yxlllc / DDSP-SVC

Real-time end-to-end singing voice conversion system based on DDSP (Differentiable Digital Signal Processing)
MIT License
1.91k stars 252 forks source link

訓練階段遇到的問題 #88

Open guitarj opened 1 year ago

guitarj commented 1 year ago

Traceback (most recent call last): File "train_diff.py", line 86, in train(args, initial_global_step, model, optimizer, scheduler, vocoder, loader_train, loader_valid) File "/home/csmxj/DDSP-SVC/diffusion/solver_new.py", line 188, in train test_ddsp_loss, test_diff_loss = test(args, model, vocoder, loader_test, saver) File "/home/csmxj/DDSP-SVC/diffusion/solver_new.py", line 25, in test for bidx, data in enumerate(loader_test): File "/root/.miniconda/envs/ddsp/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 633, in next data = self._next_data() File "/root/.miniconda/envs/ddsp/lib/python3.8/site-packages/torch/utils/data/dataloader.py", line 677, in _next_data data = self._dataset_fetcher.fetch(index) # may raise StopIteration File "/root/.miniconda/envs/ddsp/lib/python3.8/site-packages/torch/utils/data/_utils/fetch.py", line 51, in fetch data = [self.dataset[idx] for idx in possibly_batched_index] File "/root/.miniconda/envs/ddsp/lib/python3.8/site-packages/torch/utils/data/_utils/fetch.py", line 51, in data = [self.dataset[idx] for idx in possibly_batched_index] File "/home/csmxj/DDSP-SVC/diffusion/data_loaders.py", line 203, in getitem return self.getitem( (file_idx + 1) % len(self.paths)) File "/home/csmxj/DDSP-SVC/diffusion/data_loaders.py", line 203, in getitem return self.getitem( (file_idx + 1) % len(self.paths)) File "/home/csmxj/DDSP-SVC/diffusion/data_loaders.py", line 203, in getitem return self.getitem( (file_idx + 1) % len(self.paths)) [Previous line repeated 989 more times] File "/home/csmxj/DDSP-SVC/diffusion/data_loaders.py", line 202, in getitem if data_buffer['duration'] < (self.waveform_sec + 0.1): RecursionError: maximum recursion depth exceeded in comparison

在訓練完一個區間後會做一次validation,就在這邊陷入RecursionError了。

yxlllc commented 1 year ago

数据集里音频切片的长度不能小于 2s (配置文件里有相关参数)