Vasiliy Kuznetsov c193bd41f5 fake_quantize: respect device affinity (#39031) ... Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/39031 Makes the eager mode QAT prepare logic respect device affinity. This fixes the issue where a module is on `cuda:0`, and running the QAT prepare script would add observers on `cpu`. Now it will add them on the original device. Test Plan: ``` python test/test_quantization.py TestDistributed.test_device_affinity ``` Imported from OSS Differential Revision: D21729272 fbshipit-source-id: 5537bf3977ddc23412184941978bf0d1cc6fb479		2020-06-01 08:55:14 -07:00
..
__init__.py	[quant][graphmode] Rename _quantize_script.py to quantize_script.py (#39122)	2020-05-29 12:33:40 -07:00
_numeric_suite.py	[PyTorch Numeric Suite] Add module output comparison (#36701)	2020-05-03 00:04:35 -07:00
default_mappings.py	[quant] Add support for Quantized Conv1d and ConvRELU1d (#38283)	2020-05-13 16:59:13 -07:00
fake_quantize.py	fake_quant: move observer and fake_quant flags into buffers (#38368)	2020-05-18 09:30:07 -07:00
fuse_modules.py	[quant] Support for fused ConvBn1d and ConvBnRelu1d modules (#38452) (#38749)	2020-05-19 22:48:05 -07:00
observer.py	fake_quant: make qparams shape consistent (#38587)	2020-05-21 19:08:08 -07:00
qconfig.py	[quant] Return default qconfig when backend is 'none' (#38407)	2020-05-14 09:53:50 -07:00
quantize_script.py	[quant][graphmode] Rename _quantize_script.py to quantize_script.py (#39122)	2020-05-29 12:33:40 -07:00
quantize.py	fake_quantize: respect device affinity (#39031)	2020-06-01 08:55:14 -07:00
stubs.py	Factored out the default mappings	2019-10-03 11:52:21 -07:00