frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-11-04 16:04:58 +08:00
Code Issues Packages Projects Releases Wiki Activity

Specify width for st.floats in hypothesis_utils.tensor (#25188)

Browse Source 
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/25188

circleci complains about generated numbers are not representable by float32 and it pollutes the logs:
https://circleci.com/gh/pytorch/pytorch/2554740?utm_campaign=vcs-integration-link&utm_medium=referral&utm_source=github-build-link

Test Plan:
circleci

Imported from OSS

Differential Revision: D17063240

fbshipit-source-id: 0572fb810d8ccd8cdf3f3ac7efdf0cfce5aee6ca
This commit is contained in:
Jerry Zhang
2019-08-27 10:19:28 -07:00
committed by Facebook Github Bot
parent 44a7879b6e
commit c351a68f5b
1 changed files with 4 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
test/hypothesis_utils.py
View File

@ -42,7 +42,7 @@ def _get_valid_min_max(qparams):
# make sure intermediate results are within the range of long
min_value = max((long_min - zero_point) * scale, (long_min / scale + zero_point))
max_value = min((long_max - zero_point) * scale, (long_max / scale + zero_point))
return min_value, max_value
return np.float32(min_value), np.float32(max_value)
"""Hypothesis filter to avoid overflows with quantized tensors.
@ -106,7 +106,7 @@ def qparams(draw, dtypes=None, scale_min=None, scale_max=None,
scale_min = torch.finfo(torch.float).eps
if scale_max is None:
scale_max = torch.finfo(torch.float).max
scale = draw(st.floats(min_value=scale_min, max_value=scale_max))
scale = draw(st.floats(min_value=scale_min, max_value=scale_max, width=32))
return scale, zero_point, quantized_type
@ -162,7 +162,7 @@ def tensor(draw, shapes=None, elements=None, qparams=None):
_shape = draw(st.sampled_from(shapes))
if qparams is None:
if elements is None:
elements = st.floats(-1e6, 1e6, allow_nan=False)
elements = st.floats(-1e6, 1e6, allow_nan=False, width=32)
X = draw(stnp.arrays(dtype=np.float32, elements=elements, shape=_shape))
assume(not (np.isnan(X).any() or np.isinf(X).any()))
return X, None
@ -170,7 +170,7 @@ def tensor(draw, shapes=None, elements=None, qparams=None):
if elements is None:
min_value, max_value = _get_valid_min_max(qparams)
elements = st.floats(min_value, max_value, allow_infinity=False,
allow_nan=False)
allow_nan=False, width=32)
X = draw(stnp.arrays(dtype=np.float32, elements=elements, shape=_shape))
# Recompute the scale and zero_points according to the X statistics.
scale, zp = _calculate_dynamic_qparams(X, qparams[2])