
|
||||
pass
|
||||
#torchaudio.save(f'{i}_clip1_{b_}.wav', b['clip1'][b_], ds.sampling_rate)
|
||||
#torchaudio.save(f'{i}_clip2_{b_}.wav', b['clip2'][b_], ds.sampling_rate)
|
||||
#i += 1
|
||||
for b_ in range(b['clip'].shape[0]):
|
||||
#pass
|
||||
torchaudio.save(f'{i}_clip_{b_}.wav', b['clip'][b_], ds.sampling_rate)
|
||||
i += 1
|
||||
|
|
|
|||
|
|
@ -13,17 +13,10 @@ from utils.util import get_mask_from_lengths
|
|||
class DiscreteSpectrogramConditioningBlock(nn.Module):
|
||||
def __init__(self, dvae_channels, channels):
|
||||
super().__init__()
|
||||
self.emb = nn.Conv1d(dvae_channels, channels, kernel_size=1)
|
||||
self.intg = nn.Sequential(
|
||||
normalization(channels*2),
|
||||
nn.SiLU(),
|
||||
nn.Conv1d(channels*2, channels*2, kernel_size=1),
|
||||
normalization(channels*2),
|
||||
nn.SiLU(),
|
||||
nn.Conv1d(channels*2, channels, kernel_size=3, padding=1),
|
||||
self.intg = nn.Sequential(nn.Conv1d(dvae_channels, channels, kernel_size=1),
|
||||
normalization(channels),
|
||||
nn.SiLU(),
|
||||
zero_module(nn.Conv1d(channels, channels, kernel_size=1)))
|
||||
nn.Conv1d(channels, channels, kernel_size=3))
|
||||
|
||||
"""
|
||||
Embeds the given codes and concatenates them onto x. Return shape is the same as x.shape.
|
||||
|
|
@ -34,11 +27,9 @@ class DiscreteSpectrogramConditioningBlock(nn.Module):
|
|||
def forward(self, x, dvae_in):
|
||||
b, c, S = x.shape
|
||||
_, q, N = dvae_in.shape
|
||||
emb = self.emb(dvae_in)
|
||||
emb = self.intg(dvae_in)
|
||||
emb = nn.functional.interpolate(emb, size=(S,), mode='nearest')
|
||||
together = torch.cat([x, emb], dim=1)
|
||||
together = self.intg(together)
|
||||
return together + x
|
||||
return torch.cat([x, emb], dim=1)
|
||||
|
||||
|
||||
class DiffusionVocoderWithRef(nn.Module):
|
||||
|
|
@ -81,10 +72,10 @@ class DiffusionVocoderWithRef(nn.Module):
|
|||
dropout=0,
|
||||
# res 1, 2, 4, 8,16,32,64,128,256,512, 1K, 2K
|
||||
channel_mult= (1,1.5,2, 3, 4, 6, 8, 12, 16, 24, 32, 48),
|
||||
num_res_blocks=(1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2),
|
||||
num_res_blocks=(1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 2),
|
||||
# spec_cond: 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0)
|
||||
# attn: 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1
|
||||
spectrogram_conditioning_resolutions=(1,8,64,512),
|
||||
spectrogram_conditioning_resolutions=(512,),
|
||||
attention_resolutions=(512,1024,2048),
|
||||
conv_resample=True,
|
||||
dims=1,
|
||||
|
|
@ -149,6 +140,7 @@ class DiffusionVocoderWithRef(nn.Module):
|
|||
for level, (mult, num_blocks) in enumerate(zip(channel_mult, num_res_blocks)):
|
||||
if ds in spectrogram_conditioning_resolutions:
|
||||
self.input_blocks.append(DiscreteSpectrogramConditioningBlock(discrete_codes, ch))
|
||||
ch *= 2
|
||||
|
||||
for _ in range(num_blocks):
|
||||
layers = [
|
||||
|
|
@ -340,7 +332,7 @@ def register_unet_diffusion_vocoder_with_ref(opt_net, opt):
|
|||
if __name__ == '__main__':
|
||||
clip = torch.randn(2, 1, 40960)
|
||||
#spec = torch.randint(8192, (2, 40,))
|
||||
spec = torch.randn(8,512,160)
|
||||
spec = torch.randn(2,512,160)
|
||||
cond = torch.randn(2, 3, 80, 173)
|
||||
ts = torch.LongTensor([555, 556])
|
||||
model = DiffusionVocoderWithRef(32, conditioning_inputs_provided=False)
|
||||
|
|
|
|||
|
|
@ -14,7 +14,7 @@ from torchvision.transforms import ToTensor
|
|||
import utils
|
||||
import utils.options as option
|
||||
import utils.util as util
|
||||
from data.audio.wavfile_dataset import load_audio_from_wav
|
||||
from models.tacotron2.taco_utils import load_wav_to_torch
|
||||
from trainer.ExtensibleTrainer import ExtensibleTrainer
|
||||
from data import create_dataset, create_dataloader
|
||||
from tqdm import tqdm
|
||||
|
|
@ -55,7 +55,7 @@ if __name__ == "__main__":
|
|||
torch.backends.cudnn.benchmark = True
|
||||
want_metrics = False
|
||||
parser = argparse.ArgumentParser()
|
||||
parser.add_argument('-opt', type=str, help='Path to options YAML file.', default='../options/test_diffusion_vocoder.yml')
|
||||
parser.add_argument('-opt', type=str, help='Path to options YAML file.', default='../options/test_diffusion_vocoder_10-17.yml')
|
||||
opt = option.parse(parser.parse_args().opt, is_train=False)
|
||||
opt = option.dict_to_nonedict(opt)
|
||||
utils.util.loaded_options = opt
|
||||
|
|
@ -70,7 +70,9 @@ if __name__ == "__main__":
|
|||
|
||||
# Load test image
|
||||
if audio_mode:
|
||||
im = load_audio_from_wav(opt['image'], opt['sample_rate'])
|
||||
im, sr = load_wav_to_torch(opt['image'])
|
||||
assert sr == 22050
|
||||
im = im.unsqueeze(0)
|
||||
im = im[:, :(im.shape[1]//4096)*4096]
|
||||
else:
|
||||
im = ToTensor()(Image.open(opt['image'])) * 2 - 1
|
||||
|
|
|
|||
|
|
@ -96,9 +96,6 @@ class GaussianDiffusionInferenceInjector(Injector):
|
|||
self.use_ema_model = opt_get(opt, ['use_ema'], False)
|
||||
self.noise_style = opt_get(opt, ['noise_type'], 'random') # 'zero', 'fixed' or 'random'
|
||||
|
||||
self.model_fn = opt_get(opt, ['model_function'], None)
|
||||
self.model_fn = None if self.model_fn is None else getattr(self.generator, self.model_fn)
|
||||
|
||||
def forward(self, state):
|
||||
if self.use_ema_model:
|
||||
gen = self.env['emas'][self.opt['generator']]
|
||||
|
|
@ -114,6 +111,9 @@ class GaussianDiffusionInferenceInjector(Injector):
|
|||
elif 'spectrogram' in model_inputs.keys():
|
||||
output_shape = (self.output_batch_size, 1, model_inputs['spectrogram'].shape[-1]*256)
|
||||
dev = model_inputs['spectrogram'].device
|
||||
elif 'discrete_spectrogram' in model_inputs.keys():
|
||||
output_shape = (self.output_batch_size, 1, model_inputs['discrete_spectrogram'].shape[-1]*1024)
|
||||
dev = model_inputs['discrete_spectrogram'].device
|
||||
else:
|
||||
raise NotImplementedError
|
||||
noise = None
|
||||
|
|
@ -123,7 +123,7 @@ class GaussianDiffusionInferenceInjector(Injector):
|
|||
if not hasattr(self, 'fixed_noise') or self.fixed_noise.shape != output_shape:
|
||||
self.fixed_noise = torch.randn(output_shape, device=dev)
|
||||
noise = self.fixed_noise
|
||||
gen = self.sampling_fn(self.model_fn, output_shape, noise=noise, model_kwargs=model_inputs, progress=True, device=dev)
|
||||
gen = self.sampling_fn(gen, output_shape, noise=noise, model_kwargs=model_inputs, progress=True, device=dev)
|
||||
if self.undo_n1_to_1:
|
||||
gen = (gen + 1) / 2
|
||||
return {self.output: gen}
|
||||
|
|
|
|||
Loading…
Reference in New Issue
Block a user