ComplexDouble error when using for DiffSinger

[hilmiyafia]

Hello, I've trained a ddsp model and put it in the diffsinger checkpoints folder. At the validation step it errors:

Traceback (most recent call last): File "/content/DiffSinger/run.py", line 15, in run_task() File "/content/DiffSinger/run.py", line 11, in run_task task_cls.start() File "/content/DiffSinger/basics/base_task.py", line 242, in start trainer.fit(task) File "/content/DiffSinger/utils/pl_utils.py", line 497, in fit self.run_pretrain_routine(model) File "/content/DiffSinger/utils/pl_utils.py", line 573, in run_pretrain_routine self.evaluate(model, self.get_val_dataloaders(), self.num_sanity_val_steps, self.testing) File "/content/DiffSinger/utils/pl_utils.py", line 1193, in evaluate output = self.evaluation_forward(model, File "/content/DiffSinger/utils/pl_utils.py", line 1317, in evaluation_forward output = model.validation_step(args) File "/content/DiffSinger/src/naive_task.py", line 85, in validation_step self.plot_wav(batch_idx, sample['mels'], model_out['mel_out'], gt_f0=gt_f0, pred_f0=pred_f0) File "/content/DiffSinger/src/diffspeech_task.py", line 120, in plot_wav gt_wav = self.vocoder.spec2wav(gt_mel, f0=gtf0) File "/content/DiffSinger/src/vocoders/ddsp.py", line 163, in spec2wav signal, , (s_h, s_n) = self.model(mel.to(self.device), f0.to(self.device)) File "/usr/local/lib/python3.9/dist-packages/torch/nn/modules/module.py", line 889, in _call_impl result = self.forward(input, **kwargs) RuntimeError: Unsupported value kind: ComplexDouble

[yxlllc]

I guess ComplexDouble is supported in higher version of pytorch.

[hilmiyafia]

@yxlllc Thank you! It's working now! 😊