import os import shutil import sys import tempfile os.environ["CUDA_VISIBLE_DEVICES"] = "" now_dir = os.getcwd() sys.path.append(now_dir) import traceback, pdb import warnings import numpy as np import torch os.environ['OPENBLAS_NUM_THREADS'] = '1' os.environ["no_proxy"] = "localhost, 127.0.0.1, ::1" import logging import threading from random import shuffle from subprocess import Popen from time import sleep import faiss import ffmpeg import gradio as gr import soundfile as sf from config import Config from fairseq import checkpoint_utils from i18n import I18nAuto from infer_pack.models import ( SynthesizerTrnMs256NSFsid, SynthesizerTrnMs256NSFsid_nono, SynthesizerTrnMs768NSFsid, SynthesizerTrnMs768NSFsid_nono, ) from 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 logging.getLogger("numba").setLevel(logging.WARNING) tmp = os.path.join(now_dir, "TEMP") shutil.rmtree(tmp, ignore_errors=True) shutil.rmtree("%s/runtime/Lib/site-packages/infer_pack" % (now_dir), ignore_errors=True) shutil.rmtree("%s/runtime/Lib/site-packages/uvr5_pack" % (now_dir), ignore_errors=True) os.makedirs(tmp, exist_ok=True) os.makedirs(os.path.join(now_dir, "logs"), exist_ok=True) os.makedirs(os.path.join(now_dir, "weights"), exist_ok=True) os.environ["TEMP"] = tmp warnings.filterwarnings("ignore") torch.manual_seed(114514) from scipy.io import wavfile 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 if torch.cuda.is_available() or ngpu != 0: for i in range(ngpu): gpu_name = torch.cuda.get_device_name(i) if any( value in gpu_name.upper() for value in [ "10", "16", "20", "30", "40", "A2", "A3", "A4", "P4", "A50", "500", "A60", "70", "80", "90", "M4", "T4", "TITAN", ] ): # A10#A100#V100#A40#P40#M40#K80#A4500 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 / 1024 / 1024 / 1024 + 0.4 ) ) if if_gpu_ok and len(gpu_infos) > 0: gpu_info = "\n".join(gpu_infos) default_batch_size = 1 else: gpu_info = i18n("很遗憾您这没有能用的显卡来支持您训练") default_batch_size = 1 gpus = "-".join([i[0] for i in gpu_infos]) class ToolButton(gr.Button, gr.components.FormComponent): """Small button with single emoji as text, fits inside gradio forms""" def __init__(self, **kwargs): super().__init__(variant="tool", **kwargs) def get_block_name(self): return "button" 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] hubert_model = hubert_model.to(config.device) if config.is_half: hubert_model = hubert_model.half() else: hubert_model = hubert_model.float() hubert_model.eval() weight_root = "weights" weight_uvr5_root = "uvr5_weights" index_root = "logs" names = [] for name in os.listdir(weight_root): if name.endswith(".pth"): names.append(name) index_paths = [] for root, dirs, files in os.walk(index_root, topdown=False): for name in files: if name.endswith(".index") and "trained" not in name: index_paths.append("%s/%s" % (root, name)) uvr5_names = [] for name in os.listdir(weight_uvr5_root): if name.endswith(".pth") or "onnx" in name: uvr5_names.append(name.replace(".pth", "")) def vc_single( sid, input_audio_path, f0_up_key, f0_file, f0_method, file_index, file_index2, # file_big_npy, index_rate, filter_radius, resample_sr, rms_mix_rate, protect, ): # spk_item, input_audio0, vc_transform0,f0_file,f0method0 global tgt_sr, net_g, vc, hubert_model, version print(f0_up_key) if input_audio_path is None: return "You need to upload an audio", None print("input_audio_path: ", input_audio_path) print("f0_up_key: ", f0_up_key) f0_up_key = int(f0_up_key) try: audio = load_audio(input_audio_path, 16000) audio_max = np.abs(audio).max() / 0.95 if audio_max > 1: audio /= audio_max times = [0, 0, 0] if not hubert_model: load_hubert() if_f0 = cpt.get("f0", 1) file_index = ( ( file_index.strip(" ") .strip('"') .strip("\n") .strip('"') .strip(" ") .replace("trained", "added") ) if file_index != "" else file_index2 ) # 防止小白写错，自动帮他替换掉 # file_big_npy = ( # file_big_npy.strip(" ").strip('"').strip("\n").strip('"').strip(" ") # ) audio_opt = vc.pipeline( hubert_model, net_g, sid, audio, input_audio_path, times, f0_up_key, f0_method, file_index, # file_big_npy, index_rate, if_f0, filter_radius, tgt_sr, resample_sr, rms_mix_rate, version, protect, f0_file=f0_file, ) print(f0_up_key) 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." ) return "Success.\n %s\nTime:\n npy:%ss, f0:%ss, infer:%ss" % ( index_info, times[0], times[1], times[2], ), (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, ): try: print(f0_up_key) dir_path = ( dir_path.strip(" ").strip('"').strip("\n").strip('"').strip(" ") ) # 防止小白拷路径头尾带了空格和"和回车 opt_root = opt_root.strip(" ").strip('"').strip("\n").strip('"').strip(" ") os.makedirs(opt_root, exist_ok=True) try: if dir_path != "": paths = [os.path.join(dir_path, name) for name in os.listdir(dir_path)] else: paths = [path.name for path in paths] except: traceback.print_exc() paths = [path.name for path in paths] infos = [] for path in paths: info, opt = vc_single( sid, path, f0_up_key, None, f0_method, file_index, file_index2, # file_big_npy, index_rate, filter_radius, resample_sr, rms_mix_rate, protect, ) if "Success" in info: try: tgt_sr, audio_opt = opt if format1 in ["wav", "flac"]: sf.write( "%s/%s.%s" % (opt_root, os.path.basename(path), format1), audio_opt, tgt_sr, ) else: path = "%s/%s.wav" % (opt_root, os.path.basename(path)) sf.write( path, audio_opt, tgt_sr, ) if os.path.exists(path): os.system( "ffmpeg -i %s -vn %s -q:a 2 -y" % (path, path[:-4] + ".%s" % format1) ) except: info += traceback.format_exc() infos.append("%s->%s" % (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 = inp_root.strip(" ").strip('"').strip("\n").strip('"').strip(" ") save_root_vocal = ( save_root_vocal.strip(" ").strip('"').strip("\n").strip('"').strip(" ") ) save_root_ins = ( save_root_ins.strip(" ").strip('"').strip("\n").strip('"').strip(" ") ) if model_name == "onnx_dereverb_By_FoxJoy": pre_fun = MDXNetDereverb(15) else: func = _audio_pre_ if "DeEcho" not in model_name else _audio_pre_new pre_fun = func( agg=int(agg), model_path=os.path.join(weight_uvr5_root, model_name + ".pth"), device=config.device, is_half=config.is_half, ) if inp_root != "": paths = [os.path.join(inp_root, name) for name in os.listdir(inp_root)] else: paths = [path.name for path in paths] for path in paths: inp_path = os.path.join(inp_root, path) need_reformat = 1 done = 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: need_reformat = 1 traceback.print_exc() if need_reformat == 1: tmp_path = "%s/%s.reformatted.wav" % (tmp, os.path.basename(inp_path)) os.system( "ffmpeg -i %s -vn -acodec pcm_s16le -ac 2 -ar 44100 %s -y" % (inp_path, tmp_path) ) inp_path = tmp_path try: if done == 0: pre_fun._path_audio_( inp_path, save_root_ins, save_root_vocal, format0 ) infos.append("%s->Success" % (os.path.basename(inp_path))) yield "\n".join(infos) except: infos.append( "%s->%s" % (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) # 一个选项卡全局只能有一个音色 def get_vc(sid, to_return_protect0, to_return_protect1): global n_spk, tgt_sr, net_g, vc, cpt, version if sid == "" or sid == []: global hubert_model if hubert_model is not None: # 考虑到轮询, 需要加个判断看是否 sid 是由有模型切换到无模型的 print("clean_empty_cache") del net_g, n_spk, vc, hubert_model, tgt_sr # ,cpt hubert_model = net_g = n_spk = vc = hubert_model = tgt_sr = None if torch.cuda.is_available(): torch.cuda.empty_cache() ###楼下不这么折腾清理不干净 if_f0 = cpt.get("f0", 1) version = cpt.get("version", "v1") if version == "v1": if if_f0 == 1: net_g = SynthesizerTrnMs256NSFsid( *cpt["config"], is_half=config.is_half ) else: net_g = SynthesizerTrnMs256NSFsid_nono(*cpt["config"]) elif version == "v2": if if_f0 == 1: net_g = SynthesizerTrnMs768NSFsid( *cpt["config"], is_half=config.is_half ) else: net_g = SynthesizerTrnMs768NSFsid_nono(*cpt["config"]) del net_g, cpt if torch.cuda.is_available(): torch.cuda.empty_cache() cpt = None return {"visible": False, "__type__": "update"} person = "%s/%s" % (weight_root, sid) print("loading %s" % person) cpt = torch.load(person, map_location="cpu") tgt_sr = cpt["config"][-1] cpt["config"][-3] = cpt["weight"]["emb_g.weight"].shape[0] # n_spk if_f0 = cpt.get("f0", 1) if if_f0 == 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") if version == "v1": if if_f0 == 1: net_g = SynthesizerTrnMs256NSFsid(*cpt["config"], is_half=config.is_half) else: net_g = SynthesizerTrnMs256NSFsid_nono(*cpt["config"]) elif version == "v2": if if_f0 == 1: net_g = SynthesizerTrnMs768NSFsid(*cpt["config"], is_half=config.is_half) else: net_g = 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) if config.is_half: net_g = net_g.half() else: net_g = net_g.float() vc = VC(tgt_sr, config) n_spk = cpt["config"][-3] return ( {"visible": True, "maximum": n_spk, "__type__": "update"}, to_return_protect0, to_return_protect1, ) def change_choices(): names = [] for name in os.listdir(weight_root): if name.endswith(".pth"): names.append(name) index_paths = [] for root, dirs, files in os.walk(index_root, topdown=False): for name in files: if name.endswith(".index") and "trained" not in name: index_paths.append("%s/%s" % (root, name)) return {"choices": sorted(names), "__type__": "update"}, { "choices": sorted(index_paths), "__type__": "update", } def clean(): return {"value": "", "__type__": "update"} sr_dict = { "32k": 32000, "40k": 40000, "48k": 48000, } def if_done(done, p): while 1: if p.poll() is None: sleep(0.5) else: break done[0] = True def if_done_multi(done, ps): while 1: # poll==None代表进程未结束 # 只要有一个进程未结束都不停 flag = 1 for p in ps: if p.poll() is None: flag = 0 sleep(0.5) break if flag == 1: break done[0] = True def preprocess_dataset(trainset_dir, exp_dir, sr, n_p): sr = sr_dict[sr] os.makedirs("%s/logs/%s" % (now_dir, exp_dir), exist_ok=True) f = open("%s/logs/%s/preprocess.log" % (now_dir, exp_dir), "w") f.close() cmd = ( config.python_cmd + " trainset_preprocess_pipeline_print.py %s %s %s %s/logs/%s " % (trainset_dir, sr, n_p, now_dir, exp_dir) + str(config.noparallel) ) print(cmd) p = Popen(cmd, shell=True) # , stdin=PIPE, stdout=PIPE,stderr=PIPE,cwd=now_dir ###煞笔gr, popen read都非得全跑完了再一次性读取, 不用gr就正常读一句输出一句;只能额外弄出一个文本流定时读 done = [False] threading.Thread( target=if_done, args=( done, p, ), ).start() while 1: with open("%s/logs/%s/preprocess.log" % (now_dir, exp_dir), "r") as f: yield (f.read()) sleep(1) if done[0]: break with open("%s/logs/%s/preprocess.log" % (now_dir, exp_dir), "r") as f: log = f.read() print(log) yield log # but2.click(extract_f0,[gpus6,np7,f0method8,if_f0_3,trainset_dir4],[info2]) def extract_f0_feature(gpus, n_p, f0method, if_f0, exp_dir, version19): gpus = gpus.split("-") os.makedirs("%s/logs/%s" % (now_dir, exp_dir), exist_ok=True) f = open("%s/logs/%s/extract_f0_feature.log" % (now_dir, exp_dir), "w") f.close() if if_f0: cmd = config.python_cmd + " extract_f0_print.py %s/logs/%s %s %s" % ( now_dir, exp_dir, n_p, f0method, ) print(cmd) p = Popen(cmd, shell=True, cwd=now_dir) # , stdin=PIPE, stdout=PIPE,stderr=PIPE ###煞笔gr, popen read都非得全跑完了再一次性读取, 不用gr就正常读一句输出一句;只能额外弄出一个文本流定时读 done = [False] threading.Thread( target=if_done, args=( done, p, ), ).start() while 1: with open( "%s/logs/%s/extract_f0_feature.log" % (now_dir, exp_dir), "r" ) as f: yield (f.read()) sleep(1) if done[0]: break with open("%s/logs/%s/extract_f0_feature.log" % (now_dir, exp_dir), "r") as f: log = f.read() print(log) yield log ####对不同part分别开多进程 """ n_part=int(sys.argv[1]) i_part=int(sys.argv[2]) i_gpu=sys.argv[3] exp_dir=sys.argv[4] os.environ["CUDA_VISIBLE_DEVICES"]=str(i_gpu) """ leng = len(gpus) ps = [] for idx, n_g in enumerate(gpus): cmd = ( config.python_cmd + " extract_feature_print.py %s %s %s %s %s/logs/%s %s" % ( config.device, leng, idx, n_g, now_dir, exp_dir, version19, ) ) print(cmd) p = Popen( cmd, shell=True, cwd=now_dir ) # , shell=True, stdin=PIPE, stdout=PIPE, stderr=PIPE, cwd=now_dir ps.append(p) ###煞笔gr, popen read都非得全跑完了再一次性读取, 不用gr就正常读一句输出一句;只能额外弄出一个文本流定时读 done = [False] threading.Thread( target=if_done_multi, args=( done, ps, ), ).start() while 1: with open("%s/logs/%s/extract_f0_feature.log" % (now_dir, exp_dir), "r") as f: yield (f.read()) sleep(1) if done[0]: break with open("%s/logs/%s/extract_f0_feature.log" % (now_dir, exp_dir), "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 "" if_pretrained_generator_exist = os.access( "pretrained%s/%sG%s.pth" % (path_str, f0_str, sr2), os.F_OK ) if_pretrained_discriminator_exist = os.access( "pretrained%s/%sD%s.pth" % (path_str, f0_str, sr2), os.F_OK ) if not if_pretrained_generator_exist: print( "pretrained%s/%sG%s.pth" % (path_str, f0_str, sr2), "not exist, will not use pretrained model", ) if not if_pretrained_discriminator_exist: print( "pretrained%s/%sD%s.pth" % (path_str, f0_str, sr2), "not exist, will not use pretrained model", ) return ( "pretrained%s/%sG%s.pth" % (path_str, f0_str, sr2) if if_pretrained_generator_exist else "", "pretrained%s/%sD%s.pth" % (path_str, f0_str, sr2) if if_pretrained_discriminator_exist else "", ) def change_version19(sr2, if_f0_3, version19): path_str = "" if version19 == "v1" else "_v2" if sr2 == "32k" and version19 == "v1": sr2 = "40k" to_return_sr2 = ( {"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 "" if_pretrained_generator_exist = os.access( "pretrained%s/%sG%s.pth" % (path_str, f0_str, sr2), os.F_OK ) if_pretrained_discriminator_exist = os.access( "pretrained%s/%sD%s.pth" % (path_str, f0_str, sr2), os.F_OK ) if not if_pretrained_generator_exist: print( "pretrained%s/%sG%s.pth" % (path_str, f0_str, sr2), "not exist, will not use pretrained model", ) if not if_pretrained_discriminator_exist: print( "pretrained%s/%sD%s.pth" % (path_str, f0_str, sr2), "not exist, will not use pretrained model", ) return ( "pretrained%s/%sG%s.pth" % (path_str, f0_str, sr2) if if_pretrained_generator_exist else "", "pretrained%s/%sD%s.pth" % (path_str, f0_str, sr2) if if_pretrained_discriminator_exist else "", to_return_sr2, ) def change_f0(if_f0_3, sr2, version19): # f0method8,pretrained_G14,pretrained_D15 path_str = "" if version19 == "v1" else "_v2" if_pretrained_generator_exist = os.access( "pretrained%s/f0G%s.pth" % (path_str, sr2), os.F_OK ) if_pretrained_discriminator_exist = os.access( "pretrained%s/f0D%s.pth" % (path_str, sr2), os.F_OK ) if not if_pretrained_generator_exist: print( "pretrained%s/f0G%s.pth" % (path_str, sr2), "not exist, will not use pretrained model", ) if not if_pretrained_discriminator_exist: print( "pretrained%s/f0D%s.pth" % (path_str, sr2), "not exist, will not use pretrained model", ) if if_f0_3: return ( {"visible": True, "__type__": "update"}, "pretrained%s/f0G%s.pth" % (path_str, sr2) if if_pretrained_generator_exist else "", "pretrained%s/f0D%s.pth" % (path_str, sr2) if if_pretrained_discriminator_exist else "", ) return ( {"visible": False, "__type__": "update"}, ("pretrained%s/G%s.pth" % (path_str, sr2)) if if_pretrained_generator_exist else "", ("pretrained%s/D%s.pth" % (path_str, sr2)) if if_pretrained_discriminator_exist else "", ) # but3.click(click_train,[exp_dir1,sr2,if_f0_3,save_epoch10,total_epoch11,batch_size12,if_save_latest13,pretrained_G14,pretrained_D15,gpus16]) 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, ): # 生成filelist exp_dir = "%s/logs/%s" % (now_dir, exp_dir1) os.makedirs(exp_dir, exist_ok=True) gt_wavs_dir = "%s/0_gt_wavs" % (exp_dir) feature_dir = ( "%s/3_feature256" % (exp_dir) if version19 == "v1" else "%s/3_feature768" % (exp_dir) ) if if_f0_3: f0_dir = "%s/2a_f0" % (exp_dir) f0nsf_dir = "%s/2b-f0nsf" % (exp_dir) names = ( set([name.split(".")[0] for name in os.listdir(gt_wavs_dir)]) & set([name.split(".")[0] for name in os.listdir(feature_dir)]) & set([name.split(".")[0] for name in os.listdir(f0_dir)]) & set([name.split(".")[0] for name in os.listdir(f0nsf_dir)]) ) else: names = set([name.split(".")[0] for name in os.listdir(gt_wavs_dir)]) & set( [name.split(".")[0] for name in os.listdir(feature_dir)] ) opt = [] for name in names: if if_f0_3: opt.append( "%s/%s.wav|%s/%s.npy|%s/%s.wav.npy|%s/%s.wav.npy|%s" % ( gt_wavs_dir.replace("\\", "\\\\"), name, feature_dir.replace("\\", "\\\\"), name, f0_dir.replace("\\", "\\\\"), name, f0nsf_dir.replace("\\", "\\\\"), name, spk_id5, ) ) else: opt.append( "%s/%s.wav|%s/%s.npy|%s" % ( gt_wavs_dir.replace("\\", "\\\\"), name, feature_dir.replace("\\", "\\\\"), name, spk_id5, ) ) fea_dim = 256 if version19 == "v1" else 768 if if_f0_3: for _ in range(2): opt.append( "%s/logs/mute/0_gt_wavs/mute%s.wav|%s/logs/mute/3_feature%s/mute.npy|%s/logs/mute/2a_f0/mute.wav.npy|%s/logs/mute/2b-f0nsf/mute.wav.npy|%s" % (now_dir, sr2, now_dir, fea_dim, now_dir, now_dir, spk_id5) ) else: for _ in range(2): opt.append( "%s/logs/mute/0_gt_wavs/mute%s.wav|%s/logs/mute/3_feature%s/mute.npy|%s" % (now_dir, sr2, now_dir, fea_dim, spk_id5) ) shuffle(opt) with open("%s/filelist.txt" % exp_dir, "w") as f: f.write("\n".join(opt)) print("write filelist done") # 生成config#无需生成config # cmd = python_cmd + " train_nsf_sim_cache_sid_load_pretrain.py -e mi-test -sr 40k -f0 1 -bs 4 -g 0 -te 10 -se 5 -pg pretrained/f0G40k.pth -pd pretrained/f0D40k.pth -l 1 -c 0" print("use gpus:", gpus16) if pretrained_G14 == "": print("no pretrained Generator") if pretrained_D15 == "": print("no pretrained Discriminator") if gpus16: cmd = ( config.python_cmd + " train_nsf_sim_cache_sid_load_pretrain.py -e %s -sr %s -f0 %s -bs %s -g %s -te %s -se %s %s %s -l %s -c %s -sw %s -v %s" % ( exp_dir1, sr2, 1 if if_f0_3 else 0, batch_size12, gpus16, total_epoch11, save_epoch10, "-pg %s" % pretrained_G14 if pretrained_G14 != "" else "", "-pd %s" % pretrained_D15 if pretrained_D15 != "" else "", 1 if if_save_latest13 == i18n("是") else 0, 1 if if_cache_gpu17 == i18n("是") else 0, 1 if if_save_every_weights18 == i18n("是") else 0, version19, ) ) else: cmd = ( config.python_cmd + " train_nsf_sim_cache_sid_load_pretrain.py -e %s -sr %s -f0 %s -bs %s -te %s -se %s %s %s -l %s -c %s -sw %s -v %s" % ( exp_dir1, sr2, 1 if if_f0_3 else 0, batch_size12, total_epoch11, save_epoch10, "-pg %s" % pretrained_G14 if pretrained_G14 != "" else "\b", "-pd %s" % pretrained_D15 if pretrained_D15 != "" else "\b", 1 if if_save_latest13 == i18n("是") else 0, 1 if if_cache_gpu17 == i18n("是") else 0, 1 if if_save_every_weights18 == i18n("是") else 0, version19, ) ) print(cmd) p = Popen(cmd, shell=True, cwd=now_dir) p.wait() return "训练结束, 您可查看控制台训练日志或实验文件夹下的train.log" # but4.click(train_index, [exp_dir1], info3) def train_index(exp_dir1, version19): exp_dir = "%s/logs/%s" % (now_dir, exp_dir1) os.makedirs(exp_dir, exist_ok=True) feature_dir = ( "%s/3_feature256" % (exp_dir) if version19 == "v1" else "%s/3_feature768" % (exp_dir) ) if not os.path.exists(feature_dir): return "请先进行特征提取!" listdir_res = list(os.listdir(feature_dir)) if len(listdir_res) == 0: return "请先进行特征提取！" infos = [] npys = [] for name in sorted(listdir_res): phone = np.load("%s/%s" % (feature_dir, name)) npys.append(phone) big_npy = np.concatenate(npys, 0) big_npy_idx = np.arange(big_npy.shape[0]) np.random.shuffle(big_npy_idx) big_npy = big_npy[big_npy_idx] if big_npy.shape[0] > 2e5: # if(1): 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: info = traceback.format_exc() print(info) infos.append(info) yield "\n".join(infos) np.save("%s/total_fea.npy" % exp_dir, 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(256 if version19 == "v1" else 768, "IVF%s,Flat" % n_ivf) # index = faiss.index_factory(256if version19=="v1"else 768, "IVF%s,PQ128x4fs,RFlat"%n_ivf) infos.append("training") yield "\n".join(infos) index_ivf = faiss.extract_index_ivf(index) # index_ivf.nprobe = 1 index.train(big_npy) faiss.write_index( index, "%s/trained_IVF%s_Flat_nprobe_%s_%s_%s.index" % (exp_dir, n_ivf, index_ivf.nprobe, exp_dir1, version19), ) # faiss.write_index(index, '%s/trained_IVF%s_Flat_FastScan_%s.index'%(exp_dir,n_ivf,version19)) 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]) faiss.write_index( index, "%s/added_IVF%s_Flat_nprobe_%s_%s_%s.index" % (exp_dir, n_ivf, index_ivf.nprobe, exp_dir1, version19), ) infos.append( "成功构建索引，added_IVF%s_Flat_nprobe_%s_%s_%s.index" % (n_ivf, index_ivf.nprobe, exp_dir1, version19) ) # faiss.write_index(index, '%s/added_IVF%s_Flat_FastScan_%s.index'%(exp_dir,n_ivf,version19)) # infos.append("成功构建索引，added_IVF%s_Flat_FastScan_%s.index"%(n_ivf,version19)) yield "\n".join(infos) # but5.click(train1key, [exp_dir1, sr2, if_f0_3, trainset_dir4, spk_id5, gpus6, np7, f0method8, save_epoch10, total_epoch11, batch_size12, if_save_latest13, pretrained_G14, pretrained_D15, gpus16, if_cache_gpu17], info3) def train1key( exp_dir1, sr2, if_f0_3, trainset_dir4, spk_id5, np7, f0method8, save_epoch10, total_epoch11, batch_size12, if_save_latest13, pretrained_G14, pretrained_D15, gpus16, if_cache_gpu17, if_save_every_weights18, version19, ): infos = [] def get_info_str(strr): infos.append(strr) return "\n".join(infos) model_log_dir = "%s/logs/%s" % (now_dir, exp_dir1) preprocess_log_path = "%s/preprocess.log" % model_log_dir extract_f0_feature_log_path = "%s/extract_f0_feature.log" % model_log_dir gt_wavs_dir = "%s/0_gt_wavs" % model_log_dir feature_dir = ( "%s/3_feature256" % model_log_dir if version19 == "v1" else "%s/3_feature768" % model_log_dir ) os.makedirs(model_log_dir, exist_ok=True) #########step1:处理数据 open(preprocess_log_path, "w").close() cmd = ( config.python_cmd + " trainset_preprocess_pipeline_print.py %s %s %s %s " % (trainset_dir4, sr_dict[sr2], np7, model_log_dir) + str(config.noparallel) ) yield get_info_str(i18n("step1:正在处理数据")) yield get_info_str(cmd) p = Popen(cmd, shell=True) p.wait() with open(preprocess_log_path, "r") as f: print(f.read()) #########step2a:提取音高 open(extract_f0_feature_log_path, "w") if if_f0_3: yield get_info_str("step2a:正在提取音高") cmd = config.python_cmd + " extract_f0_print.py %s %s %s" % ( model_log_dir, np7, f0method8, ) yield get_info_str(cmd) p = Popen(cmd, shell=True, cwd=now_dir) p.wait() with open(extract_f0_feature_log_path, "r") as f: print(f.read()) else: yield get_info_str(i18n("step2a:无需提取音高")) #######step2b:提取特征 yield get_info_str(i18n("step2b:正在提取特征")) gpus = gpus16.split("-") leng = len(gpus) ps = [] for idx, n_g in enumerate(gpus): cmd = config.python_cmd + " extract_feature_print.py %s %s %s %s %s %s" % ( config.device, leng, idx, n_g, model_log_dir, version19, ) yield get_info_str(cmd) p = Popen( cmd, shell=True, cwd=now_dir ) # , shell=True, stdin=PIPE, stdout=PIPE, stderr=PIPE, cwd=now_dir ps.append(p) for p in ps: p.wait() with open(extract_f0_feature_log_path, "r") as f: print(f.read()) #######step3a:训练模型 yield get_info_str(i18n("step3a:正在训练模型")) # 生成filelist if if_f0_3: f0_dir = "%s/2a_f0" % model_log_dir f0nsf_dir = "%s/2b-f0nsf" % model_log_dir names = ( set([name.split(".")[0] for name in os.listdir(gt_wavs_dir)]) & set([name.split(".")[0] for name in os.listdir(feature_dir)]) & set([name.split(".")[0] for name in os.listdir(f0_dir)]) & set([name.split(".")[0] for name in os.listdir(f0nsf_dir)]) ) else: names = set([name.split(".")[0] for name in os.listdir(gt_wavs_dir)]) & set( [name.split(".")[0] for name in os.listdir(feature_dir)] ) opt = [] for name in names: if if_f0_3: opt.append( "%s/%s.wav|%s/%s.npy|%s/%s.wav.npy|%s/%s.wav.npy|%s" % ( gt_wavs_dir.replace("\\", "\\\\"), name, feature_dir.replace("\\", "\\\\"), name, f0_dir.replace("\\", "\\\\"), name, f0nsf_dir.replace("\\", "\\\\"), name, spk_id5, ) ) else: opt.append( "%s/%s.wav|%s/%s.npy|%s" % ( gt_wavs_dir.replace("\\", "\\\\"), name, feature_dir.replace("\\", "\\\\"), name, spk_id5, ) ) fea_dim = 256 if version19 == "v1" else 768 if if_f0_3: for _ in range(2): opt.append( "%s/logs/mute/0_gt_wavs/mute%s.wav|%s/logs/mute/3_feature%s/mute.npy|%s/logs/mute/2a_f0/mute.wav.npy|%s/logs/mute/2b-f0nsf/mute.wav.npy|%s" % (now_dir, sr2, now_dir, fea_dim, now_dir, now_dir, spk_id5) ) else: for _ in range(2): opt.append( "%s/logs/mute/0_gt_wavs/mute%s.wav|%s/logs/mute/3_feature%s/mute.npy|%s" % (now_dir, sr2, now_dir, fea_dim, spk_id5) ) shuffle(opt) with open("%s/filelist.txt" % model_log_dir, "w") as f: f.write("\n".join(opt)) yield get_info_str("write filelist done") if gpus16: cmd = ( config.python_cmd + " train_nsf_sim_cache_sid_load_pretrain.py -e %s -sr %s -f0 %s -bs %s -g %s -te %s -se %s %s %s -l %s -c %s -sw %s -v %s" % ( exp_dir1, sr2, 1 if if_f0_3 else 0, batch_size12, gpus16, total_epoch11, save_epoch10, "-pg %s" % pretrained_G14 if pretrained_G14 != "" else "", "-pd %s" % pretrained_D15 if pretrained_D15 != "" else "", 1 if if_save_latest13 == i18n("是") else 0, 1 if if_cache_gpu17 == i18n("是") else 0, 1 if if_save_every_weights18 == i18n("是") else 0, version19, ) ) else: cmd = ( config.python_cmd + " train_nsf_sim_cache_sid_load_pretrain.py -e %s -sr %s -f0 %s -bs %s -te %s -se %s %s %s -l %s -c %s -sw %s -v %s" % ( exp_dir1, sr2, 1 if if_f0_3 else 0, batch_size12, total_epoch11, save_epoch10, "-pg %s" % pretrained_G14 if pretrained_G14 != "" else "", "-pd %s" % pretrained_D15 if pretrained_D15 != "" else "", 1 if if_save_latest13 == i18n("是") else 0, 1 if if_cache_gpu17 == i18n("是") else 0, 1 if if_save_every_weights18 == i18n("是") else 0, version19, ) ) yield get_info_str(cmd) p = Popen(cmd, shell=True, cwd=now_dir) p.wait() yield get_info_str(i18n("训练结束, 您可查看控制台训练日志或实验文件夹下的train.log")) #######step3b:训练索引 npys = [] listdir_res = list(os.listdir(feature_dir)) for name in sorted(listdir_res): phone = np.load("%s/%s" % (feature_dir, name)) npys.append(phone) big_npy = np.concatenate(npys, 0) big_npy_idx = np.arange(big_npy.shape[0]) np.random.shuffle(big_npy_idx) big_npy = big_npy[big_npy_idx] if big_npy.shape[0] > 2e5: # if(1): info = "Trying doing kmeans %s shape to 10k centers." % big_npy.shape[0] print(info) yield get_info_str(info) 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: info = traceback.format_exc() print(info) yield get_info_str(info) np.save("%s/total_fea.npy" % model_log_dir, big_npy) # n_ivf = big_npy.shape[0] // 39 n_ivf = min(int(16 * np.sqrt(big_npy.shape[0])), big_npy.shape[0] // 39) yield get_info_str("%s,%s" % (big_npy.shape, n_ivf)) index = faiss.index_factory(256 if version19 == "v1" else 768, "IVF%s,Flat" % n_ivf) yield get_info_str("training index") index_ivf = faiss.extract_index_ivf(index) # index_ivf.nprobe = 1 index.train(big_npy) faiss.write_index( index, "%s/trained_IVF%s_Flat_nprobe_%s_%s_%s.index" % (model_log_dir, n_ivf, index_ivf.nprobe, exp_dir1, version19), ) yield get_info_str("adding index") 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]) faiss.write_index( index, "%s/added_IVF%s_Flat_nprobe_%s_%s_%s.index" % (model_log_dir, n_ivf, index_ivf.nprobe, exp_dir1, version19), ) yield get_info_str( "成功构建索引, added_IVF%s_Flat_nprobe_%s_%s_%s.index" % (n_ivf, index_ivf.nprobe, exp_dir1, version19) ) yield get_info_str(i18n("全流程结束！")) # ckpt_path2.change(change_info_,[ckpt_path2],[sr__,if_f0__]) def change_info_(ckpt_path): if not os.path.exists(ckpt_path.replace(os.path.basename(ckpt_path), "train.log")): return {"__type__": "update"}, {"__type__": "update"}, {"__type__": "update"} try: with open( ckpt_path.replace(os.path.basename(ckpt_path), "train.log"), "r" ) as f: info = eval(f.read().strip("\n").split("\n")[0].split("\t")[-1]) sr, f0 = info["sample_rate"], info["if_f0"] version = "v2" if ("version" in info and info["version"] == "v2") else "v1" return sr, str(f0), version except: traceback.print_exc() return {"__type__": "update"}, {"__type__": "update"}, {"__type__": "update"} def export_onnx(ModelPath, ExportedPath): cpt = torch.load(ModelPath, map_location="cpu") cpt["config"][-3] = cpt["weight"]["emb_g.weight"].shape[0] vec_channels = 256 if cpt.get("version", "v1") == "v1" else 768 test_phone = torch.rand(1, 200, vec_channels) # hidden unit test_phone_lengths = torch.tensor([200]).long() # hidden unit 长度（貌似没啥用） test_pitch = torch.randint(size=(1, 200), low=5, high=255) # 基频（单位赫兹） test_pitchf = torch.rand(1, 200) # nsf基频 test_ds = torch.LongTensor([0]) # 说话人ID test_rnd = torch.rand(1, 192, 200) # 噪声（加入随机因子） device = "cpu" # 导出时设备（不影响使用模型） net_g = SynthesizerTrnMsNSFsidM( *cpt["config"], is_half=False, version=cpt.get("version", "v1") ) # fp32导出（C++要支持fp16必须手动将内存重新排列所以暂时不用fp16） net_g.load_state_dict(cpt["weight"], strict=False) input_names = ["phone", "phone_lengths", "pitch", "pitchf", "ds", "rnd"] output_names = [ "audio", ] # net_g.construct_spkmixmap(n_speaker) 多角色混合轨道导出 torch.onnx.export( net_g, ( test_phone.to(device), test_phone_lengths.to(device), test_pitch.to(device), test_pitchf.to(device), test_ds.to(device), test_rnd.to(device), ), ExportedPath, dynamic_axes={ "phone": [1], "pitch": [1], "pitchf": [1], "rnd": [2], }, do_constant_folding=False, opset_version=13, verbose=False, input_names=input_names, output_names=output_names, ) return "Finished" # sid0: Inferencing voice/ model name # f0_up_key: Transpose (integer, number of semitones, raise by an octave: 12, lower by an octave: -12): # opt_root: output folder path # f0method: pitch extraction algorithm ('pm': faster extraction but lower-quality speech; 'harvest': better bass but extremely slow; 'crepe': better quality but GPU intensive) # filter_radius: If >=3: apply median filtering to the harvested pitch results. The value represents the filter radius and can reduce breathiness. # file_index: Path to the feature index file.: # Auto-detect index path # index_rate: Search feature ratio (controls accent strength, too high has artifacting): minimum=0, maximum=1 # resample_sr: Resample the output audio in post-processing to the final sample rate. Set to 0 for no resampling (time consuming) # rms_mix_rate: Adjust the volume envelope scaling. Closer to 0, the more it mimicks the volume of the original vocals. Can help mask noise and make volume sound more natural when set relatively low. Closer to 1 will be more of a consistently loud volume: # protect: Protect voiceless consonants and breath sounds to prevent artifacts such as tearing in electronic music. Set to 0.5 to disable. Decrease the value to increase protection, but it may reduce indexing accuracy # dir_path: Enter the path of the audio folder to be processed (copy it from the address bar of the file manager) # format1: choices=["wav", "flac", "mp3", "m4a"] def run(sid0, paths, dir_path=None, f0_up_key=0, opt_root="opt", f0_method="pm", filter_radius=3, file_index="", file_index2=None, index_rate=1, resample_sr=0, rms_mix_rate=1, protect=0.33, format1="wav"): if (dir_path=='' or dir_path==None) and (paths == '' or paths==None): return "must provide either dir_input or file path" if paths != None or paths != '': tempfile = [file_to_tempfile(paths)] print(paths) print(protect) get_vc(sid0, protect, protect) vc_output3 = vc_multi( 0, # sid: 0 dir_path, # dir_path: opt_root, # opt_root: opt tempfile, # paths: [] f0_up_key, # f0_up_key: -12.0 f0_method, # f0_method: pm file_index, # file_index: file_index2, # file_index2: index_rate, # index_rate: 1 filter_radius, # filter_radius: 3 resample_sr, # resample_sr: 0 rms_mix_rate, # rms_mix_rate: 1 protect, # protect: 0.33 format1 # format1: wav ) out_path = paths wavfile.write(out_path, tgt_sr, vc_output3) return vc_output3 def get_models(): return names def file_to_tempfile(file_path): with open(file_path, 'rb') as file: temp_file = tempfile.TemporaryFile() temp_file.write(file.read()) temp_file.seek(0) return temp_file print(run('mymodelimran.pth', '/home/teewhy/Desktop/RVC/Retrieval-based-Voice-Conversion-WebUI/opt/abcxot47ylz.mp3.mp3')) # sid: 0 # dir_path: # opt_root: opt # paths: [] # f0_up_key: -12.0 # f0_method: pm # file_index: # file_index2: # index_rate: 1 # filter_radius: 3 # resample_sr: 0 # rms_mix_rate: 1 # protect: 0.33 # format1: wav # -12.0 # input_audio_path: /tmp/vocal_7x2a5ym2vczt3tom5.mp3_10.wav