[torchlib] Fix and improve quantization support #7581
GitHub Actions / Test Results
failed
Jul 22, 2024 in 0s
1 fail, 3 134 skipped, 9 806 pass in 2h 7m 20s
Annotations
Check warning on line 0 in onnxscript.ir.serde_test.TensorProtoTensorTest
github-actions / Test Results
All 9 runs failed: test_tensor_proto_tensor_bfloat16 (onnxscript.ir.serde_test.TensorProtoTensorTest)
artifacts/Test Results (py311-onnx-weekly-macos-latest)/pytest.xml [took 0s]
artifacts/Test Results (py311-onnx-weekly-ubuntu-latest)/pytest.xml [took 0s]
artifacts/Test Results (py311-onnx-weekly-windows-latest)/pytest.xml [took 0s]
artifacts/Test Results (py311-torch-nightly-macos-latest)/pytest.xml [took 0s]
artifacts/Test Results (py311-torch-nightly-ubuntu-latest)/pytest.xml [took 0s]
artifacts/Test Results (py311-torch-nightly-windows-latest)/pytest.xml [took 0s]
artifacts/Test Results (py312-torch-nightly-macos-latest)/pytest.xml [took 0s]
artifacts/Test Results (py312-torch-nightly-ubuntu-latest)/pytest.xml [took 0s]
artifacts/Test Results (py312-torch-nightly-windows-latest)/pytest.xml [took 0s]
Raw output
AssertionError:
Arrays are not equal
Mismatched elements: 8 / 9 (88.9%)
Max absolute difference: 49219.
Max relative difference: 97793.
x: array([[49216, 49024, 48896, 32768, 0, 16128, 16256, 16936, 16384]],
dtype=(numpy.uint16, [('bfloat16', '<u2')]))
y: array([[-3, -1, -0.5, -0, 0, 0.5, 1, 42, 2]], dtype=bfloat16)
onnxscript/ir/serde_test.py:106: in test_tensor_proto_tensor_bfloat16
np.testing.assert_array_equal(array_from_raw_data, expected_array)
/Library/Frameworks/Python.framework/Versions/3.12/lib/python3.12/contextlib.py:81: in inner
return func(*args, **kwds)
E AssertionError:
E Arrays are not equal
E
E Mismatched elements: 8 / 9 (88.9%)
E Max absolute difference: 49219.
E Max relative difference: 97793.
E x: array([[49216, 49024, 48896, 32768, 0, 16128, 16256, 16936, 16384]],
E dtype=(numpy.uint16, [('bfloat16', '<u2')]))
E y: array([[-3, -1, -0.5, -0, 0, 0.5, 1, 42, 2]], dtype=bfloat16)
Loading