Closed QiaoYang-CV closed 5 months ago
when I run train.py, some error was reported:
if score_t[idx] > score_s[idx]:
IndexError: too many indices for array: array is 0-dimensional, but 1 were indexed
but, when I modify one code in trainer.py, ie:
self.iqa_metric = pyiqa.create_metric('musiq', as_loss=True).cuda(),
---->
self.iqa_metric = pyiqa.create_metric('musiq', as_loss=False).cuda(),
this error is disappears and code can be trained.
Specifically, I think this error is caused by "N = teacher_predict.shape[0]". But I don't know if modifying "as_loss" is correct.
Can you explain it to me ? Look forward to your reply.
when I run train.py, some error was reported:
IndexError: too many indices for array: array is 0-dimensional, but 1 were indexed
but, when I modify one code in trainer.py, ie:
self.iqa_metric = pyiqa.create_metric('musiq', as_loss=True).cuda(),
---->
self.iqa_metric = pyiqa.create_metric('musiq', as_loss=False).cuda(),
this error is disappears and code can be trained.
Specifically, I think this error is caused by "N = teacher_predict.shape[0]". But I don't know if modifying "as_loss" is correct.
Can you explain it to me ? Look forward to your reply.