We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
ops/tilelang/torch.py
1 parent 8fcb667 commit d9b4a82Copy full SHA for d9b4a82
1 file changed
ops/tilelang/torch.py
@@ -0,0 +1,12 @@
1
+import torch
2
+
3
+import ops.tilelang.kernels.mm
4
5
6
+def mm(input, other):
7
+ output_shape = (input.shape[0], other.shape[1])
8
+ output = torch.empty(output_shape, dtype=input.dtype, device=input.device)
9
10
+ ops.tilelang.kernels.mm.kernel(input, other, output)
11
12
+ return output
0 commit comments