search

CompVis / adaptive-style-transfer

source code for the ECCV18 paper A Style-Aware Content Loss for Real-time HD Style Transfer
https://compvis.github.io/adaptive-style-transfer/
GNU General Public License v3.0
730 stars 140 forks source link

colab ports: #33

Open Maven-art opened 4 years ago

Maven-art commented 4 years ago

Rough google colab ports of trainer and model. It worked, but i had to abandon due to my poor internet constantly disconnecting my colab instances and requiring babysitting, hope this helps others who want to play with the code:

Model: https://colab.research.google.com/drive/1Aie8A2wphXpkIGJ4ePcBlTnnSYSmF_Er

Trainer: https://colab.research.google.com/drive/1aL-fXNS7Jqjo0rqU4hEkOdpk6Ti5wgdy

lorrp1 commented 4 years ago

colab disconnect ip using too resources. dont use a vpn they are already blacklisted which result in almost instantly disconnection if you train something.

marlluslustosa commented 4 years ago

The original article or the github documentation does not mention this. But how long does it take to train with a dataset of 100 images of an artist, using a low-cost graphics card, such as a GeForce GTX?

can you cite your case?