search

primepake / wav2lip_288x288

MIT License
560 stars 143 forks source link

dataset standard #77

Closed SuperGoodGame closed 11 months ago

SuperGoodGame commented 11 months ago

Traceback (most recent call last): File "color_syncnet_train.py", line 279, in nepochs=hparams.nepochs) File "color_syncnet_train.py", line 162, in train loss.backward() File "/home/jiahuaw/anaconda3/envs/wav2lip/lib/python3.7/site-packages/torch/tensor.py", line 221, in backward torch.autograd.backward(self, gradient, retain_graph, create_graph) File "/home/jiahuaw/anaconda3/envs/wav2lip/lib/python3.7/site-packages/torch/autograd/init.py", line 132, in backward allow_unreachable=True) # allow_unreachable flag RuntimeError: CUDA error: device-side assert triggered /pytorch/aten/src/ATen/native/cuda/Loss.cu:102: operator(): block: [0,0,0], thread: [8,0,0] Assertion input_val >= zero && input_val <= one failed. i think my dataset have some problem.And i don.t know how to prepare. i just preprocess it with the orignal wav2lip program 我只用了原版wav2lip 的数据处理，不知道怎么再对数据集处理，有人知道数据集需要的标准是什么吗

SuperGoodGame commented 11 months ago

我看了cleardata.py 上面好像是需要对帧数有要求，有人知道标准吗，十分感谢

Arvin-keke commented 11 months ago

25帧

SuperGoodGame commented 11 months ago

我的数据确实是25帧，也都有人脸，有人遇到过这个报错吗 mydataset seems match the standard ，have someone meet this error

SuperGoodGame commented 11 months ago

the problem has been soloved the method #60