leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 28 Packages Projects Releases Wiki Activity

CUTLASS 3.4.0 (#1286)

Browse Source 
* CUTLASS 3.4.0

* Update CHANGELOG.md

---------

Co-authored-by: Pradeep Ramani <prramani@nvidia.com>
This commit is contained in:
Pradeep Ramani
2023-12-29 15:21:31 -05:00
committed by GitHub
co-authored by Pradeep Ramani
parent b7508e3379
commit 8236f30675
211 changed files with 11409 additions and 2763 deletions
Download Patch File Download Diff File Expand all files Collapse all files
test/python/cutlass/gemm/gemm_f64_sm90.py
+4 -2
View File
@@ -46,8 +46,11 @@ from utils import LayoutCombination, add_test_gemm
cutlass.set_log_level(logging.WARNING)
cc = 90
dtype = cutlass.DataType.f64
@unittest.skipIf(device_cc() < cc, 'Device compute capability is insufficient for SM90 tests.')
@unittest.skipIf(cutlass.utils.datatypes.torch_type(dtype) is None, f'Version of torch installed does not contain a datatype match for {dtype}')
class GemmF64Sm90(unittest.TestCase):
"""
Wrapper class to which tests will be added dynamically in __main__
@@ -56,8 +59,7 @@ class GemmF64Sm90(unittest.TestCase):
add_test_specialized = partial(add_test_gemm, cls=GemmF64Sm90, alignments=[1, 1, 1], cluster_shape=[1, 1, 1],
element=cutlass.DataType.f64, element_output=cutlass.DataType.f64,
element_accumulator=cutlass.DataType.f64, compilation_modes=['nvcc'])
element=dtype, element_output=dtype, element_accumulator=dtype, compilation_modes=['nvcc'])
add_test_specialized(opclass=cutlass.OpcodeClass.TensorOp, layouts=LayoutCombination.NNT, threadblock_shape=[128, 128, 32], stages=3)
add_test_specialized(opclass=cutlass.OpcodeClass.TensorOp, layouts=LayoutCombination.TNN, threadblock_shape=[128, 128, 32], stages=3)