Spandrel gives your project support for various PyTorch architectures meant for AI Super-Resolution, restoration, and inpainting. Based on the model support implemented in chaiNNer.
MIT License
105
stars
7
forks
source link
Improved error message for failed size fuzzing #131
Instead of telling you the first size that fails, it will now go through all sizes and tell you which ones failed and which ones were valid.
This makes it easy to use the fuzz test as a way to figure out the size requirements of a model. E.g. in #130, I just ran the fuzz test with default size req (so none at all), and it told me that all sizes failed except for 8. So I required multiples of 8, and then the fuzzing said everything is valid.
Basically, I used the fuzz test to automatically determine the size requirements.
Instead of telling you the first size that fails, it will now go through all sizes and tell you which ones failed and which ones were valid.
This makes it easy to use the fuzz test as a way to figure out the size requirements of a model. E.g. in #130, I just ran the fuzz test with default size req (so none at all), and it told me that all sizes failed except for 8. So I required multiples of 8, and then the fuzzing said everything is valid. Basically, I used the fuzz test to automatically determine the size requirements.