Open peterslany opened 4 years ago
Probably it would be good practice specify versions for all requirements. Thanks
Any advice on how to downgrade? Getting stuck on going to 1.15
It would be useful to put into readme that this repo works only with TensorFlow version below 2.0, because the
contrib
attribute is no longer supported in later versions.Probably it would be good practice specify versions for all requirements. Thanks
maybe ive missed something - but: Implementation Details
All images were rendered on a machine with:
CPU: Intel Core i7-6800K @ 3.40GHz × 12
GPU: NVIDIA GeForce GTX 1080/PCIe/SSE2
OS: Linux Ubuntu 16.04.1 LTS 64-bit
CUDA: 8.0
python: 2.7.12
tensorflow: 0.10.0rc
opencv: 2.4.9.1
I did a little change on the code then run on tf 2.3.1 and it works fine.
First I simply change neural_style.py
into neural_style_v2.py
by using tf_upgrade_v2
(reference here)
Then delete all of the ".value" (e.g. h.value, w.value ...)
Finally use the adam opt instead of lbfgs
As @peterslany said the contrib
attribute is no longer supported in tf 2+ versions, but I think maybe it could be replaced with the opt here? If you think the result of using the adam opt isn't good enough. (I haven't tried it)
the env I'm using:
python=3.7.9
tensorflow=2.3.1
opencv=4.5.2
CUDA=10.0
cuDNN=7.6.5
Hope this helps!
I did a little change on the code then run on tf 2.3.1 and it works fine. First I simply change
neural_style.py
intoneural_style_v2.py
by usingtf_upgrade_v2
(reference here) Then delete all of the ".value" (e.g. h.value, w.value ...) Finally use the adam opt instead of lbfgs As @peterslany said thecontrib
attribute is no longer supported in tf 2+ versions, but I think maybe it could be replaced with the opt here? If you think the result of using the adam opt isn't good enough. (I haven't tried it)the env I'm using:
python=3.7.9 tensorflow=2.3.1 opencv=4.5.2 CUDA=10.0 cuDNN=7.6.5
Hope this helps!
Hey, would you mind sharing your version somewhere? I'm trying to make this work, it would help me a lot!
I did a little change on the code then run on tf 2.3.1 and it works fine. First I simply change
neural_style.py
intoneural_style_v2.py
by usingtf_upgrade_v2
(reference here) Then delete all of the ".value" (e.g. h.value, w.value ...) Finally use the adam opt instead of lbfgs As @peterslany said thecontrib
attribute is no longer supported in tf 2+ versions, but I think maybe it could be replaced with the opt here? If you think the result of using the adam opt isn't good enough. (I haven't tried it) the env I'm using:python=3.7.9 tensorflow=2.3.1 opencv=4.5.2 CUDA=10.0 cuDNN=7.6.5
Hope this helps!
Hey, would you mind sharing your version somewhere? I'm trying to make this work, it would help me a lot! neural_style.txt
I attached my version as .txt. It worked for me, maybe it'll help someone else, too
@BDinspel Do you have forked this with the whole update? Do you mind sharing it?
@BDinspel Thanks so much for this! You saved me a lot of effort. Works great.
I attached my version as .txt. It worked for me, maybe it'll help someone else, too
It would be useful to put into readme that this repo works only with TensorFlow version below 2.0, because the
contrib
attribute is no longer supported in later versions.