from multiprocessing import cpu_count import threading, pdb, librosa from time import sleep from subprocess import Popen from time import sleep import torch, os, traceback, sys, warnings, shutil, numpy as np import faiss from random import shuffle now_dir = os.getcwd() sys.path.append(now_dir) tmp = os.path.join(now_dir, "TEMP") shutil.rmtree(tmp, 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 i18n import I18nAuto import ffmpeg i18n = I18nAuto() # 判断是否有能用来训练和加速推理的N卡 ncpu = cpu_count() ngpu = torch.cuda.device_count() gpu_infos = [] mem = [] if (not torch.cuda.is_available()) or ngpu == 0: if_gpu_ok = False else: if_gpu_ok = False for i in range(ngpu): gpu_name = torch.cuda.get_device_name(i) if ( "10" in gpu_name or "16" in gpu_name or "20" in gpu_name or "30" in gpu_name or "40" in gpu_name or "A2" in gpu_name.upper() or "A3" in gpu_name.upper() or "A4" in gpu_name.upper() or "P4" in gpu_name.upper() or "A50" in gpu_name.upper() or "70" in gpu_name or "80" in gpu_name or "90" in gpu_name or "M4" in gpu_name.upper() or "T4" in gpu_name.upper() or "TITAN" in gpu_name.upper() ): # 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 == True and len(gpu_infos) > 0: gpu_info = "\n".join(gpu_infos) default_batch_size = min(mem) // 2 else: gpu_info = "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]) from infer_pack.models import SynthesizerTrnMs256NSFsid, SynthesizerTrnMs256NSFsid_nono from scipy.io import wavfile from fairseq import checkpoint_utils import gradio as gr import logging from vc_infer_pipeline import VC from config import ( is_half, device, python_cmd, listen_port, iscolab, noparallel, noautoopen, ) from infer_uvr5 import _audio_pre_ from my_utils import load_audio from train.process_ckpt import show_info, change_info, merge, extract_small_model # from trainset_preprocess_pipeline import PreProcess logging.getLogger("numba").setLevel(logging.WARNING) 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(device) if is_half: hubert_model = hubert_model.half() else: hubert_model = hubert_model.float() hubert_model.eval() weight_root = "weights" weight_uvr5_root = "uvr5_weights" names = [] for name in os.listdir(weight_root): if name.endswith(".pth"): names.append(name) uvr5_names = [] for name in os.listdir(weight_uvr5_root): if name.endswith(".pth"): uvr5_names.append(name.replace(".pth", "")) def vc_single( sid, input_audio, f0_up_key, f0_file, f0_method, file_index, # file_big_npy, index_rate, ): # spk_item, input_audio0, vc_transform0,f0_file,f0method0 global tgt_sr, net_g, vc, hubert_model if input_audio is None: return "You need to upload an audio", None f0_up_key = int(f0_up_key) try: audio = load_audio(input_audio, 16000) times = [0, 0, 0] if hubert_model == None: load_hubert() if_f0 = cpt.get("f0", 1) file_index = ( file_index.strip(" ") .strip('"') .strip("\n") .strip('"') .strip(" ") .replace("trained", "added") ) # 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, times, f0_up_key, f0_method, file_index, # file_big_npy, index_rate, if_f0, f0_file=f0_file, ) print( "npy: ", times[0], "s, f0: ", times[1], "s, infer: ", times[2], "s", sep="" ) return "Success", (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_big_npy, index_rate, ): try: 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 = [] file_index = ( file_index.strip(" ") .strip('"') .strip("\n") .strip('"') .strip(" ") .replace("trained", "added") ) # 防止小白写错,自动帮他替换掉 for path in paths: info, opt = vc_single( sid, path, f0_up_key, None, f0_method, file_index, # file_big_npy, index_rate, ) if info == "Success": try: tgt_sr, audio_opt = opt wavfile.write( "%s/%s" % (opt_root, os.path.basename(path)), tgt_sr, audio_opt ) 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): 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(" ") ) pre_fun = _audio_pre_( agg=int(agg), model_path=os.path.join(weight_uvr5_root, model_name + ".pth"), device=device, is_half=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) 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) 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: 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): global n_spk, tgt_sr, net_g, vc, cpt if sid == []: global hubert_model if hubert_model != 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) if if_f0 == 1: net_g = SynthesizerTrnMs256NSFsid(*cpt["config"], is_half=is_half) else: net_g = SynthesizerTrnMs256NSFsid_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 == 1: net_g = SynthesizerTrnMs256NSFsid(*cpt["config"], is_half=is_half) else: net_g = SynthesizerTrnMs256NSFsid_nono(*cpt["config"]) del net_g.enc_q print(net_g.load_state_dict(cpt["weight"], strict=False)) # 不加这一行清不干净, 真奇葩 net_g.eval().to(device) if is_half: net_g = net_g.half() else: net_g = net_g.float() vc = VC(tgt_sr, device, is_half) n_spk = cpt["config"][-3] return {"visible": True, "maximum": n_spk, "__type__": "update"} def change_choices(): names = [] for name in os.listdir(weight_root): if name.endswith(".pth"): names.append(name) return {"choices": sorted(names), "__type__": "update"} def clean(): return {"value": "", "__type__": "update"} def change_f0(if_f0_3, sr2): # np7, f0method8,pretrained_G14,pretrained_D15 if if_f0_3 == "是": return ( {"visible": True, "__type__": "update"}, {"visible": True, "__type__": "update"}, "pretrained/f0G%s.pth" % sr2, "pretrained/f0D%s.pth" % sr2, ) return ( {"visible": False, "__type__": "update"}, {"visible": False, "__type__": "update"}, "pretrained/G%s.pth" % sr2, "pretrained/D%s.pth" % sr2, ) sr_dict = { "32k": 32000, "40k": 40000, "48k": 48000, } def if_done(done, p): while 1: if p.poll() == 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() == None: flag = 0 sleep(0.5) break if flag == 1: break done[0] = True def preprocess_dataset(trainset_dir, exp_dir, sr, n_p=ncpu): 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 = ( python_cmd + " trainset_preprocess_pipeline_print.py %s %s %s %s/logs/%s " % (trainset_dir, sr, n_p, now_dir, exp_dir) + str(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] == True: 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): 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 == "是": cmd = 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] == True: 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 = python_cmd + " extract_feature_print.py %s %s %s %s %s/logs/%s" % ( device, leng, idx, n_g, now_dir, exp_dir, ) 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] == True: 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): if if_f0_3 == "是": return "pretrained/f0G%s.pth" % sr2, "pretrained/f0D%s.pth" % sr2 else: return "pretrained/G%s.pth" % sr2, "pretrained/D%s.pth" % sr2 # 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, ): # 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) co256_dir = "%s/3_feature256" % (exp_dir) if if_f0_3 == "是": 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(co256_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(co256_dir)] ) opt = [] for name in names: if if_f0_3 == "是": opt.append( "%s/%s.wav|%s/%s.npy|%s/%s.wav.npy|%s/%s.wav.npy|%s" % ( gt_wavs_dir.replace("\\", "\\\\"), name, co256_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, co256_dir.replace("\\", "\\\\"), name, spk_id5, ) ) if if_f0_3 == "是": for _ in range(2): opt.append( "%s/logs/mute/0_gt_wavs/mute%s.wav|%s/logs/mute/3_feature256/mute.npy|%s/logs/mute/2a_f0/mute.wav.npy|%s/logs/mute/2b-f0nsf/mute.wav.npy|%s" % (now_dir, sr2, now_dir, 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_feature256/mute.npy|%s" % (now_dir, sr2, now_dir, 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 gpus16: cmd = ( python_cmd + " train_nsf_sim_cache_sid_load_pretrain.py -e %s -sr %s -f0 %s -bs %s -g %s -te %s -se %s -pg %s -pd %s -l %s -c %s" % ( exp_dir1, sr2, 1 if if_f0_3 == "是" else 0, batch_size12, gpus16, total_epoch11, save_epoch10, pretrained_G14, pretrained_D15, 1 if if_save_latest13 == "是" else 0, 1 if if_cache_gpu17 == "是" else 0, ) ) else: cmd = ( python_cmd + " train_nsf_sim_cache_sid_load_pretrain.py -e %s -sr %s -f0 %s -bs %s -te %s -se %s -pg %s -pd %s -l %s -c %s" % ( exp_dir1, sr2, 1 if if_f0_3 == "是" else 0, batch_size12, total_epoch11, save_epoch10, pretrained_G14, pretrained_D15, 1 if if_save_latest13 == "是" else 0, 1 if if_cache_gpu17 == "是" else 0, ) ) 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): exp_dir = "%s/logs/%s" % (now_dir, exp_dir1) os.makedirs(exp_dir, exist_ok=True) feature_dir = "%s/3_feature256" % (exp_dir) if os.path.exists(feature_dir) == False: 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!" npys = [] for name in sorted(listdir_res): phone = np.load("%s/%s" % (feature_dir, name)) npys.append(phone) big_npy = np.concatenate(npys, 0) # np.save("%s/total_fea.npy" % exp_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) infos = [] infos.append("%s,%s" % (big_npy.shape, n_ivf)) yield "\n".join(infos) index = faiss.index_factory(256, "IVF%s,Flat" % n_ivf) # index = faiss.index_factory(256, "IVF%s,PQ128x4fs,RFlat"%n_ivf) infos.append("training") yield "\n".join(infos) index_ivf = faiss.extract_index_ivf(index) # # index_ivf.nprobe = int(np.power(n_ivf,0.3)) index_ivf.nprobe = 1 index.train(big_npy) faiss.write_index( index, "%s/trained_IVF%s_Flat_nprobe_%s.index" % (exp_dir, n_ivf, index_ivf.nprobe), ) # faiss.write_index(index, '%s/trained_IVF%s_Flat_FastScan.index'%(exp_dir,n_ivf)) infos.append("adding") yield "\n".join(infos) index.add(big_npy) faiss.write_index( index, "%s/added_IVF%s_Flat_nprobe_%s.index" % (exp_dir, n_ivf, index_ivf.nprobe), ) infos.append("成功构建索引,added_IVF%s_Flat_nprobe_%s.index" % (n_ivf, index_ivf.nprobe)) # faiss.write_index(index, '%s/added_IVF%s_Flat_FastScan.index'%(exp_dir,n_ivf)) # infos.append("成功构建索引,added_IVF%s_Flat_FastScan.index"%(n_ivf)) 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, gpus6, np7, f0method8, save_epoch10, total_epoch11, batch_size12, if_save_latest13, pretrained_G14, pretrained_D15, gpus16, if_cache_gpu17, ): infos = [] def get_info_str(strr): infos.append(strr) return "\n".join(infos) os.makedirs("%s/logs/%s" % (now_dir, exp_dir1), exist_ok=True) #########step1:处理数据 open("%s/logs/%s/preprocess.log" % (now_dir, exp_dir1), "w").close() cmd = ( python_cmd + " trainset_preprocess_pipeline_print.py %s %s %s %s/logs/%s " % (trainset_dir4, sr_dict[sr2], ncpu, now_dir, exp_dir1) + str(noparallel) ) yield get_info_str("step1:正在处理数据") yield get_info_str(cmd) p = Popen(cmd, shell=True) p.wait() with open("%s/logs/%s/preprocess.log" % (now_dir, exp_dir1), "r") as f: print(f.read()) #########step2a:提取音高 open("%s/logs/%s/extract_f0_feature.log" % (now_dir, exp_dir1), "w") if if_f0_3 == "是": yield get_info_str("step2a:正在提取音高") cmd = python_cmd + " extract_f0_print.py %s/logs/%s %s %s" % ( now_dir, exp_dir1, np7, f0method8, ) yield get_info_str(cmd) p = Popen(cmd, shell=True, cwd=now_dir) p.wait() with open("%s/logs/%s/extract_f0_feature.log" % (now_dir, exp_dir1), "r") as f: print(f.read()) else: yield get_info_str("step2a:无需提取音高") #######step2b:提取特征 yield get_info_str("step2b:正在提取特征") gpus = gpus16.split("-") leng = len(gpus) ps = [] for idx, n_g in enumerate(gpus): cmd = python_cmd + " extract_feature_print.py %s %s %s %s %s/logs/%s" % ( device, leng, idx, n_g, now_dir, exp_dir1, ) 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("%s/logs/%s/extract_f0_feature.log" % (now_dir, exp_dir1), "r") as f: print(f.read()) #######step3a:训练模型 yield get_info_str("step3a:正在训练模型") # 生成filelist exp_dir = "%s/logs/%s" % (now_dir, exp_dir1) gt_wavs_dir = "%s/0_gt_wavs" % (exp_dir) co256_dir = "%s/3_feature256" % (exp_dir) if if_f0_3 == "是": 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(co256_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(co256_dir)] ) opt = [] for name in names: if if_f0_3 == "是": opt.append( "%s/%s.wav|%s/%s.npy|%s/%s.wav.npy|%s/%s.wav.npy|%s" % ( gt_wavs_dir.replace("\\", "\\\\"), name, co256_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, co256_dir.replace("\\", "\\\\"), name, spk_id5, ) ) if if_f0_3 == "是": for _ in range(2): opt.append( "%s/logs/mute/0_gt_wavs/mute%s.wav|%s/logs/mute/3_feature256/mute.npy|%s/logs/mute/2a_f0/mute.wav.npy|%s/logs/mute/2b-f0nsf/mute.wav.npy|%s" % (now_dir, sr2, now_dir, 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_feature256/mute.npy|%s" % (now_dir, sr2, now_dir, spk_id5) ) shuffle(opt) with open("%s/filelist.txt" % exp_dir, "w") as f: f.write("\n".join(opt)) yield get_info_str("write filelist done") if gpus16: cmd = ( python_cmd + " train_nsf_sim_cache_sid_load_pretrain.py -e %s -sr %s -f0 %s -bs %s -g %s -te %s -se %s -pg %s -pd %s -l %s -c %s" % ( exp_dir1, sr2, 1 if if_f0_3 == "是" else 0, batch_size12, gpus16, total_epoch11, save_epoch10, pretrained_G14, pretrained_D15, 1 if if_save_latest13 == "是" else 0, 1 if if_cache_gpu17 == "是" else 0, ) ) else: cmd = ( python_cmd + " train_nsf_sim_cache_sid_load_pretrain.py -e %s -sr %s -f0 %s -bs %s -te %s -se %s -pg %s -pd %s -l %s -c %s" % ( exp_dir1, sr2, 1 if if_f0_3 == "是" else 0, batch_size12, total_epoch11, save_epoch10, pretrained_G14, pretrained_D15, 1 if if_save_latest13 == "是" else 0, 1 if if_cache_gpu17 == "是" else 0, ) ) yield get_info_str(cmd) p = Popen(cmd, shell=True, cwd=now_dir) p.wait() yield get_info_str("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 feature_dir = "%s/3_feature256" % (exp_dir) 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) # np.save("%s/total_fea.npy" % exp_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, "IVF%s,Flat" % n_ivf) yield get_info_str("training index") index_ivf = faiss.extract_index_ivf(index) # # index_ivf.nprobe = int(np.power(n_ivf,0.3)) index_ivf.nprobe = 1 index.train(big_npy) faiss.write_index( index, "%s/trained_IVF%s_Flat_nprobe_%s.index" % (exp_dir, n_ivf, index_ivf.nprobe), ) yield get_info_str("adding index") index.add(big_npy) faiss.write_index( index, "%s/added_IVF%s_Flat_nprobe_%s.index" % (exp_dir, n_ivf, index_ivf.nprobe), ) yield get_info_str( "成功构建索引, added_IVF%s_Flat_nprobe_%s.index" % (n_ivf, index_ivf.nprobe) ) yield get_info_str("全流程结束!") # ckpt_path2.change(change_info_,[ckpt_path2],[sr__,if_f0__]) def change_info_(ckpt_path): if ( os.path.exists(ckpt_path.replace(os.path.basename(ckpt_path), "train.log")) == False ): return {"__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"] return sr, str(f0) except: traceback.print_exc() return {"__type__": "update"}, {"__type__": "update"} from infer_pack.models_onnx_moess import SynthesizerTrnMs256NSFsidM from infer_pack.models_onnx import SynthesizerTrnMs256NSFsidO def export_onnx(ModelPath, ExportedPath, MoeVS=True): hidden_channels = 256 # hidden_channels,为768Vec做准备 cpt = torch.load(ModelPath, map_location="cpu") cpt["config"][-3] = cpt["weight"]["emb_g.weight"].shape[0] # n_spk print(*cpt["config"]) test_phone = torch.rand(1, 200, hidden_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) if MoeVS: net_g = SynthesizerTrnMs256NSFsidM( *cpt["config"], is_half=False ) # 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", ] 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=16, verbose=False, input_names=input_names, output_names=output_names, ) else: net_g = SynthesizerTrnMs256NSFsidO( *cpt["config"], is_half=False ) # 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"] output_names = [ "audio", ] 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), ), ExportedPath, dynamic_axes={ "phone": [1], "pitch": [1], "pitchf": [1], }, do_constant_folding=False, opset_version=16, verbose=False, input_names=input_names, output_names=output_names, ) return "Finished" with gr.Blocks() as app: gr.Markdown( value=i18n( "本软件以MIT协议开源, 作者不对软件具备任何控制力, 使用软件者、传播软件导出的声音者自负全责.
如不认可该条款, 则不能使用或引用软件包内任何代码和文件. 详见根目录使用需遵守的协议-LICENSE.txt." ) ) with gr.Tabs(): with gr.TabItem(i18n("模型推理")): with gr.Row(): sid0 = gr.Dropdown(label=i18n("推理音色"), choices=sorted(names)) refresh_button = gr.Button(i18n("刷新音色列表"), variant="primary") refresh_button.click(fn=change_choices, inputs=[], outputs=[sid0]) clean_button = gr.Button(i18n("卸载音色省显存"), variant="primary") spk_item = gr.Slider( minimum=0, maximum=2333, step=1, label=i18n("请选择说话人id"), value=0, visible=False, interactive=True, ) clean_button.click(fn=clean, inputs=[], outputs=[sid0]) sid0.change( fn=get_vc, inputs=[sid0], outputs=[spk_item], ) with gr.Group(): gr.Markdown( value=i18n("男转女推荐+12key, 女转男推荐-12key, 如果音域爆炸导致音色失真也可以自己调整到合适音域. ") ) with gr.Row(): with gr.Column(): vc_transform0 = gr.Number( label=i18n("变调(整数, 半音数量, 升八度12降八度-12)"), value=0 ) input_audio0 = gr.Textbox( label=i18n("输入待处理音频文件路径(默认是正确格式示例)"), value="E:\\codes\\py39\\vits_vc_gpu_train\\todo-songs\\冬之花clip1.wav", ) f0method0 = gr.Radio( label=i18n("选择音高提取算法,输入歌声可用pm提速,harvest低音好但巨慢无比"), choices=["pm", "harvest"], value="pm", interactive=True, ) with gr.Column(): file_index1 = gr.Textbox( label=i18n("特征检索库文件路径"), value="E:\\codes\\py39\\vits_vc_gpu_train\\logs\\mi-test-1key\\added_IVF677_Flat_nprobe_7.index", interactive=True, ) # file_big_npy1 = gr.Textbox( # label=i18n("特征文件路径"), # value="E:\\codes\py39\\vits_vc_gpu_train\\logs\\mi-test-1key\\total_fea.npy", # interactive=True, # ) index_rate1 = gr.Slider( minimum=0, maximum=1, label="检索特征占比", value=0.76, interactive=True, ) f0_file = gr.File(label=i18n("F0曲线文件, 可选, 一行一个音高, 代替默认F0及升降调")) but0 = gr.Button(i18n("转换"), variant="primary") with gr.Column(): vc_output1 = gr.Textbox(label=i18n("输出信息")) vc_output2 = gr.Audio(label=i18n("输出音频(右下角三个点,点了可以下载)")) but0.click( vc_single, [ spk_item, input_audio0, vc_transform0, f0_file, f0method0, file_index1, # file_big_npy1, index_rate1, ], [vc_output1, vc_output2], ) with gr.Group(): gr.Markdown( value=i18n("批量转换, 输入待转换音频文件夹, 或上传多个音频文件, 在指定文件夹(默认opt)下输出转换的音频. ") ) with gr.Row(): with gr.Column(): vc_transform1 = gr.Number( label=i18n("变调(整数, 半音数量, 升八度12降八度-12)"), value=0 ) opt_input = gr.Textbox(label=i18n("指定输出文件夹"), value="opt") f0method1 = gr.Radio( label=i18n("选择音高提取算法,输入歌声可用pm提速,harvest低音好但巨慢无比"), choices=["pm", "harvest"], value="pm", interactive=True, ) with gr.Column(): file_index2 = gr.Textbox( label=i18n("特征检索库文件路径"), value="E:\\codes\\py39\\vits_vc_gpu_train\\logs\\mi-test-1key\\added_IVF677_Flat_nprobe_7.index", interactive=True, ) # file_big_npy2 = gr.Textbox( # label=i18n("特征文件路径"), # value="E:\\codes\\py39\\vits_vc_gpu_train\\logs\\mi-test-1key\\total_fea.npy", # interactive=True, # ) index_rate2 = gr.Slider( minimum=0, maximum=1, label=i18n("检索特征占比"), value=1, interactive=True, ) with gr.Column(): dir_input = gr.Textbox( label=i18n("输入待处理音频文件夹路径(去文件管理器地址栏拷就行了)"), value="E:\codes\py39\\vits_vc_gpu_train\\todo-songs", ) inputs = gr.File( file_count="multiple", label=i18n("也可批量输入音频文件, 二选一, 优先读文件夹") ) but1 = gr.Button(i18n("转换"), variant="primary") vc_output3 = gr.Textbox(label=i18n("输出信息")) but1.click( vc_multi, [ spk_item, dir_input, opt_input, inputs, vc_transform1, f0method1, file_index2, # file_big_npy2, index_rate2, ], [vc_output3], ) with gr.TabItem(i18n("伴奏人声分离")): with gr.Group(): gr.Markdown( value=i18n( "人声伴奏分离批量处理, 使用UVR5模型.
不带和声用HP2, 带和声且提取的人声不需要和声用HP5
合格的文件夹路径格式举例: E:\\codes\\py39\\vits_vc_gpu\\白鹭霜华测试样例(去文件管理器地址栏拷就行了)" ) ) with gr.Row(): with gr.Column(): dir_wav_input = gr.Textbox( label=i18n("输入待处理音频文件夹路径"), value="E:\\codes\\py39\\vits_vc_gpu_train\\todo-songs", ) wav_inputs = gr.File( file_count="multiple", label=i18n("也可批量输入音频文件, 二选一, 优先读文件夹") ) with gr.Column(): model_choose = gr.Dropdown(label=i18n("模型"), choices=uvr5_names) agg = gr.Slider( minimum=0, maximum=20, step=1, label="人声提取激进程度", value=10, interactive=True, visible=False, # 先不开放调整 ) opt_vocal_root = gr.Textbox( label=i18n("指定输出人声文件夹"), value="opt" ) opt_ins_root = gr.Textbox(label=i18n("指定输出乐器文件夹"), value="opt") but2 = gr.Button(i18n("转换"), variant="primary") vc_output4 = gr.Textbox(label=i18n("输出信息")) but2.click( uvr, [ model_choose, dir_wav_input, opt_vocal_root, wav_inputs, opt_ins_root, agg, ], [vc_output4], ) with gr.TabItem(i18n("训练")): gr.Markdown( value=i18n( "step1: 填写实验配置. 实验数据放在logs下, 每个实验一个文件夹, 需手工输入实验名路径, 内含实验配置, 日志, 训练得到的模型文件. " ) ) with gr.Row(): exp_dir1 = gr.Textbox(label=i18n("输入实验名"), value="mi-test") sr2 = gr.Radio( label=i18n("目标采样率"), choices=["32k", "40k", "48k"], value="40k", interactive=True, ) if_f0_3 = gr.Radio( label=i18n("模型是否带音高指导(唱歌一定要, 语音可以不要)"), choices=["是", "否"], value="是", interactive=True, ) with gr.Group(): # 暂时单人的, 后面支持最多4人的#数据处理 gr.Markdown( value=i18n( "step2a: 自动遍历训练文件夹下所有可解码成音频的文件并进行切片归一化, 在实验目录下生成2个wav文件夹; 暂时只支持单人训练. " ) ) with gr.Row(): trainset_dir4 = gr.Textbox( label=i18n("输入训练文件夹路径"), value="E:\\语音音频+标注\\米津玄师\\src" ) spk_id5 = gr.Slider( minimum=0, maximum=4, step=1, label=i18n("请指定说话人id"), value=0, interactive=True, ) but1 = gr.Button(i18n("处理数据"), variant="primary") info1 = gr.Textbox(label=i18n("输出信息"), value="") but1.click( preprocess_dataset, [trainset_dir4, exp_dir1, sr2], [info1] ) with gr.Group(): gr.Markdown(value=i18n("step2b: 使用CPU提取音高(如果模型带音高), 使用GPU提取特征(选择卡号)")) with gr.Row(): with gr.Column(): gpus6 = gr.Textbox( label=i18n("以-分隔输入使用的卡号, 例如 0-1-2 使用卡0和卡1和卡2"), value=gpus, interactive=True, ) gpu_info9 = gr.Textbox(label=i18n("显卡信息"), value=gpu_info) with gr.Column(): np7 = gr.Slider( minimum=0, maximum=ncpu, step=1, label=i18n("提取音高使用的CPU进程数"), value=ncpu, interactive=True, ) f0method8 = gr.Radio( label=i18n( "选择音高提取算法:输入歌声可用pm提速,高质量语音但CPU差可用dio提速,harvest质量更好但慢" ), choices=["pm", "harvest", "dio"], value="harvest", interactive=True, ) but2 = gr.Button(i18n("特征提取"), variant="primary") info2 = gr.Textbox(label=i18n("输出信息"), value="", max_lines=8) but2.click( extract_f0_feature, [gpus6, np7, f0method8, if_f0_3, exp_dir1], [info2], ) with gr.Group(): gr.Markdown(value=i18n("step3: 填写训练设置, 开始训练模型和索引")) with gr.Row(): save_epoch10 = gr.Slider( minimum=0, maximum=50, step=1, label=i18n("保存频率save_every_epoch"), value=5, interactive=True, ) total_epoch11 = gr.Slider( minimum=0, maximum=1000, step=1, label=i18n("总训练轮数total_epoch"), value=20, interactive=True, ) batch_size12 = gr.Slider( minimum=0, maximum=40, step=1, label="每张显卡的batch_size", value=default_batch_size, interactive=True, ) if_save_latest13 = gr.Radio( label=i18n("是否仅保存最新的ckpt文件以节省硬盘空间"), choices=["是", "否"], value="否", interactive=True, ) if_cache_gpu17 = gr.Radio( label=i18n( "是否缓存所有训练集至显存. 10min以下小数据可缓存以加速训练, 大数据缓存会炸显存也加不了多少速" ), choices=["是", "否"], value="否", interactive=True, ) with gr.Row(): pretrained_G14 = gr.Textbox( label=i18n("加载预训练底模G路径"), value="pretrained/f0G40k.pth", interactive=True, ) pretrained_D15 = gr.Textbox( label=i18n("加载预训练底模D路径"), value="pretrained/f0D40k.pth", interactive=True, ) sr2.change( change_sr2, [sr2, if_f0_3], [pretrained_G14, pretrained_D15] ) if_f0_3.change( change_f0, [if_f0_3, sr2], [np7, f0method8, pretrained_G14, pretrained_D15], ) gpus16 = gr.Textbox( label=i18n("以-分隔输入使用的卡号, 例如 0-1-2 使用卡0和卡1和卡2"), value=gpus, interactive=True, ) but3 = gr.Button(i18n("训练模型"), variant="primary") but4 = gr.Button(i18n("训练特征索引"), variant="primary") but5 = gr.Button(i18n("一键训练"), variant="primary") info3 = gr.Textbox(label=i18n("输出信息"), value="", max_lines=10) but3.click( 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, ], info3, ) but4.click(train_index, [exp_dir1], info3) 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, ) with gr.TabItem(i18n("ckpt处理")): with gr.Group(): gr.Markdown(value=i18n("模型融合, 可用于测试音色融合")) with gr.Row(): ckpt_a = gr.Textbox(label=i18n("A模型路径"), value="", interactive=True) ckpt_b = gr.Textbox(label=i18n("B模型路径"), value="", interactive=True) alpha_a = gr.Slider( minimum=0, maximum=1, label=i18n("A模型权重"), value=0.5, interactive=True, ) with gr.Row(): sr_ = gr.Radio( label=i18n("目标采样率"), choices=["32k", "40k", "48k"], value="40k", interactive=True, ) if_f0_ = gr.Radio( label=i18n("模型是否带音高指导"), choices=["是", "否"], value="是", interactive=True, ) info__ = gr.Textbox( label=i18n("要置入的模型信息"), value="", max_lines=8, interactive=True ) name_to_save0 = gr.Textbox( label=i18n("保存的模型名不带后缀"), value="", max_lines=1, interactive=True, ) with gr.Row(): but6 = gr.Button(i18n("融合"), variant="primary") info4 = gr.Textbox(label=i18n("输出信息"), value="", max_lines=8) but6.click( merge, [ckpt_a, ckpt_b, alpha_a, sr_, if_f0_, info__, name_to_save0], info4, ) # def merge(path1,path2,alpha1,sr,f0,info): with gr.Group(): gr.Markdown(value=i18n("修改模型信息(仅支持weights文件夹下提取的小模型文件)")) with gr.Row(): ckpt_path0 = gr.Textbox( label=i18n("模型路径"), value="", interactive=True ) info_ = gr.Textbox( label=i18n("要改的模型信息"), value="", max_lines=8, interactive=True ) name_to_save1 = gr.Textbox( label=i18n("保存的文件名, 默认空为和源文件同名"), value="", max_lines=8, interactive=True, ) with gr.Row(): but7 = gr.Button(i18n("修改"), variant="primary") info5 = gr.Textbox(label=i18n("输出信息"), value="", max_lines=8) but7.click(change_info, [ckpt_path0, info_, name_to_save1], info5) with gr.Group(): gr.Markdown(value=i18n("查看模型信息(仅支持weights文件夹下提取的小模型文件)")) with gr.Row(): ckpt_path1 = gr.Textbox( label=i18n("模型路径"), value="", interactive=True ) but8 = gr.Button(i18n("查看"), variant="primary") info6 = gr.Textbox(label=i18n("输出信息"), value="", max_lines=8) but8.click(show_info, [ckpt_path1], info6) with gr.Group(): gr.Markdown( value=i18n( "模型提取(输入logs文件夹下大文件模型路径),适用于训一半不想训了模型没有自动提取保存小文件模型,或者想测试中间模型的情况" ) ) with gr.Row(): ckpt_path2 = gr.Textbox( label=i18n("模型路径"), value="E:\\codes\\py39\\logs\\mi-test_f0_48k\\G_23333.pth", interactive=True, ) save_name = gr.Textbox( label=i18n("保存名"), value="", interactive=True ) sr__ = gr.Radio( label=i18n("目标采样率"), choices=["32k", "40k", "48k"], value="40k", interactive=True, ) if_f0__ = gr.Radio( label=i18n("模型是否带音高指导,1是0否"), choices=["1", "0"], value="1", interactive=True, ) info___ = gr.Textbox( label=i18n("要置入的模型信息"), value="", max_lines=8, interactive=True ) but9 = gr.Button(i18n("提取"), variant="primary") info7 = gr.Textbox(label=i18n("输出信息"), value="", max_lines=8) ckpt_path2.change(change_info_, [ckpt_path2], [sr__, if_f0__]) but9.click( extract_small_model, [ckpt_path2, save_name, sr__, if_f0__, info___], info7, ) with gr.TabItem(i18n("Onnx导出")): with gr.Row(): ckpt_dir = gr.Textbox(label=i18n("RVC模型路径"), value="", interactive=True) with gr.Row(): onnx_dir = gr.Textbox( label=i18n("Onnx输出路径"), value="", interactive=True ) with gr.Row(): moevs = gr.Checkbox(label=i18n("MoeVS模型"), value=True) infoOnnx = gr.Label(label="Null") with gr.Row(): butOnnx = gr.Button(i18n("导出Onnx模型"), variant="primary") butOnnx.click(export_onnx, [ckpt_dir, onnx_dir, moevs], infoOnnx) # with gr.TabItem(i18n("招募音高曲线前端编辑器")): # gr.Markdown(value=i18n("加开发群联系我xxxxx")) # with gr.TabItem(i18n("点击查看交流、问题反馈群号")): # gr.Markdown(value=i18n("xxxxx")) if iscolab: app.queue(concurrency_count=511, max_size=1022).launch(share=True) else: app.queue(concurrency_count=511, max_size=1022).launch( server_name="0.0.0.0", inbrowser=not noautoopen, server_port=listen_port, quiet=True, )