Docker problem with runGan.py 4

runGan.py 4 produces some error I cannot fathom. †

root@b55c13b7b90f:/TecoGAN# python3 runGan.py 4 /usr/local/lib/python3.5/dist-packages/h5py/init.py:36: FutureWarning: Conversion of the second argument of issubdtype from float to np.floating is deprecated. In future, it will be treated as np.float64 == np.dtype(float).type. from ._conv import register_converters as _register_converters Using TensorFlow backend.

Preparing train_data [Config] Use random crop [Config] Use random crop [Config] Use random flip Sequenced batches: 28250, sequence length: 7 Preparing validation_data [Config] Use random crop [Config] Use random crop [Config] Use random flip Sequenced batches: 5085, sequence length: 7 tData count = 28250, steps per epoch 28250 Traceback (most recent call last): File “main.py”, line 287, in <module> Net = FRVSR( rdata.s_inputs, rdata.s_targets, FLAGS ) File “/TecoGAN/lib/Teco.py”, line 522, in FRVSR return TecoGAN(r_inputs, r_targets, FLAGS, False) File “/TecoGAN/lib/Teco.py”, line 434, in TecoGAN update_loss = exp_averager.apply(update_list) File “/usr/local/lib/python3.5/dist-packages/tensorflow/python/training/moving_averages.py”, line 388, in apply raise ValueError(“Moving average already computed for: %s” % var.name) ValueError: Moving average already computed for: generator_loss/content_loss/Mean:0

† probably just stupid.

[flutide]

Some questionable hack made it work.

Changed the Dockerfile a bit:

from tensorflow/tensorflow:1.15.0rc2-gpu-py3

run pip3 install \
numpy>=1.14.3 \
scipy>=1.0.1 \
scikit-image>=0.13.0 \
matplotlib>=1.5.1 \
pandas>=0.23.1 \
Keras==2.1.3 \
torch>=0.4.0 \
torchvision>=0.2.1 \
opencv-python>=4.2.0.34 \
ipython>=7.4.0

and changed lib\ops.py according to suggestion in https://github.com/thunil/TecoGAN/issues/20#issuecomment-632592051

[flutide]

A bit perhaps ¯_(ツ)_/¯ But than again, I’m surprised it works at all (at my end) with GPU support under Win10 (20215.1000), WSL2, Docker etc. Thanks again Tom for all valuable input, in this thread and the other ones!