File size: 11,297 Bytes
1e4a2ab |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 |
import os
import sys
import onnx
import torch
import platform
import warnings
import onnx2torch
import numpy as np
import onnxruntime as ort
from tqdm import tqdm
sys.path.append(os.getcwd())
from main.library.uvr5_lib import spec_utils
from main.library.uvr5_lib.common_separator import CommonSeparator
warnings.filterwarnings("ignore")
class MDXSeparator(CommonSeparator):
def __init__(self, common_config, arch_config):
super().__init__(config=common_config)
self.segment_size = arch_config.get("segment_size")
self.overlap = arch_config.get("overlap")
self.batch_size = arch_config.get("batch_size", 1)
self.hop_length = arch_config.get("hop_length")
self.enable_denoise = arch_config.get("enable_denoise")
self.compensate = self.model_data["compensate"]
self.dim_f = self.model_data["mdx_dim_f_set"]
self.dim_t = 2 ** self.model_data["mdx_dim_t_set"]
self.n_fft = self.model_data["mdx_n_fft_scale_set"]
self.config_yaml = self.model_data.get("config_yaml", None)
self.load_model()
self.n_bins = 0
self.trim = 0
self.chunk_size = 0
self.gen_size = 0
self.stft = None
self.primary_source = None
self.secondary_source = None
self.audio_file_path = None
self.audio_file_base = None
def load_model(self):
if self.segment_size == self.dim_t:
ort_session_options = ort.SessionOptions()
ort_session_options.log_severity_level = 3
ort_inference_session = ort.InferenceSession(self.model_path, providers=self.onnx_execution_provider, sess_options=ort_session_options)
self.model_run = lambda spek: ort_inference_session.run(None, {"input": spek.cpu().numpy()})[0]
else:
self.model_run = onnx2torch.convert(onnx.load(self.model_path)) if platform.system() == 'Windows' else onnx2torch.convert(self.model_path)
self.model_run.to(self.torch_device).eval()
def separate(self, audio_file_path):
self.audio_file_path = audio_file_path
self.audio_file_base = os.path.splitext(os.path.basename(audio_file_path))[0]
mix = self.prepare_mix(self.audio_file_path)
mix = spec_utils.normalize(wave=mix, max_peak=self.normalization_threshold)
source = self.demix(mix)
output_files = []
if not isinstance(self.primary_source, np.ndarray):
self.primary_source = spec_utils.normalize(wave=source, max_peak=self.normalization_threshold).T
if not isinstance(self.secondary_source, np.ndarray):
raw_mix = self.demix(mix, is_match_mix=True)
if self.invert_using_spec:
self.secondary_source = spec_utils.invert_stem(raw_mix, source)
else:
self.secondary_source = mix.T - source.T
if not self.output_single_stem or self.output_single_stem.lower() == self.secondary_stem_name.lower():
self.secondary_stem_output_path = os.path.join(f"{self.audio_file_base}_({self.secondary_stem_name})_{self.model_name}.{self.output_format.lower()}")
self.final_process(self.secondary_stem_output_path, self.secondary_source, self.secondary_stem_name)
output_files.append(self.secondary_stem_output_path)
if not self.output_single_stem or self.output_single_stem.lower() == self.primary_stem_name.lower():
self.primary_stem_output_path = os.path.join(f"{self.audio_file_base}_({self.primary_stem_name})_{self.model_name}.{self.output_format.lower()}")
if not isinstance(self.primary_source, np.ndarray): self.primary_source = source.T
self.final_process(self.primary_stem_output_path, self.primary_source, self.primary_stem_name)
output_files.append(self.primary_stem_output_path)
return output_files
def initialize_model_settings(self):
self.n_bins = self.n_fft // 2 + 1
self.trim = self.n_fft // 2
self.chunk_size = self.hop_length * (self.segment_size - 1)
self.gen_size = self.chunk_size - 2 * self.trim
self.stft = STFT(self.n_fft, self.hop_length, self.dim_f, self.torch_device)
def initialize_mix(self, mix, is_ckpt=False):
if is_ckpt:
pad = self.gen_size + self.trim - (mix.shape[-1] % self.gen_size)
mixture = np.concatenate((np.zeros((2, self.trim), dtype="float32"), mix, np.zeros((2, pad), dtype="float32")), 1)
num_chunks = mixture.shape[-1] // self.gen_size
mix_waves = [mixture[:, i * self.gen_size : i * self.gen_size + self.chunk_size] for i in range(num_chunks)]
else:
mix_waves = []
n_sample = mix.shape[1]
pad = self.gen_size - n_sample % self.gen_size
mix_p = np.concatenate((np.zeros((2, self.trim)), mix, np.zeros((2, pad)), np.zeros((2, self.trim))), 1)
i = 0
while i < n_sample + pad:
mix_waves.append(np.array(mix_p[:, i : i + self.chunk_size]))
i += self.gen_size
mix_waves_tensor = torch.tensor(mix_waves, dtype=torch.float32).to(self.torch_device)
return mix_waves_tensor, pad
def demix(self, mix, is_match_mix=False):
self.initialize_model_settings()
tar_waves_ = []
if is_match_mix:
chunk_size = self.hop_length * (self.segment_size - 1)
overlap = 0.02
else:
chunk_size = self.chunk_size
overlap = self.overlap
gen_size = chunk_size - 2 * self.trim
mixture = np.concatenate((np.zeros((2, self.trim), dtype="float32"), mix, np.zeros((2, gen_size + self.trim - ((mix.shape[-1]) % gen_size)), dtype="float32")), 1)
step = int((1 - overlap) * chunk_size)
result = np.zeros((1, 2, mixture.shape[-1]), dtype=np.float32)
divider = np.zeros((1, 2, mixture.shape[-1]), dtype=np.float32)
total = 0
for i in tqdm(range(0, mixture.shape[-1], step), ncols=100, unit="f"):
total += 1
start = i
end = min(i + chunk_size, mixture.shape[-1])
chunk_size_actual = end - start
window = None
if overlap != 0:
window = np.hanning(chunk_size_actual)
window = np.tile(window[None, None, :], (1, 2, 1))
mix_part_ = mixture[:, start:end]
if end != i + chunk_size:
pad_size = (i + chunk_size) - end
mix_part_ = np.concatenate((mix_part_, np.zeros((2, pad_size), dtype="float32")), axis=-1)
mix_waves = torch.tensor([mix_part_], dtype=torch.float32).to(self.torch_device).split(self.batch_size)
with torch.no_grad():
batches_processed = 0
for mix_wave in mix_waves:
batches_processed += 1
tar_waves = self.run_model(mix_wave, is_match_mix=is_match_mix)
if window is not None:
tar_waves[..., :chunk_size_actual] *= window
divider[..., start:end] += window
else: divider[..., start:end] += 1
result[..., start:end] += tar_waves[..., : end - start]
tar_waves = result / divider
tar_waves_.append(tar_waves)
tar_waves = np.concatenate(np.vstack(tar_waves_)[:, :, self.trim : -self.trim], axis=-1)[:, : mix.shape[-1]]
source = tar_waves[:, 0:None]
if not is_match_mix:
source *= self.compensate
return source
def run_model(self, mix, is_match_mix=False):
spek = self.stft(mix.to(self.torch_device))
spek[:, :, :3, :] *= 0
if is_match_mix:
spec_pred = spek.cpu().numpy()
else:
if self.enable_denoise:
spec_pred_neg = self.model_run(-spek)
spec_pred_pos = self.model_run(spek)
spec_pred = (spec_pred_neg * -0.5) + (spec_pred_pos * 0.5)
else:
spec_pred = self.model_run(spek)
result = self.stft.inverse(torch.tensor(spec_pred).to(self.torch_device)).cpu().detach().numpy()
return result
class STFT:
def __init__(self, n_fft, hop_length, dim_f, device):
self.n_fft = n_fft
self.hop_length = hop_length
self.dim_f = dim_f
self.device = device
self.hann_window = torch.hann_window(window_length=self.n_fft, periodic=True)
def __call__(self, input_tensor):
is_non_standard_device = not input_tensor.device.type in ["cuda", "cpu"]
if is_non_standard_device: input_tensor = input_tensor.cpu()
batch_dimensions = input_tensor.shape[:-2]
channel_dim, time_dim = input_tensor.shape[-2:]
permuted_stft_output = torch.stft(input_tensor.reshape([-1, time_dim]), n_fft=self.n_fft, hop_length=self.hop_length, window=self.hann_window.to(input_tensor.device), center=True, return_complex=False).permute([0, 3, 1, 2])
final_output = permuted_stft_output.reshape([*batch_dimensions, channel_dim, 2, -1, permuted_stft_output.shape[-1]]).reshape([*batch_dimensions, channel_dim * 2, -1, permuted_stft_output.shape[-1]])
if is_non_standard_device: final_output = final_output.to(self.device)
return final_output[..., : self.dim_f, :]
def pad_frequency_dimension(self, input_tensor, batch_dimensions, channel_dim, freq_dim, time_dim, num_freq_bins):
return torch.cat([input_tensor, torch.zeros([*batch_dimensions, channel_dim, num_freq_bins - freq_dim, time_dim]).to(input_tensor.device)], -2)
def calculate_inverse_dimensions(self, input_tensor):
channel_dim, freq_dim, time_dim = input_tensor.shape[-3:]
return input_tensor.shape[:-3], channel_dim, freq_dim, time_dim, self.n_fft // 2 + 1
def prepare_for_istft(self, padded_tensor, batch_dimensions, channel_dim, num_freq_bins, time_dim):
permuted_tensor = padded_tensor.reshape([*batch_dimensions, channel_dim // 2, 2, num_freq_bins, time_dim]).reshape([-1, 2, num_freq_bins, time_dim]).permute([0, 2, 3, 1])
return permuted_tensor[..., 0] + permuted_tensor[..., 1] * 1.0j
def inverse(self, input_tensor):
is_non_standard_device = not input_tensor.device.type in ["cuda", "cpu"]
if is_non_standard_device: input_tensor = input_tensor.cpu()
batch_dimensions, channel_dim, freq_dim, time_dim, num_freq_bins = self.calculate_inverse_dimensions(input_tensor)
final_output = torch.istft(self.prepare_for_istft(self.pad_frequency_dimension(input_tensor, batch_dimensions, channel_dim, freq_dim, time_dim, num_freq_bins), batch_dimensions, channel_dim, num_freq_bins, time_dim), n_fft=self.n_fft, hop_length=self.hop_length, window=self.hann_window.to(input_tensor.device), center=True).reshape([*batch_dimensions, 2, -1])
if is_non_standard_device: final_output = final_output.to(self.device)
return final_output |