lengstrom / fast-style-transfer

TensorFlow CNN for fast style transfer ⚡🖥🎨🖼
10.91k stars 2.6k forks source link

error running style.py #270

Open YvesBB opened 3 years ago

YvesBB commented 3 years ago

I' m getting an error when i want to train with style.py , it runs perfectly if i evaluate a style transfer with evaluate.py

python style.py --style images/emoticon.jpg --checkpoint-dir checkpoint/ --content-weight 1.5e1 --checkpoint-iterations 1000 --batch-size 20

CUDA Version: 11.2 Python 3.9.4 Tensorflow 2.5.0-rc0 GeForce gtx 1060

pb.txt

keremkurtulus commented 3 years ago

I fixed it with https://github.com/lengstrom/fast-style-transfer/issues/10#issuecomment-257462430

mmethot commented 2 years ago

I had to specify a smaller batch-size. I removed the --batch-size command line option entirely from my command to let it take the default, which is 4 and which worked.