Open sminy67 opened 1 year ago
There is Tensor core in NVIDIA GPUs but as the architectures get different, so as the number of tensor cores. Then how do we fully utilize tensor core when coding CUDA? I'm aiming to use tensor cores with quantization and sparse operations.
There is Tensor core in NVIDIA GPUs but as the architectures get different, so as the number of tensor cores. Then how do we fully utilize tensor core when coding CUDA? I'm aiming to use tensor cores with quantization and sparse operations.