diff --git a/GPT_SoVITS/s2_train.py b/GPT_SoVITS/s2_train.py index 45b3bc05..bb91ac94 100644 --- a/GPT_SoVITS/s2_train.py +++ b/GPT_SoVITS/s2_train.py @@ -3,7 +3,7 @@ import warnings warnings.filterwarnings("ignore") import os -import utils +import GPT_SoVITS.utils as utils hps = utils.get_hparams(stage=2) os.environ["CUDA_VISIBLE_DEVICES"] = hps.train.gpu_numbers.replace("-", ",") @@ -71,7 +71,7 @@ def main(): def run(rank, n_gpus, hps): global global_step if rank == 0: - logger = GPT_SoVITS.utils.get_logger(hps.data.exp_dir) + logger = utils.get_logger(hps.data.exp_dir) logger.info(hps) # utils.check_git_hash(hps.s2_ckpt_dir) writer = SummaryWriter(log_dir=hps.s2_ckpt_dir) @@ -204,7 +204,7 @@ def run(rank, n_gpus, hps): net_d = net_d.to(device) try: # 如果能加载自动resume - _, _, _, epoch_str = GPT_SoVITS.utils.load_checkpoint( + _, _, _, epoch_str = utils.load_checkpoint( utils.latest_checkpoint_path("%s/logs_s2_%s" % (hps.data.exp_dir, hps.model.version), "D_*.pth"), net_d, optim_d, @@ -212,7 +212,7 @@ def run(rank, n_gpus, hps): if rank == 0: logger.info("loaded D") # _, _, _, epoch_str = utils.load_checkpoint(utils.latest_checkpoint_path(hps.model_dir, "G_*.pth"), net_g, optim_g,load_opt=0) - _, _, _, epoch_str = GPT_SoVITS.utils.load_checkpoint( + _, _, _, epoch_str = utils.load_checkpoint( utils.latest_checkpoint_path("%s/logs_s2_%s" % (hps.data.exp_dir, hps.model.version), "G_*.pth"), net_g, optim_g, @@ -479,30 +479,30 @@ def train_and_evaluate(rank, epoch, hps, nets, optims, schedulers, scaler, loade image_dict = None try: ###Some people installed the wrong version of matplotlib. image_dict = { - "slice/mel_org": GPT_SoVITS.utils.plot_spectrogram_to_numpy( + "slice/mel_org": utils.plot_spectrogram_to_numpy( y_mel[0].data.cpu().numpy(), ), - "slice/mel_gen": GPT_SoVITS.utils.plot_spectrogram_to_numpy( + "slice/mel_gen": utils.plot_spectrogram_to_numpy( y_hat_mel[0].data.cpu().numpy(), ), - "all/mel": GPT_SoVITS.utils.plot_spectrogram_to_numpy( + "all/mel": utils.plot_spectrogram_to_numpy( mel[0].data.cpu().numpy(), ), - "all/stats_ssl": GPT_SoVITS.utils.plot_spectrogram_to_numpy( + "all/stats_ssl": utils.plot_spectrogram_to_numpy( stats_ssl[0].data.cpu().numpy(), ), } except: pass if image_dict: - GPT_SoVITS.utils.summarize( + utils.summarize( writer=writer, global_step=global_step, images=image_dict, scalars=scalar_dict, ) else: - GPT_SoVITS.utils.summarize( + utils.summarize( writer=writer, global_step=global_step, scalars=scalar_dict, @@ -510,7 +510,7 @@ def train_and_evaluate(rank, epoch, hps, nets, optims, schedulers, scaler, loade global_step += 1 if epoch % hps.train.save_every_epoch == 0 and rank == 0: if hps.train.if_save_latest == 0: - GPT_SoVITS.utils.save_checkpoint( + utils.save_checkpoint( net_g, optim_g, hps.train.learning_rate, @@ -520,7 +520,7 @@ def train_and_evaluate(rank, epoch, hps, nets, optims, schedulers, scaler, loade "G_{}.pth".format(global_step), ), ) - GPT_SoVITS.utils.save_checkpoint( + utils.save_checkpoint( net_d, optim_d, hps.train.learning_rate, @@ -531,7 +531,7 @@ def train_and_evaluate(rank, epoch, hps, nets, optims, schedulers, scaler, loade ), ) else: - GPT_SoVITS.utils.save_checkpoint( + utils.save_checkpoint( net_g, optim_g, hps.train.learning_rate, @@ -541,7 +541,7 @@ def train_and_evaluate(rank, epoch, hps, nets, optims, schedulers, scaler, loade "G_{}.pth".format(233333333333), ), ) - GPT_SoVITS.utils.save_checkpoint( + utils.save_checkpoint( net_d, optim_d, hps.train.learning_rate, @@ -644,7 +644,7 @@ def evaluate(hps, generator, eval_loader, writer_eval): ) image_dict.update( { - f"gen/mel_{batch_idx}_{test}": GPT_SoVITS.utils.plot_spectrogram_to_numpy( + f"gen/mel_{batch_idx}_{test}": utils.plot_spectrogram_to_numpy( y_hat_mel[0].cpu().numpy(), ), } @@ -656,7 +656,7 @@ def evaluate(hps, generator, eval_loader, writer_eval): ) image_dict.update( { - f"gt/mel_{batch_idx}": GPT_SoVITS.utils.plot_spectrogram_to_numpy(mel[0].cpu().numpy()), + f"gt/mel_{batch_idx}": utils.plot_spectrogram_to_numpy(mel[0].cpu().numpy()), }, ) audio_dict.update({f"gt/audio_{batch_idx}": y[0, :, : y_lengths[0]]}) @@ -666,7 +666,7 @@ def evaluate(hps, generator, eval_loader, writer_eval): # f"gen/audio_{batch_idx}_style_pred": y_hat[0, :, :] # }) - GPT_SoVITS.utils.summarize( + utils.summarize( writer=writer_eval, global_step=global_step, images=image_dict, diff --git a/GPT_SoVITS/s2_train_v3.py b/GPT_SoVITS/s2_train_v3.py index 71d21967..620feff7 100644 --- a/GPT_SoVITS/s2_train_v3.py +++ b/GPT_SoVITS/s2_train_v3.py @@ -3,7 +3,7 @@ import warnings warnings.filterwarnings("ignore") import os -import utils +import GPT_SoVITS.utils as utils hps = utils.get_hparams(stage=2) os.environ["CUDA_VISIBLE_DEVICES"] = hps.train.gpu_numbers.replace("-", ",") @@ -23,20 +23,20 @@ logging.getLogger("h5py").setLevel(logging.INFO) logging.getLogger("numba").setLevel(logging.INFO) from random import randint -from module import commons -from module.data_utils import ( +from GPT_SoVITS.module import commons +from GPT_SoVITS.module.data_utils import ( DistributedBucketSampler, ) -from module.data_utils import ( +from GPT_SoVITS.module.data_utils import ( TextAudioSpeakerCollateV3 as TextAudioSpeakerCollate, ) -from module.data_utils import ( +from GPT_SoVITS.module.data_utils import ( TextAudioSpeakerLoaderV3 as TextAudioSpeakerLoader, ) -from module.models import ( +from GPT_SoVITS.module.models import ( SynthesizerTrnV3 as SynthesizerTrn, ) -from process_ckpt import savee +from GPT_SoVITS.process_ckpt import savee torch.backends.cudnn.benchmark = False torch.backends.cudnn.deterministic = False diff --git a/GPT_SoVITS/s2_train_v3_lora.py b/GPT_SoVITS/s2_train_v3_lora.py index ddeec4fc..cf07f828 100644 --- a/GPT_SoVITS/s2_train_v3_lora.py +++ b/GPT_SoVITS/s2_train_v3_lora.py @@ -3,7 +3,7 @@ import warnings warnings.filterwarnings("ignore") import os -import utils +import GPT_SoVITS.utils as utils hps = utils.get_hparams(stage=2) os.environ["CUDA_VISIBLE_DEVICES"] = hps.train.gpu_numbers.replace("-", ",") @@ -24,8 +24,8 @@ logging.getLogger("numba").setLevel(logging.INFO) from collections import OrderedDict as od from random import randint -from module import commons -from module.data_utils import ( +from GPT_SoVITS.module import commons +from GPT_SoVITS.module.data_utils import ( DistributedBucketSampler, TextAudioSpeakerCollateV3, TextAudioSpeakerLoaderV3, @@ -33,7 +33,7 @@ from module.data_utils import ( TextAudioSpeakerLoaderV4, ) -from module.models import ( +from GPT_SoVITS.module.models import ( SynthesizerTrnV3 as SynthesizerTrn, ) from peft import LoraConfig, get_peft_model