Jerry Zhang 7ddf212f33 [quant][fx] Fully align convert with the reference model design and simplify the implementation (#73863) ... Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/73863 This PR fully aligns the convert function with the design: https://github.com/pytorch/rfcs/blob/master/RFC-0019-Extending-PyTorch-Quantization-to-Custom-Backends.md and simplifies the implementation of convert function by always produce a reference quantized model (with reference patterns) first, and then lower the model to a quantized model that is runnable with PyTorch native backend (fbgemm/qnnpack). This PR makes the convert.py much easier to understand than the previous implementation, and we are able to remove majority of code in quantization_patterns.py as well (in followup PRs). Test Plan: ``` python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps python test/test_quantization.py TestFXNumericSuiteCoreAPIs python test/test_quantization.py TestFXNumericSuiteCoreAPIsModels ``` and other internal/oss regression tests Imported from OSS Reviewed By: andrewor14 Differential Revision: D34778506 fbshipit-source-id: 0678b66addf736039a8749b352f6f569caca962b (cherry picked from commit 33ec9caf23f3ab373d827117efbd9db0668b2437)		2022-03-11 17:11:30 +00:00
..
_dbr	dbr quant: enable reference module support for torch.qint32 (#73493)	2022-03-04 17:35:31 +00:00
fx	[quant][fx] Fully align convert with the reference model design and simplify the implementation (#73863)	2022-03-11 17:11:30 +00:00
__init__.py	[reland][bc-breaking][quant][be] Refactor fuser_method to include is_qat argument" (#71956)	2022-01-31 23:02:22 +00:00
_correct_bias.py	[quant] Fix the parts that were missing after initial migration (#66058)	2021-10-05 11:45:37 -07:00
_equalize.py	[quant] AO migration of the _correct_bias.py, _equalize.py, and _learnable_fake_quantize.py (#64917)	2021-09-15 18:15:39 -07:00
_learnable_fake_quantize.py	[quant] Fix the parts that were missing after initial migration (#66058)	2021-10-05 11:45:37 -07:00
_quantize_dbr.py	dbr quant: insert activation obs explicitly, instead of relying on hooks (#73492)	2022-03-04 17:35:31 +00:00
_quantize_fx_do_not_use.py	[quant][fx] Fully align convert with the reference model design and simplify the implementation (#73863)	2022-03-11 17:11:30 +00:00
fake_quantize.py	[ao] Removing memoryless observer args for MovingAverage (#73947)	2022-03-11 00:21:49 +00:00
fuse_modules.py	[reland][bc-breaking][quant][be] Refactor fuser_method to include is_qat argument" (#71956)	2022-01-31 23:02:22 +00:00
fuser_method_mappings.py	[qunat][fx][fix] Fix get_module_type for fusion (#72735)	2022-02-25 18:37:31 +00:00
observer.py	[ao] Removing memoryless observer args for MovingAverage (#73947)	2022-03-11 00:21:49 +00:00
pattern.md	[quant][refactor] Move pattern type definition to ao/quantization/utils.py (#68769)	2021-12-07 11:00:22 -08:00
qconfig_dict_utils.py	fx quant: move _parent_name to common utils (#69720)	2021-12-17 05:59:46 -08:00
qconfig.py	[quant][fx] Fully align convert with the reference model design and simplify the implementation (#73863)	2022-03-11 17:11:30 +00:00
quant_type.py	[quant] AO migration of the quant_types.py (phase 1) (#64916)	2021-09-15 17:30:00 -07:00
quantization_mappings.py	[quant][fx] Fully align convert with the reference model design and simplify the implementation (#73863)	2022-03-11 17:11:30 +00:00
quantize_fx.py	[quant][fx] Fully align convert with the reference model design and simplify the implementation (#73863)	2022-03-11 17:11:30 +00:00
quantize_jit.py	[quant] Fix the parts that were missing after initial migration (#66058)	2021-10-05 11:45:37 -07:00
quantize.py	[Qunat] Refactor reference module mapping (#72755)	2022-03-08 06:48:04 +00:00
stubs.py	quantization: fix bug in QuantWrapper with DeQuant qconfig (#73671)	2022-03-03 15:31:53 +00:00
utils.py	[quant][fx] Fully align convert with the reference model design and simplify the implementation (#73863)	2022-03-11 17:11:30 +00:00