-
Notifications
You must be signed in to change notification settings - Fork 0
/
infer-web.py
2294 lines (1984 loc) · 99.5 KB
/
infer-web.py
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
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
import sys
from shutil import rmtree
import shutil
import json # Mangio fork using json for preset saving
from glob import glob1
from signal import SIGTERM
import os
now_dir = os.getcwd()
sys.path.append(now_dir)
import lib.globals.globals as rvc_globals
from LazyImport import lazyload
math = lazyload('math')
import traceback
import warnings
tensorlowest = lazyload('tensorlowest')
import faiss
ffmpeg = lazyload('ffmpeg')
np = lazyload("numpy")
torch = lazyload('torch')
re = lazyload('regex')
os.environ["OPENBLAS_NUM_THREADS"] = "1"
os.environ["no_proxy"] = "localhost, 127.0.0.1, ::1"
import logging
from random import shuffle
from subprocess import Popen
gr = lazyload("gradio")
SF = lazyload("soundfile")
SFWrite = SF.write
from config import Config
from fairseq import checkpoint_utils
from i18n import I18nAuto
from lib.infer_pack.models import (
SynthesizerTrnMs256NSFsid,
SynthesizerTrnMs256NSFsid_nono,
SynthesizerTrnMs768NSFsid,
SynthesizerTrnMs768NSFsid_nono,
)
from lib.infer_pack.models_onnx import SynthesizerTrnMsNSFsidM
from infer_uvr5 import _audio_pre_, _audio_pre_new
from MDXNet import MDXNetDereverb
from my_utils import load_audio
from train.process_ckpt import change_info, extract_small_model, merge, show_info
from vc_infer_pipeline import VC
from sklearn.cluster import MiniBatchKMeans
import time
import threading
from shlex import quote as SQuote
RQuote = lambda val: SQuote(str(val))
tmp = os.path.join(now_dir, "TEMP")
runtime_dir = os.path.join(now_dir, "runtime/Lib/site-packages")
directories = ['logs', 'audios', 'datasets', 'weights']
rmtree(tmp, ignore_errors=True)
rmtree(os.path.join(runtime_dir, "infer_pack"), ignore_errors=True)
rmtree(os.path.join(runtime_dir, "uvr5_pack"), ignore_errors=True)
os.makedirs(tmp, exist_ok=True)
for folder in directories:
os.makedirs(os.path.join(now_dir, folder), exist_ok=True)
os.environ["TEMP"] = tmp
warnings.filterwarnings("ignore")
torch.manual_seed(114514)
logging.getLogger("numba").setLevel(logging.WARNING)
try:
file = open('csvdb/stop.csv', 'x')
file.close()
except FileExistsError: pass
global DoFormant, Quefrency, Timbre
DoFormant = rvc_globals.DoFormant
Quefrency = rvc_globals.Quefrency
Timbre = rvc_globals.Timbre
config = Config()
i18n = I18nAuto()
i18n.print()
# 判断是否有能用来训练和加速推理的N卡
ngpu = torch.cuda.device_count()
gpu_infos = []
mem = []
if_gpu_ok = False
keywords = ["10", "16", "20", "30", "40", "A2", "A3", "A4", "P4", "A50", "500", "A60",
"70", "80", "90", "M4", "T4", "TITAN"]
if torch.cuda.is_available() or ngpu != 0:
for i in range(ngpu):
gpu_name = torch.cuda.get_device_name(i).upper()
if any(keyword in gpu_name for keyword in keywords):
if_gpu_ok = True # 至少有一张能用的N卡
gpu_infos.append("%s\t%s" % (i, gpu_name))
mem.append(int(torch.cuda.get_device_properties(i).total_memory / 1e9 + 0.4))
gpu_info = "\n".join(gpu_infos) if if_gpu_ok and gpu_infos else i18n("很遗憾您这没有能用的显卡来支持您训练")
default_batch_size = min(mem) // 2 if if_gpu_ok and gpu_infos else 1
gpus = "-".join(i[0] for i in gpu_infos)
hubert_model = None
def load_hubert():
global hubert_model
models, _, _ = checkpoint_utils.load_model_ensemble_and_task(["hubert_base.pt"], suffix="")
hubert_model = models[0].to(config.device)
if config.is_half:
hubert_model = hubert_model.half()
hubert_model.eval()
datasets_root = "datasets"
weight_root = "weights"
weight_uvr5_root = "uvr5_weights"
index_root = "logs"
fshift_root = "formantshiftcfg"
audio_root = "audios"
sup_audioext = {'wav', 'mp3', 'flac', 'ogg', 'opus',
'm4a', 'mp4', 'aac', 'alac', 'wma',
'aiff', 'webm', 'ac3'}
names = [os.path.join(root, file)
for root, _, files in os.walk(weight_root)
for file in files
if file.endswith((".pth", ".onnx"))]
indexes_list = [os.path.join(root, name)
for root, _, files in os.walk(index_root, topdown=False)
for name in files
if name.endswith(".index") and "trained" not in name]
audio_paths = [os.path.join(root, name)
for root, _, files in os.walk(audio_root, topdown=False)
for name in files
if name.endswith(tuple(sup_audioext))]
uvr5_names = [name.replace(".pth", "")
for name in os.listdir(weight_uvr5_root)
if name.endswith(".pth") or "onnx" in name]
check_for_name = lambda: sorted(names)[0] if names else ''
def get_indexes():
indexes_list = [
os.path.join(dirpath, filename)
for dirpath, _, filenames in os.walk(index_root)
for filename in filenames
if filename.endswith(".index") and "trained" not in filename
]
return indexes_list if indexes_list else ''
def get_fshift_presets():
fshift_presets_list = [
os.path.join(dirpath, filename)
for dirpath, _, filenames in os.walk(fshift_root)
for filename in filenames
if filename.endswith(".txt")
]
return fshift_presets_list if fshift_presets_list else ''
def vc_single(
sid: str,
input_audio_path0: str,
input_audio_path1: str,
f0_up_key: int,
f0_file: str,
f0_method: str,
file_index: str,
file_index2: str,
index_rate: float,
filter_radius: int,
resample_sr: int,
rms_mix_rate: float,
protect: float,
crepe_hop_length: int,
f0_min: int,
note_min: str,
f0_max: int,
note_max: str,
):
global total_time
total_time = 0
start_time = time.time()
global tgt_sr, net_g, vc, hubert_model, version
if not input_audio_path0 and not input_audio_path1:
return "You need to upload an audio", None
if (not os.path.exists(input_audio_path0)) and (not os.path.exists(os.path.join(now_dir, input_audio_path0))):
return "Audio was not properly selected or doesn't exist", None
# This might be jank, but I'm trying to make sure this gets the right file...
input_audio_path1 = input_audio_path1 or input_audio_path0
print(f"\nStarting inference for '{os.path.basename(input_audio_path1)}'")
print("-------------------")
f0_up_key = int(f0_up_key)
if rvc_globals.NotesOrHertz and f0_method != 'rmvpe':
f0_min = note_to_hz(note_min) if note_min else 50
f0_max = note_to_hz(note_max) if note_max else 1100
print(f"Converted min pitch freq - {f0_min}\n"
f"Converted max pitch freq - {f0_max}")
else:
f0_min = f0_min or 50
f0_max = f0_max or 1100
try:
print(f"Attempting to load {input_audio_path1}....")
audio = load_audio(input_audio_path1,
16000,
DoFormant=rvc_globals.DoFormant,
Quefrency=rvc_globals.Quefrency,
Timbre=rvc_globals.Timbre)
audio_max = np.abs(audio).max() / 0.95
if audio_max > 1:
audio /= audio_max
times = [0, 0, 0]
if not hubert_model:
print("Loading HuBERT for the first time...")
load_hubert()
try:
if_f0 = cpt.get("f0", 1)
except NameError:
message = "Model was not properly selected"
print(message)
return message, None
file_index = (
file_index.strip(" ").strip('"').strip("\n").strip('"').strip(" ").replace("trained", "added")
) if file_index != "" else file_index2
try:
audio_opt = vc.pipeline(
hubert_model,
net_g,
sid,
audio,
input_audio_path1,
times,
f0_up_key,
f0_method,
file_index,
index_rate,
if_f0,
filter_radius,
tgt_sr,
resample_sr,
rms_mix_rate,
version,
protect,
crepe_hop_length,
f0_file=f0_file,
f0_min=f0_min,
f0_max=f0_max
)
except AssertionError:
message = "Mismatching index version detected (v1 with v2, or v2 with v1)."
print(message)
return message, None
except NameError:
message = "RVC libraries are still loading. Please try again in a few seconds."
print(message)
return message, None
if tgt_sr != resample_sr >= 16000:
tgt_sr = resample_sr
index_info = "Using index:%s." % file_index if os.path.exists(file_index) else "Index not used."
end_time = time.time()
total_time = end_time - start_time
return f"Success.\n {index_info}\nTime:\n npy:{times[0]}, f0:{times[1]}, infer:{times[2]}\nTotal Time: {total_time} seconds", (tgt_sr, audio_opt)
except:
info = traceback.format_exc()
print(info)
return info, (None, None)
def vc_multi(
sid,
dir_path,
opt_root,
paths,
f0_up_key,
f0_method,
file_index,
file_index2,
# file_big_npy,
index_rate,
filter_radius,
resample_sr,
rms_mix_rate,
protect,
format1,
crepe_hop_length,
f0_min,
note_min,
f0_max,
note_max,
):
if rvc_globals.NotesOrHertz and f0_method != 'rmvpe':
f0_min = note_to_hz(note_min) if note_min else 50
f0_max = note_to_hz(note_max) if note_max else 1100
print(f"Converted min pitch freq - {f0_min}\n"
f"Converted max pitch freq - {f0_max}")
else:
f0_min = f0_min or 50
f0_max = f0_max or 1100
try:
dir_path, opt_root = [x.strip(" ").strip('"').strip("\n").strip('"').strip(" ") for x in [dir_path, opt_root]]
os.makedirs(opt_root, exist_ok=True)
paths = [os.path.join(dir_path, name) for name in os.listdir(dir_path)] if dir_path else [path.name for path in paths]
infos = []
for path in paths:
info, opt = vc_single(sid, path, None, f0_up_key, None, f0_method, file_index, file_index2, index_rate, filter_radius,
resample_sr, rms_mix_rate, protect, crepe_hop_length, f0_min, note_min, f0_max, note_max)
if "Success" in info:
try:
tgt_sr, audio_opt = opt
#sys.stdout.write(f"\nTarget Sample Rate (tgt_sr): {tgt_sr}") # Debugging print
base_name = os.path.splitext(os.path.basename(path))[0]
output_path = f"{opt_root}/{base_name}.{format1}"
path, extension = output_path, format1
path, extension = output_path if format1 in ["wav", "flac", "mp3", "ogg", "aac", "m4a"] else f"{output_path}.wav", format1
#sys.stdout.write(f"\nOutput Path: {path}") # Debugging print
#sys.stdout.write(f"\nFile Extension: {extension}") # Debugging print
SFWrite(path, audio_opt, tgt_sr)
#sys.stdout.write("\nFile Written Successfully with SFWrite") # Debugging print
if os.path.exists(path) and extension not in ["wav", "flac", "mp3", "ogg", "aac", "m4a"]:
sys.stdout.write(f"Running command: ffmpeg -i {RQuote(path)} -vn {RQuote(path[:-4] + '.' + extension)} -q:a 2 -y")
os.system(f"ffmpeg -i {RQuote(path)} -vn {RQuote(path[:-4] + '.' + extension)} -q:a 2 -y")
#print(f"\nFile Converted to {extension} using ffmpeg") # Debugging print
except:
info += traceback.format_exc()
print(f"\nException encountered: {info}") # Debugging print
infos.append(f"{os.path.basename(path)}->{info}")
yield "\n".join(infos)
yield "\n".join(infos)
except:
yield traceback.format_exc()
def uvr(model_name, inp_root, save_root_vocal, paths, save_root_ins, agg, format0):
infos = []
try:
inp_root, save_root_vocal, save_root_ins = [x.strip(" ").strip('"').strip("\n").strip('"').strip(" ") for x in [inp_root, save_root_vocal, save_root_ins]]
pre_fun = MDXNetDereverb(15) if model_name == "onnx_dereverb_By_FoxJoy" else (_audio_pre_ if "DeEcho" not in model_name else _audio_pre_new)(
agg=int(agg),
model_path=os.path.join(weight_uvr5_root, model_name + ".pth"),
device=config.device,
is_half=config.is_half,
)
paths = [os.path.join(inp_root, name) for name in os.listdir(inp_root)] if inp_root else [path.name for path in paths]
for path in paths:
inp_path = os.path.join(inp_root, path)
need_reformat, done = 1, 0
try:
info = ffmpeg.probe(inp_path, cmd="ffprobe")
if info["streams"][0]["channels"] == 2 and info["streams"][0]["sample_rate"] == "44100":
need_reformat = 0
pre_fun._path_audio_(inp_path, save_root_ins, save_root_vocal, format0)
done = 1
except:
traceback.print_exc()
if need_reformat:
tmp_path = f"{tmp}/{os.path.basename(RQuote(inp_path))}.reformatted.wav"
os.system(f"ffmpeg -i {RQuote(inp_path)} -vn -acodec pcm_s16le -ac 2 -ar 44100 {RQuote(tmp_path)} -y")
inp_path = tmp_path
try:
if not done:
pre_fun._path_audio_(inp_path, save_root_ins, save_root_vocal, format0)
infos.append(f"{os.path.basename(inp_path)}->Success")
yield "\n".join(infos)
except:
infos.append(f"{os.path.basename(inp_path)}->{traceback.format_exc()}")
yield "\n".join(infos)
except:
infos.append(traceback.format_exc())
yield "\n".join(infos)
finally:
try:
if model_name == "onnx_dereverb_By_FoxJoy":
del pre_fun.pred.model
del pre_fun.pred.model_
else:
del pre_fun.model
del pre_fun
except: traceback.print_exc()
print("clean_empty_cache")
if torch.cuda.is_available(): torch.cuda.empty_cache()
yield "\n".join(infos)
def get_vc(sid, to_return_protect0, to_return_protect1):
global n_spk, tgt_sr, net_g, vc, cpt, version, hubert_model
if not sid:
if hubert_model is not None:
print("clean_empty_cache")
del net_g, n_spk, vc, hubert_model, tgt_sr
hubert_model = net_g = n_spk = vc = hubert_model = tgt_sr = None
if torch.cuda.is_available():
torch.cuda.empty_cache()
if_f0, version = cpt.get("f0", 1), cpt.get("version", "v1")
net_g = (SynthesizerTrnMs256NSFsid if version == "v1" else SynthesizerTrnMs768NSFsid)(
*cpt["config"], is_half=config.is_half) if if_f0 == 1 else (SynthesizerTrnMs256NSFsid_nono if version == "v1" else SynthesizerTrnMs768NSFsid_nono)(*cpt["config"])
del net_g, cpt
if torch.cuda.is_available():
torch.cuda.empty_cache()
cpt = None
return ({"visible": False, "__type__": "update"},) * 3
print(f"loading {sid}")
cpt = torch.load(sid, map_location="cpu")
tgt_sr = cpt["config"][-1]
cpt["config"][-3] = cpt["weight"]["emb_g.weight"].shape[0]
if cpt.get("f0", 1) == 0:
to_return_protect0 = to_return_protect1 = {"visible": False, "value": 0.5, "__type__": "update"}
else:
to_return_protect0 = {"visible": True, "value": to_return_protect0, "__type__": "update"}
to_return_protect1 = {"visible": True, "value": to_return_protect1, "__type__": "update"}
version = cpt.get("version", "v1")
net_g = (SynthesizerTrnMs256NSFsid if version == "v1" else SynthesizerTrnMs768NSFsid)(
*cpt["config"], is_half=config.is_half) if cpt.get("f0", 1) == 1 else (SynthesizerTrnMs256NSFsid_nono if version == "v1" else SynthesizerTrnMs768NSFsid_nono)(*cpt["config"])
del net_g.enc_q
print(net_g.load_state_dict(cpt["weight"], strict=False))
net_g.eval().to(config.device)
net_g = net_g.half() if config.is_half else net_g.float()
vc = VC(tgt_sr, config)
n_spk = cpt["config"][-3]
return (
{"visible": False, "maximum": n_spk, "__type__": "update"},
to_return_protect0,
to_return_protect1
)
def change_choices():
names = [os.path.join(root, file)
for root, _, files in os.walk(weight_root)
for file in files
if file.endswith((".pth", ".onnx"))]
indexes_list = [os.path.join(root, name) for root, _, files in os.walk(index_root, topdown=False) for name in files if name.endswith(".index") and "trained" not in name]
audio_paths = [os.path.join(audio_root, file) for file in os.listdir(os.path.join(now_dir, "audios"))]
return (
{"choices": sorted(names), "__type__": "update"},
{"choices": sorted(indexes_list), "__type__": "update"},
{"choices": sorted(audio_paths), "__type__": "update"}
)
sr_dict = {
"32k": 32000,
"40k": 40000,
"48k": 48000,
}
def if_done(done, p):
while p.poll() is None:
time.sleep(0.5)
done[0] = True
def if_done_multi(done, ps):
while not all(p.poll() is not None for p in ps):
time.sleep(0.5)
done[0] = True
def formant_enabled(cbox, qfrency, tmbre):
global DoFormant, Quefrency, Timbre
DoFormant = cbox
Quefrency = qfrency
Timbre = tmbre
rvc_globals.DoFormant = cbox
rvc_globals.Quefrency = qfrency
rvc_globals.Timbre = tmbre
visibility_update = {"visible": DoFormant, "__type__": "update"}
return (
{"value": DoFormant, "__type__": "update"},
) + (visibility_update,) * 6
def formant_apply(qfrency, tmbre):
global Quefrency, Timbre, DoFormant
Quefrency = qfrency
Timbre = tmbre
DoFormant = True
rvc_globals.DoFormant = True
rvc_globals.Quefrency = qfrency
rvc_globals.Timbre = tmbre
return ({"value": Quefrency, "__type__": "update"}, {"value": Timbre, "__type__": "update"})
def update_fshift_presets(preset, qfrency, tmbre):
if preset:
with open(preset, 'r') as p:
content = p.readlines()
qfrency, tmbre = content[0].strip(), content[1]
formant_apply(qfrency, tmbre)
else:
qfrency, tmbre = preset_apply(preset, qfrency, tmbre)
return (
{"choices": get_fshift_presets(), "__type__": "update"},
{"value": qfrency, "__type__": "update"},
{"value": tmbre, "__type__": "update"},
)
def preprocess_dataset(trainset_dir, exp_dir, sr, n_p):
sr = sr_dict[sr]
log_dir = os.path.join(now_dir, "logs", exp_dir)
log_file = os.path.join(log_dir, "preprocess.log")
os.makedirs(log_dir, exist_ok=True)
with open(log_file, "w") as f: pass
cmd = (
f"{config.python_cmd} "
"trainset_preprocess_pipeline_print.py "
f"{trainset_dir} "
f"{RQuote(sr)} "
f"{RQuote(n_p)} "
f"{log_dir} "
f"{RQuote(config.noparallel)}"
)
print(cmd)
p = Popen(cmd, shell=True)
done = [False]
threading.Thread(target=if_done, args=(done,p,)).start()
while not done[0]:
with open(log_file, "r") as f:
yield f.read()
time.sleep(1)
with open(log_file, "r") as f:
log = f.read()
print(log)
yield log
def extract_f0_feature(gpus, n_p, f0method, if_f0, exp_dir, version19, echl):
gpus = gpus.split("-")
log_dir = f"{now_dir}/logs/{exp_dir}"
log_file = f"{log_dir}/extract_f0_feature.log"
os.makedirs(log_dir, exist_ok=True)
with open(log_file, "w") as f: pass
if if_f0:
cmd = (
f"{config.python_cmd} extract_f0_print.py {log_dir} "
f"{RQuote(n_p)} {RQuote(f0method)} {RQuote(echl)}"
)
print(cmd)
p = Popen(cmd, shell=True, cwd=now_dir)
done = [False]
threading.Thread(target=if_done, args=(done, p)).start()
while not done[0]:
with open(log_file, "r") as f:
yield f.read()
time.sleep(1)
leng = len(gpus)
ps = []
for idx, n_g in enumerate(gpus):
cmd = (
f"{config.python_cmd} extract_feature_print.py {RQuote(config.device)} "
f"{RQuote(leng)} {RQuote(idx)} {RQuote(n_g)} {log_dir} {RQuote(version19)}"
)
print(cmd)
p = Popen(cmd, shell=True, cwd=now_dir)
ps.append(p)
done = [False]
threading.Thread(target=if_done_multi, args=(done, ps)).start()
while not done[0]:
with open(log_file, "r") as f:
yield f.read()
time.sleep(1)
with open(log_file, "r") as f:
log = f.read()
print(log)
yield log
def change_sr2(sr2, if_f0_3, version19):
path_str = "" if version19 == "v1" else "_v2"
f0_str = "f0" if if_f0_3 else ""
model_paths = {"G": "", "D": ""}
for model_type in model_paths:
file_path = f"pretrained{path_str}/{f0_str}{model_type}{sr2}.pth"
if os.access(file_path, os.F_OK):
model_paths[model_type] = file_path
else:
print(f"{file_path} doesn't exist, will not use pretrained model.")
return (model_paths["G"], model_paths["D"])
def change_version19(sr2, if_f0_3, version19):
path_str = "" if version19 == "v1" else "_v2"
sr2 = "40k" if (sr2 == "32k" and version19 == "v1") else sr2
choices_update = {
"choices": ["40k", "48k"], "__type__": "update", "value": sr2
} if version19 == "v1" else {
"choices": ["40k", "48k", "32k"], "__type__": "update", "value": sr2}
f0_str = "f0" if if_f0_3 else ""
model_paths = {"G": "", "D": ""}
for model_type in model_paths:
file_path = f"pretrained{path_str}/{f0_str}{model_type}{sr2}.pth"
if os.access(file_path, os.F_OK):
model_paths[model_type] = file_path
else:
print(f"{file_path} doesn't exist, will not use pretrained model.")
return (model_paths["G"], model_paths["D"], choices_update)
def change_f0(if_f0_3, sr2, version19): # f0method8,pretrained_G14,pretrained_D15
path_str = "" if version19 == "v1" else "_v2"
pth_format = "pretrained%s/f0%s%s.pth"
model_desc = { "G": "", "D": "" }
for model_type in model_desc:
file_path = pth_format % (path_str, model_type, sr2)
if os.access(file_path, os.F_OK):
model_desc[model_type] = file_path
else:
print(file_path, "doesn't exist, will not use pretrained model")
return (
{"visible": if_f0_3, "__type__": "update"},
model_desc["G"],
model_desc["D"],
{"visible": if_f0_3, "__type__": "update"}
)
global log_interval
def set_log_interval(exp_dir, batch_size12):
log_interval = 1
folder_path = os.path.join(exp_dir, "1_16k_wavs")
if os.path.isdir(folder_path):
wav_files_num = len(glob1(folder_path,"*.wav"))
if wav_files_num > 0:
log_interval = math.ceil(wav_files_num / batch_size12)
if log_interval > 1:
log_interval += 1
return log_interval
global PID, PROCESS
def click_train(
exp_dir1,
sr2,
if_f0_3,
spk_id5,
save_epoch10,
total_epoch11,
batch_size12,
if_save_latest13,
pretrained_G14,
pretrained_D15,
gpus16,
if_cache_gpu17,
if_save_every_weights18,
version19,
):
with open('csvdb/stop.csv', 'w+') as file: file.write("False")
log_dir = os.path.join(now_dir, "logs", exp_dir1)
os.makedirs(log_dir, exist_ok=True)
gt_wavs_dir = os.path.join(log_dir, "0_gt_wavs")
feature_dim = "256" if version19 == "v1" else "768"
feature_dir = os.path.join(log_dir, f"3_feature{feature_dim}")
log_interval = set_log_interval(log_dir, batch_size12)
required_dirs = [gt_wavs_dir, feature_dir]
if if_f0_3:
f0_dir = f"{log_dir}/2a_f0"
f0nsf_dir = f"{log_dir}/2b-f0nsf"
required_dirs.extend([f0_dir, f0nsf_dir])
names = set(name.split(".")[0] for directory in required_dirs for name in os.listdir(directory))
def generate_paths(name):
paths = [gt_wavs_dir, feature_dir]
if if_f0_3:
paths.extend([f0_dir, f0nsf_dir])
return '|'.join([path.replace('\\', '\\\\') + '/' + name + ('.wav.npy' if path in [f0_dir, f0nsf_dir] else '.wav' if path == gt_wavs_dir else '.npy') for path in paths])
opt = [f"{generate_paths(name)}|{spk_id5}" for name in names]
mute_dir = f"{now_dir}/logs/mute"
for _ in range(2):
mute_string = f"{mute_dir}/0_gt_wavs/mute{sr2}.wav|{mute_dir}/3_feature{feature_dim}/mute.npy"
if if_f0_3:
mute_string += f"|{mute_dir}/2a_f0/mute.wav.npy|{mute_dir}/2b-f0nsf/mute.wav.npy"
opt.append(mute_string+f"|{spk_id5}")
shuffle(opt)
with open(f"{log_dir}/filelist.txt", "w") as f:
f.write("\n".join(opt))
print("write filelist done")
print("use gpus:", gpus16)
if pretrained_G14 == "":
print("no pretrained Generator")
if pretrained_D15 == "":
print("no pretrained Discriminator")
G_train = f"-pg {pretrained_G14}" if pretrained_G14 else ""
D_train = f"-pd {pretrained_D15}" if pretrained_D15 else ""
cmd = (
f"{config.python_cmd} train_nsf_sim_cache_sid_load_pretrain.py -e {exp_dir1} -sr {sr2} -f0 {int(if_f0_3)} -bs {batch_size12}"
f" -g {gpus16 if gpus16 is not None else ''} -te {total_epoch11} -se {save_epoch10} {G_train} {D_train} -l {int(if_save_latest13)}"
f" -c {int(if_cache_gpu17)} -sw {int(if_save_every_weights18)} -v {version19} -li {log_interval}"
)
print(cmd)
global p
p = Popen(cmd, shell=True, cwd=now_dir)
global PID
PID = p.pid
p.wait()
return "Training is done, check train.log", {"visible": False, "__type__": "update"}, {"visible": True, "__type__": "update"}
def train_index(exp_dir1, version19):
exp_dir = os.path.join(now_dir, 'logs', exp_dir1)
os.makedirs(exp_dir, exist_ok=True)
feature_dim = '256' if version19 == "v1" else '768'
feature_dir = os.path.join(exp_dir, f"3_feature{feature_dim}")
if not os.path.exists(feature_dir) or len(os.listdir(feature_dir)) == 0:
return "请先进行特征提取!"
npys = [np.load(os.path.join(feature_dir, name)) for name in sorted(os.listdir(feature_dir))]
big_npy = np.concatenate(npys, 0)
np.random.shuffle(big_npy)
infos = []
if big_npy.shape[0] > 2*10**5:
infos.append("Trying doing kmeans %s shape to 10k centers." % big_npy.shape[0])
yield "\n".join(infos)
try:
big_npy = MiniBatchKMeans(n_clusters=10000, verbose=True, batch_size=256 * config.n_cpu,
compute_labels=False,init="random").fit(big_npy).cluster_centers_
except Exception as e:
infos.append(str(e))
yield "\n".join(infos)
np.save(os.path.join(exp_dir, "total_fea.npy"), big_npy)
n_ivf = min(int(16 * np.sqrt(big_npy.shape[0])), big_npy.shape[0] // 39)
infos.append("%s,%s" % (big_npy.shape, n_ivf))
yield "\n".join(infos)
index = faiss.index_factory(int(feature_dim), f"IVF{n_ivf},Flat")
index_ivf = faiss.extract_index_ivf(index)
index_ivf.nprobe = 1
index.train(big_npy)
index_file_base = f"{exp_dir}/trained_IVF{n_ivf}_Flat_nprobe_{index_ivf.nprobe}_{exp_dir1}_{version19}.index"
faiss.write_index(index, index_file_base)
infos.append("adding")
yield "\n".join(infos)
batch_size_add = 8192
for i in range(0, big_npy.shape[0], batch_size_add):
index.add(big_npy[i:i + batch_size_add])
index_file_base = f"{exp_dir}/added_IVF{n_ivf}_Flat_nprobe_{index_ivf.nprobe}_{exp_dir1}_{version19}.index"
faiss.write_index(index, index_file_base)
infos.append(f"Successful Index Construction，added_IVF{n_ivf}_Flat_nprobe_{index_ivf.nprobe}_{exp_dir1}_{version19}.index")
yield "\n".join(infos)
#def setBoolean(status): #true to false and vice versa / not implemented yet, dont touch!!!!!!!
# status = not status
# return status
def change_info_(ckpt_path):
train_log_path = os.path.join(os.path.dirname(ckpt_path), "train.log")
if not os.path.exists(train_log_path):
return {"__type__": "update"}, {"__type__": "update"}, {"__type__": "update"}
try:
with open(train_log_path, "r") as f:
info_line = next(f).strip()
info = eval(info_line.split("\t")[-1])
sr, f0 = info.get("sample_rate"), info.get("if_f0")
version = "v2" if info.get("version") == "v2" else "v1"
return sr, str(f0), version
except Exception as e:
print(f"Exception occurred: {str(e)}, Traceback: {traceback.format_exc()}")
return {"__type__": "update"}, {"__type__": "update"}, {"__type__": "update"}
def export_onnx(model_path, exported_path):
device = torch.device("cpu")
checkpoint = torch.load(model_path, map_location=device)
vec_channels = 256 if checkpoint.get("version", "v1") == "v1" else 768
test_inputs = {
"phone": torch.rand(1, 200, vec_channels),
"phone_lengths": torch.LongTensor([200]),
"pitch": torch.randint(5, 255, (1, 200)),
"pitchf": torch.rand(1, 200),
"ds": torch.zeros(1).long(),
"rnd": torch.rand(1, 192, 200)
}
checkpoint["config"][-3] = checkpoint["weight"]["emb_g.weight"].shape[0]
net_g = SynthesizerTrnMsNSFsidM(*checkpoint["config"], is_half=False, version=checkpoint.get("version", "v1"))
net_g.load_state_dict(checkpoint["weight"], strict=False)
net_g = net_g.to(device)
dynamic_axes = {"phone": [1], "pitch": [1], "pitchf": [1], "rnd": [2]}
torch.onnx.export(
net_g,
tuple(value.to(device) for value in test_inputs.values()),
exported_path,
dynamic_axes=dynamic_axes,
do_constant_folding=False,
opset_version=13,
verbose=False,
input_names=list(test_inputs.keys()),
output_names=["audio"],
)
return "Finished"
#region Mangio-RVC-Fork CLI App
import scipy.io.wavfile as wavfile
cli_current_page = "HOME"
def cli_split_command(com):
exp = r'(?:(?<=\s)|^)"(.*?)"(?=\s|$)|(\S+)'
split_array = re.findall(exp, com)
split_array = [group[0] if group[0] else group[1] for group in split_array]
return split_array
execute_generator_function = lambda genObject: all(x is not None for x in genObject)
def cli_infer(com):
model_name, source_audio_path, output_file_name, feature_index_path, speaker_id, transposition, f0_method, crepe_hop_length, harvest_median_filter, resample, mix, feature_ratio, protection_amnt, _, f0_min, f0_max, do_formant = cli_split_command(com)[:17]
speaker_id, crepe_hop_length, harvest_median_filter, resample = map(int, [speaker_id, crepe_hop_length, harvest_median_filter, resample])
transposition, mix, feature_ratio, protection_amnt = map(float, [transposition, mix, feature_ratio, protection_amnt])
if do_formant.lower() == 'false':
Quefrency = 1.0
Timbre = 1.0
else:
Quefrency, Timbre = map(float, cli_split_command(com)[17:19])
rvc_globals.DoFormant = do_formant.lower() == 'true'
rvc_globals.Quefrency = Quefrency
rvc_globals.Timbre = Timbre
output_message = 'Mangio-RVC-Fork Infer-CLI:'
output_path = f'audio-outputs/{output_file_name}'
print(f"{output_message} Starting the inference...")
vc_data = get_vc(model_name, protection_amnt, protection_amnt)
print(vc_data)
print(f"{output_message} Performing inference...")
conversion_data = vc_single(
speaker_id,
source_audio_path,
source_audio_path,
transposition,
None, # f0 file support not implemented
f0_method,
feature_index_path,
feature_index_path,
feature_ratio,
harvest_median_filter,
resample,
mix,
protection_amnt,
crepe_hop_length,
f0_min=f0_min,
note_min=None,
f0_max=f0_max,
note_max=None
)
if "Success." in conversion_data[0]:
print(f"{output_message} Inference succeeded. Writing to {output_path}...")
wavfile.write(output_path, conversion_data[1][0], conversion_data[1][1])
print(f"{output_message} Finished! Saved output to {output_path}")
else:
print(f"{output_message} Inference failed. Here's the traceback: {conversion_data[0]}")
def cli_pre_process(com):
print("Mangio-RVC-Fork Pre-process: Starting...")
execute_generator_function(
preprocess_dataset(
*cli_split_command(com)[:3],
int(cli_split_command(com)[3])
)
)
print("Mangio-RVC-Fork Pre-process: Finished")
def cli_extract_feature(com):
model_name, gpus, num_processes, has_pitch_guidance, f0_method, crepe_hop_length, version = cli_split_command(com)
num_processes = int(num_processes)
has_pitch_guidance = bool(int(has_pitch_guidance))
crepe_hop_length = int(crepe_hop_length)
print(
f"Mangio-RVC-CLI: Extract Feature Has Pitch: {has_pitch_guidance}"
f"Mangio-RVC-CLI: Extract Feature Version: {version}"
"Mangio-RVC-Fork Feature Extraction: Starting..."
)
generator = extract_f0_feature(
gpus,
num_processes,
f0_method,
has_pitch_guidance,
model_name,