MCPcopy
hub / github.com/deepspeedai/DeepSpeed / skip_on_arch

Function skip_on_arch

tests/unit/util.py:18–24  ·  view source on GitHub ↗
(min_arch=7)

Source from the content-addressed store, hash-verified

16
17
18def skip_on_arch(min_arch=7):
19 if get_accelerator().device_name() == 'cuda':
20 if torch.cuda.get_device_capability()[0] < min_arch: #ignore-cuda
21 pytest.skip(f"needs higher compute capability than {min_arch}")
22 else:
23 assert is_current_accelerator_supported()
24 return
25
26
27def skip_on_cuda(valid_cuda):

Callers 15

test_softmaxFunction · 0.90
test_matmulFunction · 0.90
testMethod · 0.90
test_compile_zeroMethod · 0.90
testMethod · 0.90

Calls 3

get_acceleratorFunction · 0.90
device_nameMethod · 0.45

Tested by 15

test_softmaxFunction · 0.72
test_matmulFunction · 0.72
testMethod · 0.72
test_compile_zeroMethod · 0.72
testMethod · 0.72

Used in the wild real call sites across dependent graphs

searching dependent graphs…