mirror of
https://github.com/tile-ai/tilelang.git
synced 2026-10-22 06:50:25 +08:00
45 lines
1.3 KiB
Python
45 lines
1.3 KiB
Python
import tilelang
|
|
import tilelang.language as T
|
|
import torch
|
|
import tilelang.testing
|
|
|
|
|
|
@tilelang.jit(out_idx=[1],)
|
|
def tilelang_ternary(M, N, block_M, block_N, dtype="float16"):
|
|
|
|
@T.prim_func
|
|
def main(
|
|
A: T.Tensor((M, N), dtype),
|
|
B: T.Tensor((M, N), dtype),
|
|
):
|
|
# Initialize Kernel Context
|
|
with T.Kernel(T.ceildiv(N, block_N), T.ceildiv(M, block_M), threads=128) as (bx, by):
|
|
for i, j in T.Parallel(block_M, block_N):
|
|
B[by * block_M + i, bx * block_N + j] = (
|
|
A[by * block_M + i, bx * block_N + j] if (by * block_M + i) < (M // 2) else 0)
|
|
|
|
return main
|
|
|
|
|
|
def run_tilelang_ternary(M=128, N=128, block_M=32, block_N=32, dtype="float16"):
|
|
kernel = tilelang_ternary(M, N, block_M, block_N, dtype)
|
|
a = torch.randn(M, N, device="cuda", dtype=getattr(torch, dtype))
|
|
b = kernel(a)
|
|
ref_b = torch.zeros_like(b)
|
|
for i in range(M):
|
|
for j in range(N):
|
|
if i < M // 2:
|
|
ref_b[i, j] = a[i, j]
|
|
else:
|
|
ref_b[i, j] = 0
|
|
|
|
torch.testing.assert_close(b, ref_b, rtol=1e-2, atol=1e-2)
|
|
|
|
|
|
def test_tilelang_ternary():
|
|
run_tilelang_ternary(M=128, N=128, block_M=32, block_N=32)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
tilelang.testing.main()
|