TypeError: prepare_vertices() missing 1 required positional argument: 'camera_rot'

[M202071895]

I ran into this problem when I was training. I looked at the source file of this function：def prepare_vertices(vertices, faces, camera_rot, camera_trans, camera_proj) （in kaolin.render.mesh）

Why can you successfully run this code? I strictly follow the installation process you provided

Thank you so much !

[tau-yihouxiang]

Hundred of problems come from mismatching package version. It is very likely that you didn't correctly install of Kaolin v0.9.0. Please first check whether you can run kaolin example normally.

[zhixuanli]

Hi, @tau-yihouxiang I have encountered this problem too, which cost me almost a whole day.

You have listed two versions of kaolin, including v0.9.0 and 5554bce48187d2acfcbce62cc5f8c511a082de92. And I have tried both of them.

If I use version v0.9.0, the same problem with this issue will happen. And I'm sure I have followed the official instructions(of yours and kaolin library).

If I use the second version 5554bce48187d2acfcbce62cc5f8c511a082de92, there is a problem when compiling the kaolin code. The error is more than one instance of overloaded function "log2" matches the argument list:.

Could you please experiment on a new virtual env and provide a foo version to us? I believe SMR is an essential work in this field and thanks for your patience.

Hoping for your replying.

[tau-yihouxiang]

Thank you for your reporting, I will reinstall kaolin v0.9.0 to check.

[zhixuanli]

The second version 5554bce48187d2acfcbce62cc5f8c511a082de92 is correct.

install kaolin

$ git clone --recursive https://github.com/NVIDIAGameWorks/kaolin
$ git checkout 5554bce48187d2acfcbce62cc5f8c511a082de92
$ python setup.py develop

fix the bug of log2

There is a type error in the code of kaolin/csrc/ops/conversions/unbatched_mcube/unbatched_mcube_cuda.cu

You can change #line 658~660 as following:

  gridSizeLog2.x = (int) log2(i * 1.0);
  gridSizeLog2.y = (int) log2(j * 1.0);
  gridSizeLog2.z = (int) log2(k * 1.0);

And now we can compile and run the training code.

Name:  ./log/Bird/SMR
[0/500][0/248]
LossD: 0.0010 lossD_real: -0.0000 lossD_fake: 0.0000 lossD_gp: 0.0010
lossR: 0.9353 lossR_fake: -0.0000 lossR_reg: 0.2225 lossR_data: 0.5515 lossR_IC: 0.1407 

Elapsed time in update: 0.844983
Name:  ./log/Bird/SMR
[0/500][1/248]
LossD: 0.0010 lossD_real: -0.0000 lossD_fake: 0.0000 lossD_gp: 0.0010
lossR: 1.0701 lossR_fake: -0.0000 lossR_reg: 0.3620 lossR_data: 0.5807 lossR_IC: 0.1064 

Elapsed time in update: 0.748916
Name:  ./log/Bird/SMR
[0/500][2/248]
LossD: 0.0010 lossD_real: -0.0000 lossD_fake: 0.0000 lossD_gp: 0.0010
lossR: 0.8825 lossR_fake: -0.0000 lossR_reg: 0.1489 lossR_data: 0.6115 lossR_IC: 0.1025 

Elapsed time in update: 0.744629
Name:  ./log/Bird/SMR
[0/500][3/248]
LossD: 0.0010 lossD_real: -0.0000 lossD_fake: 0.0000 lossD_gp: 0.0010
lossR: 0.8643 lossR_fake: -0.0000 lossR_reg: 0.1289 lossR_data: 0.5920 lossR_IC: 0.1240 

Elapsed time in update: 0.741169
Name:  ./log/Bird/SMR
[0/500][4/248]
LossD: 0.0010 lossD_real: -0.0000 lossD_fake: 0.0000 lossD_gp: 0.0010
lossR: 0.8159 lossR_fake: -0.0000 lossR_reg: 0.0868 lossR_data: 0.5955 lossR_IC: 0.1148 

Thanks for your awesome work!

[tau-yihouxiang]

I have built a new environment and set up the packages from pip install -r requirement , and it works. Also, thank you for your reminding of the existing of different versions of kaolin v0.9.0. (5554bce48187d2acfcbce62cc5f8c511a082de92) is the correct one.

[zhixuanli]

Bravo! Thanks for your rapid reply!