[PyTorch] Utilize preswizzled inputs in GroupedGemm #2464
Description
Enable and optimize the use of preswizzled input tensors in GroupedGemm operations in PyTorch, reducing input transform overhead and maximizing device throughput.
{{ message }}
Enable and optimize the use of preswizzled input tensors in GroupedGemm operations in PyTorch, reducing input transform overhead and maximizing device throughput.