RuntimeError: Could not compile function about neural_kernels_code HOT 3 OPEN

Ah I am pretty sure the code only works on V100s or TitanV right now.

I am working on an implementation based on Jax that supports more GPU types but alas it is not out yet.

from neural_kernels_code.

@Vaishaal Yes, V100 can indeed run the code. But I find your generate_kernel_parallel function broke for some new dataset (MNIST-style) on a single GPU, while the single-thread generate_kernel works fine. So I suggest you make the single-thread one as the default for the single GPU case.

from neural_kernels_code.

Noted! What was the exception, that should not happen. I can try to fix it.

from neural_kernels_code.