Problem when inference with the model converted to pb

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

723 stars 139 forks source link

Problem when inference with the model converted to pb #37

Closed MaTJ89 closed 4 years ago

MaTJ89 commented 4 years ago

Hi, I find this project very interesting, so I hope to deploy this on web (non-commercial use) for people to have fun with this. After I convert the model to pb, the inference doesn't work. It shows Input 0 of node encoder/g_e0_bn/scale/Assign_1 was passed float from encoder/g_e0_bn/scale_1:0 incompatible with expected float_ref

I don't know much about tensorflow, or AI. I'm not sure if I did this persistence right. Appreciating for anyone that could help.

asanakoy commented 4 years ago

HI,

Thanks for the interest! I'm very excited to see your web-app!

I'm not sure abot your error. But maybe you have to convert your input tensor to float32. We could help you even more if you provided some minimal code example to reproduce your erorr.

MaTJ89 commented 4 years ago

Never mind, I figure it out. It turns out that I used a placeholder with wrong type as input.