Some BUG about CUDA KERNEL

MatthieuCourbariaux / BinaryNet

Training Deep Neural Networks with Weights and Activations Constrained to +1 or -1

BSD 3-Clause "New" or "Revised" License

1.04k stars 346 forks source link

Closed tairenpiao closed 6 months ago

tairenpiao commented 4 years ago

If set matrix dimension N smaller than 512(e.g. 256, 128) when running benchmark, the result of the xnor_gemm has error.

MatthieuCourbariaux commented 6 months ago

Yeah, it was really just a proof of concept cuda kernel. Not really optimized or useful for production.