error training my own dataset

kotaxyz commented 1 year ago

i exported dataset of my own animation from unity but when i try to train it i give this error

(DeepPhaseSubmission) C:\Users\Genesis\deep-motion\PyTorch\PAE>python Train.py
Loading Dataset/DataShape.txt
Loading Dataset/Sequences.txt
Generating Data Sequences
Data Sequences: 25650
Test Sequences: 250
Training Phases
Traceback (most recent call last):
  File "Train.py", line 142, in <module>
    yPred, latent, signal, params = network(train_batch)
  File "C:\Users\Genesis\anaconda3\envs\DeepPhaseSubmission\lib\site-packages\torch\nn\modules\module.py", line 1102, in _call_impl
    return forward_call(*input, **kwargs)
  File "C:\Users\Genesis\deep-motion\PyTorch\PAE\PAE.py", line 79, in forward
    y = y.reshape(y.shape[0], self.input_channels, self.time_range)
RuntimeError: shape '[32, 78, 121]' is invalid for input of size 441408

(DeepPhaseSubmission) C:\Users\Genesis\deep-motion\PyTorch\PAE>

note that i use the 2022 updated ai4animation

sebastianstarke commented 1 year ago

Did you doublecheck the number of joints and framerate, which may be different for your own dataset? Could it be that your character has 38 joints? If so, you need to adjust the hyperparameter in the Train.py file.

kotaxyz commented 1 year ago

thanks i will try it out

kotaxyz commented 1 year ago

it worked by adjusting the joint number in the parameters thanks alot

sebastianstarke / AI4Animation

error training my own dataset #100