a-One-Fan/DL-Art-School
1
0
Fork 0
forked from mrq/DL-Art-School
Code Pull Requests Activity

Allow flat dvae

Browse Source
This commit is contained in:
James Betker 2021-11-18 00:53:42 -07:00
parent f3db41f125
commit 019acfa4c5
1 changed files with 30 additions and 23 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

53
codes/models/gpt_voice/lucidrains_dvae.py
View File

@ -78,7 +78,6 @@ class DiscreteVAE(nn.Module):
discretization_loss_averaging_steps = 100, discretization_loss_averaging_steps = 100,
): ):
super().__init__() super().__init__()
assert num_layers >= 1, 'number of layers must be greater than or equal to 1'
has_resblocks = num_resnet_blocks > 0 has_resblocks = num_resnet_blocks > 0
self.num_tokens = num_tokens self.num_tokens = num_tokens
@ -106,35 +105,43 @@ class DiscreteVAE(nn.Module):
assert NotImplementedError() assert NotImplementedError()
enc_chans = [hidden_dim * 2 ** i for i in range(num_layers)]
dec_chans = list(reversed(enc_chans))
enc_chans = [channels, *enc_chans]
dec_init_chan = codebook_dim if not has_resblocks else dec_chans[0]
dec_chans = [dec_init_chan, *dec_chans]
enc_chans_io, dec_chans_io = map(lambda t: list(zip(t[:-1], t[1:])), (enc_chans, dec_chans))
enc_layers = [] enc_layers = []
dec_layers = [] dec_layers = []
pad = (kernel_size - 1) // 2 if num_layers > 0:
for (enc_in, enc_out), (dec_in, dec_out) in zip(enc_chans_io, dec_chans_io): enc_chans = [hidden_dim * 2 ** i for i in range(num_layers)]
enc_layers.append(nn.Sequential(conv(enc_in, enc_out, kernel_size, stride = stride, padding = pad), act())) dec_chans = list(reversed(enc_chans))
if encoder_norm:
enc_layers.append(nn.GroupNorm(8, enc_out)) enc_chans = [channels, *enc_chans]
dec_layers.append(nn.Sequential(conv_transpose(dec_in, dec_out, kernel_size, stride = stride, padding = pad), act()))
dec_init_chan = codebook_dim if not has_resblocks else dec_chans[0]
dec_chans = [dec_init_chan, *dec_chans]
enc_chans_io, dec_chans_io = map(lambda t: list(zip(t[:-1], t[1:])), (enc_chans, dec_chans))
pad = (kernel_size - 1) // 2
for (enc_in, enc_out), (dec_in, dec_out) in zip(enc_chans_io, dec_chans_io):
enc_layers.append(nn.Sequential(conv(enc_in, enc_out, kernel_size, stride = stride, padding = pad), act()))
if encoder_norm:
enc_layers.append(nn.GroupNorm(8, enc_out))
dec_layers.append(nn.Sequential(conv_transpose(dec_in, dec_out, kernel_size, stride = stride, padding = pad), act()))
dec_out_chans = dec_chans[-1]
innermost_dim = dec_chans[0]
else:
enc_layers.append(nn.Sequential(conv(channels, hidden_dim, 1), act()))
dec_out_chans = hidden_dim
innermost_dim = hidden_dim
for _ in range(num_resnet_blocks): for _ in range(num_resnet_blocks):
dec_layers.insert(0, ResBlock(dec_chans[1], conv, act)) dec_layers.insert(0, ResBlock(innermost_dim, conv, act))
enc_layers.append(ResBlock(enc_chans[-1], conv, act)) enc_layers.append(ResBlock(innermost_dim, conv, act))
if num_resnet_blocks > 0: if num_resnet_blocks > 0:
dec_layers.insert(0, conv(codebook_dim, dec_chans[1], 1)) dec_layers.insert(0, conv(codebook_dim, innermost_dim, 1))
enc_layers.append(conv(enc_chans[-1], codebook_dim, 1)) enc_layers.append(conv(innermost_dim, codebook_dim, 1))
dec_layers.append(conv(dec_chans[-1], channels, 1)) dec_layers.append(conv(dec_out_chans, channels, 1))
self.encoder = nn.Sequential(*enc_layers) self.encoder = nn.Sequential(*enc_layers)
self.decoder = nn.Sequential(*dec_layers) self.decoder = nn.Sequential(*dec_layers)
@ -258,7 +265,7 @@ if __name__ == '__main__':
#o=v(torch.randn(1,3,256,256)) #o=v(torch.randn(1,3,256,256))
#print(o.shape) #print(o.shape)
v = DiscreteVAE(channels=80, normalization=None, positional_dims=1, num_tokens=4096, codebook_dim=4096, v = DiscreteVAE(channels=80, normalization=None, positional_dims=1, num_tokens=4096, codebook_dim=4096,
hidden_dim=256, stride=2, num_resnet_blocks=2, kernel_size=3, num_layers=2, use_transposed_convs=False) hidden_dim=256, stride=2, num_resnet_blocks=2, kernel_size=3, num_layers=0, use_transposed_convs=False)
#v.eval() #v.eval()
o=v(torch.randn(1,80,256)) o=v(torch.randn(1,80,256))
print(o[-1].shape) print(o[-1].shape)