Hi @Timen. I also built dirt with the changes in this pull request, but running square_test.py failed. The rasteriser didn't seem to give the expected result. Did you try running it?

Hey @Grieverheart I mostly tested it with the deferred.py sample, so maybe try that one :)

@Timen I had to make some changes to deferred.py, including adding 'tf.compat.v1.disable_eager_execution()' at the start of main to make it work, but the image I get does not like what you would expect:

@Grieverheart I updated the samples to tf 2.4.1 and fixed some issues saving images. However all 3 samples produced correct results for me as can be seen in these images:

I can't say what the issue on your side is but I hope these results will encourage you to dig a bit deeper, as it should be possible to get it to work 👍

@Timen Thanks for your help. After a lot of digging around, I found out that the problem I was having was that there seems to be an incompatibility of the default generated PTX with the GPU I'm using (a Tesla T4). Basically, I have to generate PTX for at least compute_70. In the end, I just set '-arch sm_75', and it works like a charm.

Thanks for the contribution @Timen -- I'll test it with some other versions of tensorflow before merging, to check everything is still ok there. In the changes to the samples, please don't add a dependency on PIL and numpy -- only some small changes to the tf calls for 2.x compatibility -- all the required functions are still present, just with slightly different names.