leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 27 Packages Projects Releases Wiki Activity

v4.3 update. (#2709)

Browse Source 
* v4.3 update.

* Update the cute_dsl_api changelog's doc link

* Update version to 4.3.0

* Update the example link

* Update doc to encourage user to install DSL from requirements.txt

---------

Co-authored-by: Larry Wu <larwu@nvidia.com>
This commit is contained in:
Junkai-Wu
2025-10-21 14:26:30 -04:00
committed by GitHub
co-authored by Larry Wu
parent e6e2cc29f5
commit b1d6e2c9b3
244 changed files with 59272 additions and 10455 deletions
Download Patch File Download Diff File Expand all files Collapse all files
examples/python/CuTeDSL/ampere/call_bypass_dlpack.py
+6 -6
View File
@@ -89,7 +89,7 @@ def tensor_op_gemm_wrapper(
k: cutlass.Int32,
l: cutlass.Int32,
):
print(f"\n[DSL INFO] Input Parameters:")
print("\n[DSL INFO] Input Parameters:")
print(f"[DSL INFO] mnkl: {(m, n, k, l)}")
# Assume alignment of shape to call tensorop_gemm example
@@ -111,7 +111,7 @@ def tensor_op_gemm_wrapper(
tensor_op_gemm = TensorOpGemm(
a_ptr.value_type, c_ptr.value_type, cutlass.Float32, (2, 2, 1)
)
print(f"\n[DSL INFO] Created TensorOpGemm instance")
print("\n[DSL INFO] Created TensorOpGemm instance")
print(f"[DSL INFO] Input dtype: {a_ptr.value_type}")
print(f"[DSL INFO] Output dtype: {c_ptr.value_type}")
print(f"[DSL INFO] Accumulation dtype: {cutlass.Float32}")
@@ -119,11 +119,11 @@ def tensor_op_gemm_wrapper(
# No need to compile inside jit function
tensor_op_gemm(mA, mB, mC)
print(f"\n[DSL INFO] Executed TensorOpGemm")
print("\n[DSL INFO] Executed TensorOpGemm")
def run_tensor_op_gemm_wrapper(mnkl: Tuple[int, int, int, int]):
print(f"\nRunning TensorOpGemm test with:")
print("\nRunning TensorOpGemm test with:")
print(f"Tensor dimensions: {mnkl}")
# (M,K,L)
@@ -139,7 +139,7 @@ def run_tensor_op_gemm_wrapper(mnkl: Tuple[int, int, int, int]):
mnkl[3], mnkl[0], mnkl[1], dtype=torch.float16, device="cuda"
).permute(1, 2, 0)
print(f"Input tensor shapes:")
print("Input tensor shapes:")
print(f"a: {a.shape}, dtype: {a.dtype}")
print(f"b: {b.shape}, dtype: {b.dtype}")
print(f"c: {c.shape}, dtype: {c.dtype}\n")
@@ -158,7 +158,7 @@ def run_tensor_op_gemm_wrapper(mnkl: Tuple[int, int, int, int]):
ref = torch.einsum("mkl,nkl->mnl", a, b)
torch.testing.assert_close(c, ref, atol=1e-05, rtol=1e-05)
print(f"\n[DSL INFO] Results verified successfully!")
print("\n[DSL INFO] Results verified successfully!")
print(f"First few elements of result: \n{c[:3, :3, :3]}")