Open agustincaniglia opened 2 years ago
Same issue here, I even tried the trick to uninstall taichi, it launches with numba and says it's working but this time it's not actually making any images...
Same issue.
[W 10/24/22 22:23:18.542 43780] [D:/a/taichi/taichi/taichi/transforms/offload.cpp:emit_struct_for@236] Specified block dim 1024 is bigger than SNode element size 64. Clipping. [PIE] not working, photometric correction is disabled
Same issue in docker version running on WSL on Windows 11. Even happens when PIE is disabled on first generation.
I have exactly the same problem
Same issue
[W 10/26/22 22:46:52.339 392] [offload.cpp:emit_struct_for@236] Specified block dim 1024 is bigger than SNode element size 64. Clipping.
[PIE] not working, photometric correction is disabled
WSL, Windows 10, Photometric correction is not working, fpie is installed
Uninstalling taichi did work for me, but do note numba is very slow so depending on your system and your batch size you may need to wait a while--for me it was almost as long a wait per image as the initial generation.
Everything works fine, except for the photometric correction in my install on windows. I don't know aything about code or anything I just wanted to get some help. Is this expected as I saw something related to this not working in the original post, but I don't understand if there is a fix or not. This is the line I get:
100%|██████████████████████████████████████████████████████████████████████████████████| 51/51 [00:09<00:00, 5.13it/s] [W 10/24/22 09:48:57.309 6460] [D:/a/taichi/taichi/taichi/transforms/offload.cpp:emit_struct_for@236] Specified block dim 1024 is bigger than SNode element size 64. Clipping. [PIE] not working, photometric correction is disabled
The image is still generated but photometric correction is not being applied.
EDIT: Also sometimes I get the error even if I never enabled the photometric correction. And images seem to be generated but don't appear on the GUI. And I have to restart.. It happens when I used the inpainting+img2img (which btw I don't understand how to do img2img)