Spaces:
Running
Running
File size: 2,818 Bytes
3aa4060 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 |
import torch
import numpy as np
import inspect
def sequence_mask(length, max_length=None):
if max_length is None:
max_length = length.max()
x = torch.arange(int(max_length), dtype=length.dtype, device=length.device)
return x.unsqueeze(0) < length.unsqueeze(1)
def fix_len_compatibility(length, num_downsamplings_in_unet=2):
while True:
if length % (2**num_downsamplings_in_unet) == 0:
return length
length += 1
def convert_pad_shape(pad_shape):
l = pad_shape[::-1]
pad_shape = [item for sublist in l for item in sublist]
return pad_shape
def generate_path(duration, mask):
device = duration.device
b, t_x, t_y = mask.shape
cum_duration = torch.cumsum(duration, 1)
path = torch.zeros(b, t_x, t_y, dtype=mask.dtype).to(device=device)
cum_duration_flat = cum_duration.view(b * t_x)
path = sequence_mask(cum_duration_flat, t_y).to(mask.dtype)
path = path.view(b, t_x, t_y)
path = path - torch.nn.functional.pad(path, convert_pad_shape([[0, 0],
[1, 0], [0, 0]]))[:, :-1]
path = path * mask
return path
def duration_loss(logw, logw_, lengths):
loss = torch.sum((logw - logw_)**2) / torch.sum(lengths)
return loss
f0_bin = 256
f0_max = 1100.0
f0_min = 50.0
f0_mel_min = 1127 * np.log(1 + f0_min / 700)
f0_mel_max = 1127 * np.log(1 + f0_max / 700)
def f0_to_coarse(f0):
is_torch = isinstance(f0, torch.Tensor)
f0_mel = 1127 * (1 + f0 / 700).log() if is_torch else 1127 * \
np.log(1 + f0 / 700)
f0_mel[f0_mel > 0] = (f0_mel[f0_mel > 0] - f0_mel_min) * \
(f0_bin - 2) / (f0_mel_max - f0_mel_min) + 1
f0_mel[f0_mel <= 1] = 1
f0_mel[f0_mel > f0_bin - 1] = f0_bin - 1
f0_coarse = (
f0_mel + 0.5).long() if is_torch else np.rint(f0_mel).astype(np.int)
assert f0_coarse.max() <= 255 and f0_coarse.min(
) >= 1, (f0_coarse.max(), f0_coarse.min())
return f0_coarse
def rand_ids_segments(lengths, segment_size=200):
b = lengths.shape[0]
ids_str_max = lengths - segment_size
ids_str = (torch.rand([b]).to(device=lengths.device) * ids_str_max).to(dtype=torch.long)
return ids_str
def slice_segments(x, ids_str, segment_size=200):
ret = torch.zeros_like(x[:, :, :segment_size])
for i in range(x.size(0)):
idx_str = ids_str[i]
idx_end = idx_str + segment_size
ret[i] = x[i, :, idx_str:idx_end]
return ret
def retrieve_name(var):
for fi in reversed(inspect.stack()):
names = [var_name for var_name,
var_val in fi.frame.f_locals.items() if var_val is var]
if len(names) > 0:
return names[0]
Debug_Enable = True
def debug_shapes(var):
if Debug_Enable:
print(retrieve_name(var), var.shape)
|