Iamanorange comments

Results 60 comments of


                                            Iamanorange

Cannot compile on ubuntu 18.04

Use gcc 5.

Cannot compile on ubuntu 18.04

https://github.com/sampepose/flownet2-tf/issues/43#issuecomment-406945155

Cannot compile on ubuntu 18.04

https://github.com/sampepose/flownet2-tf/issues/28#issuecomment-406941839

Compiling problem: invalid redeclaration of type name & cannot overload functions distinguished by return type alone

Try CUDA 9? Make sure your cuda, nvcc and tensorflow have compatible versions. e.g. cuda 10 with tf 1.13-1.14, cuda 9 with tf 1.5-1.12, cuda 8 with tf below 1.4....

Can not download weights?

The download link is down. I have a backup [here](https://drive.google.com/file/d/1cft8EvnsBL5w4-REUeAaVWLeRx39hyHE/view?usp=sharing).

Can not download weights?

OK. Pick one that suits you.

Can not download weights?

Mega: https://mega.nz/#!xk5DVYwB!hG2u_3ho8LWJl660Z5e0z-C0SlytB4SvlSkuoLEd568

There was a problem in successive calling to the function that generated the optical flow

Paste full error log. Or at least one line including the error.

fatal error: nsync_cv.h: No such file or directory

`nsync_cv.h` should be in `$TF_INC/external/nsync/public`. You can manually include it. For reference: https://github.com/tensorflow/tensorflow/issues/12482

Is flownet2-tf cuda 10 compatible?

Yes, CUDA 10 is fine. 1. Add `-DNDEBUG` to CFLAGS. For more information: https://github.com/tensorflow/tensorflow/issues/22766 2. Add `--expt-relaxed-constexpr` to GPUCFLAGS as it required.