[Quant][fx] Add default configs for fixed qparams ops #80184
Closed
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Summary: This commit adds qconfigs with special observers for fixed qparams ops in get_default_qconfig_mapping and get_default_qat_qconfig_mapping. For correctness, we also require users to use these special observers if we detect these fixed qparams ops in prepare. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo [ghstack-poisoned]
🔗 Helpful links
✅ No Failures (0 Pending)As of commit e7b651e (more details on the Dr. CI page): Expand to see more💚 💚 Looks good so far! There are no failures yet. 💚 💚 This comment was automatically generated by Dr. CI (expand for details).Please report bugs/suggestions to the (internal) Dr. CI Users group. |
Summary: This commit adds qconfigs with special observers for fixed qparams ops in get_default_qconfig_mapping and get_default_qat_qconfig_mapping. For correctness, we also require users to use these special observers if we detect these fixed qparams ops in prepare. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo [ghstack-poisoned]
Summary: This commit adds qconfigs with special observers for fixed qparams ops in get_default_qconfig_mapping and get_default_qat_qconfig_mapping. For correctness, we also require users to use these special observers if we detect these fixed qparams ops in prepare. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo [ghstack-poisoned]
andrewor14
added a commit
that referenced
this pull request
Jun 23, 2022
Summary: This commit adds qconfigs with special observers for fixed qparams ops in get_default_qconfig_mapping and get_default_qat_qconfig_mapping. For correctness, we also require users to use these special observers if we detect these fixed qparams ops in prepare. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo ghstack-source-id: cbb3fd2 Pull Request resolved: #80184
|
@andrewor14 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
Summary: This commit adds qconfigs with special observers for fixed qparams ops in get_default_qconfig_mapping and get_default_qat_qconfig_mapping. For correctness, we also require users to use these special observers if we detect these fixed qparams ops in prepare. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo Differential Revision: [D37396379](https://our.internmc.facebook.com/intern/diff/D37396379) [ghstack-poisoned]
andrewor14
added a commit
that referenced
this pull request
Jun 24, 2022
Summary: This commit adds qconfigs with special observers for fixed qparams ops in get_default_qconfig_mapping and get_default_qat_qconfig_mapping. For correctness, we also require users to use these special observers if we detect these fixed qparams ops in prepare. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo ghstack-source-id: 2588bf5 Pull Request resolved: #80184
|
@andrewor14 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
jerryzh168
reviewed
Jun 24, 2022
jerryzh168
reviewed
Jun 24, 2022
jerryzh168
reviewed
Jun 24, 2022
jerryzh168
approved these changes
Jun 24, 2022
Summary: This commit adds qconfigs with special observers for fixed qparams ops in get_default_qconfig_mapping and get_default_qat_qconfig_mapping. For correctness, we also require users to use these special observers if we detect these fixed qparams ops in prepare. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo Differential Revision: [D37396379](https://our.internmc.facebook.com/intern/diff/D37396379) [ghstack-poisoned]
Summary: This commit adds qconfigs with special observers for fixed qparams ops in get_default_qconfig_mapping and get_default_qat_qconfig_mapping. For correctness, we also require users to use these special observers if we detect these fixed qparams ops in prepare. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo Differential Revision: [D37396379](https://our.internmc.facebook.com/intern/diff/D37396379) [ghstack-poisoned]
andrewor14
added a commit
that referenced
this pull request
Jun 28, 2022
Summary: This commit adds qconfigs with special observers for fixed qparams ops in get_default_qconfig_mapping and get_default_qat_qconfig_mapping. For correctness, we also require users to use these special observers if we detect these fixed qparams ops in prepare. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo ghstack-source-id: 2e830b6 Pull Request resolved: #80184
|
@andrewor14 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
Summary: This commit adds qconfigs with special observers for fixed qparams ops in get_default_qconfig_mapping and get_default_qat_qconfig_mapping. For correctness, we also require users to use these special observers if we detect these fixed qparams ops in prepare. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo Differential Revision: [D37396379](https://our.internmc.facebook.com/intern/diff/D37396379) [ghstack-poisoned]
andrewor14
added a commit
that referenced
this pull request
Jun 29, 2022
Summary: This commit adds qconfigs with special observers for fixed qparams ops in get_default_qconfig_mapping and get_default_qat_qconfig_mapping. For correctness, we also require users to use these special observers if we detect these fixed qparams ops in prepare. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo ghstack-source-id: a16d8cf Pull Request resolved: #80184
|
@andrewor14 has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator. |
pytorchmergebot
pushed a commit
that referenced
this pull request
Jul 14, 2022
…ops (#81010) Summary: This adds a few things on top of #80184, 1). node.target was assumed to be "tanh", torch.nn.Tanh etc. this PR handles that properly 2). adds FixedQParamsFakeQuantize support 3). extends the comparison function _partial_wrapper_equals to work with FakeQuantize.with_args(observer=...) Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: Subscribers: Tasks: Tags: Differential Revision: [D37735193](https://our.internmc.facebook.com/intern/diff/D37735193) Pull Request resolved: #81010 Approved by: https://github.com/andrewor14
facebook-github-bot
pushed a commit
that referenced
this pull request
Jul 14, 2022
…ops (#81010) Summary: Pull Request resolved: #81010 This adds a few things on top of #80184, 1). node.target was assumed to be "tanh", torch.nn.Tanh etc. this PR handles that properly 2). adds FixedQParamsFakeQuantize support 3). extends the comparison function _partial_wrapper_equals to work with FakeQuantize.with_args(observer=...) Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Imported from OSS **Static Docs Preview: classyvision** |[Full Site](https://our.intern.facebook.com/intern/staticdocs/eph/D37735193/V9/classyvision/)| |**Modified Pages**| Reviewed By: cpuhrsch Differential Revision: D37735193 fbshipit-source-id: 337e42dd9001ed12f1f901764a0f5644e75dfc3c
This was referenced Nov 7, 2022
andrewor14
added a commit
that referenced
this pull request
Nov 7, 2022
Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should only specify requirements on observers, but not the observer constructors themselves. Later, #80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers specified by the user. This commit removes these observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. To be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo [ghstack-poisoned]
andrewor14
added a commit
that referenced
this pull request
Nov 8, 2022
…ut_observer" Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should only specify requirements on observers, but not the observer constructors themselves. Later, #80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers specified by the user. This commit removes these observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. To be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo cc jerryzh168 jianyuh raghuramank100 jamesr66a vkuzo jgong5 Xia-Weiwen leslie-fang-intel [ghstack-poisoned]
andrewor14
added a commit
that referenced
this pull request
Nov 8, 2022
Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should only specify requirements on observers, but not the observer constructors themselves. Later, #80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers specified by the user. This commit removes these observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. To be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo cc jerryzh168 jianyuh raghuramank100 jamesr66a vkuzo jgong5 Xia-Weiwen leslie-fang-intel [ghstack-poisoned]
andrewor14
added a commit
that referenced
this pull request
Nov 8, 2022
Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should only specify requirements on observers, but not the observer constructors themselves. Later, #80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers specified by the user. This commit removes these observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. To be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo ghstack-source-id: fb72078 Pull Request resolved: #88620
andrewor14
added a commit
that referenced
this pull request
Nov 8, 2022
…ut_observer" Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should not set the observer constructors themselves, but should instead specify only requirements on the observers. Later, #80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers from the BackendConfig, since we expect the user to pass in the correct observers for these ops. This commit removes these overwrite observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. Note that, however, to be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo [ghstack-poisoned]
andrewor14
added a commit
that referenced
this pull request
Nov 8, 2022
Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should not set the observer constructors themselves, but should instead specify only requirements on the observers. Later, #80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers from the BackendConfig, since we expect the user to pass in the correct observers for these ops. This commit removes these overwrite observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. Note that, however, to be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo [ghstack-poisoned]
andrewor14
added a commit
that referenced
this pull request
Nov 8, 2022
Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should not set the observer constructors themselves, but should instead specify only requirements on the observers. Later, #80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers from the BackendConfig, since we expect the user to pass in the correct observers for these ops. This commit removes these overwrite observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. Note that, however, to be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo ghstack-source-id: fb72078 Pull Request resolved: #88620
andrewor14
added a commit
that referenced
this pull request
Nov 14, 2022
…ut_observer" Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should not set the observer constructors themselves, but should instead specify only requirements on the observers. Later, #80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers from the BackendConfig, since we expect the user to pass in the correct observers for these ops. This commit removes these overwrite observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. Note that, however, to be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo [ghstack-poisoned]
andrewor14
added a commit
that referenced
this pull request
Nov 14, 2022
Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should not set the observer constructors themselves, but should instead specify only requirements on the observers. Later, #80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers from the BackendConfig, since we expect the user to pass in the correct observers for these ops. This commit removes these overwrite observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. Note that, however, to be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo [ghstack-poisoned]
andrewor14
added a commit
that referenced
this pull request
Nov 14, 2022
…ut_observer" Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should not set the observer constructors themselves, but should instead specify only requirements on the observers. Later, #80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers from the BackendConfig, since we expect the user to pass in the correct observers for these ops. This commit removes these overwrite observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. Note that, however, to be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo [ghstack-poisoned]
andrewor14
added a commit
that referenced
this pull request
Nov 14, 2022
Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should not set the observer constructors themselves, but should instead specify only requirements on the observers. Later, #80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers from the BackendConfig, since we expect the user to pass in the correct observers for these ops. This commit removes these overwrite observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. Note that, however, to be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo [ghstack-poisoned]
andrewor14
added a commit
that referenced
this pull request
Nov 14, 2022
…ut_observer" Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should not set the observer constructors themselves, but should instead specify only requirements on the observers. Later, #80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers from the BackendConfig, since we expect the user to pass in the correct observers for these ops. This commit removes these overwrite observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. Note that, however, to be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo [ghstack-poisoned]
andrewor14
added a commit
that referenced
this pull request
Nov 14, 2022
Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should not set the observer constructors themselves, but should instead specify only requirements on the observers. Later, #80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers from the BackendConfig, since we expect the user to pass in the correct observers for these ops. This commit removes these overwrite observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. Note that, however, to be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo [ghstack-poisoned]
andrewor14
added a commit
that referenced
this pull request
Nov 14, 2022
Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should not set the observer constructors themselves, but should instead specify only requirements on the observers. Later, #80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers from the BackendConfig, since we expect the user to pass in the correct observers for these ops. This commit removes these overwrite observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. Note that, however, to be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo ghstack-source-id: bf01ac4 Pull Request resolved: #88620
andrewor14
added a commit
that referenced
this pull request
Nov 15, 2022
…ut_observer" Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should not set the observer constructors themselves, but should instead specify only requirements on the observers. Later, #80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers from the BackendConfig, since we expect the user to pass in the correct observers for these ops. This commit removes these overwrite observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. Note that, however, to be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo [ghstack-poisoned]
andrewor14
added a commit
that referenced
this pull request
Nov 15, 2022
Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should not set the observer constructors themselves, but should instead specify only requirements on the observers. Later, #80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers from the BackendConfig, since we expect the user to pass in the correct observers for these ops. This commit removes these overwrite observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. Note that, however, to be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo [ghstack-poisoned]
andrewor14
added a commit
that referenced
this pull request
Nov 15, 2022
Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should not set the observer constructors themselves, but should instead specify only requirements on the observers. Later, #80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers from the BackendConfig, since we expect the user to pass in the correct observers for these ops. This commit removes these overwrite observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. Note that, however, to be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo ghstack-source-id: 5599150 Pull Request resolved: #88620
andrewor14
added a commit
that referenced
this pull request
Nov 15, 2022
…ut_observer" Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should not set the observer constructors themselves, but should instead specify only requirements on the observers. Later, #80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers from the BackendConfig, since we expect the user to pass in the correct observers for these ops. This commit removes these overwrite observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. Note that, however, to be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo [ghstack-poisoned]
andrewor14
added a commit
that referenced
this pull request
Nov 15, 2022
Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should not set the observer constructors themselves, but should instead specify only requirements on the observers. Later, #80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers from the BackendConfig, since we expect the user to pass in the correct observers for these ops. This commit removes these overwrite observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. Note that, however, to be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo [ghstack-poisoned]
andrewor14
added a commit
that referenced
this pull request
Nov 15, 2022
Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should not set the observer constructors themselves, but should instead specify only requirements on the observers. Later, #80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers from the BackendConfig, since we expect the user to pass in the correct observers for these ops. This commit removes these overwrite observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. Note that, however, to be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo ghstack-source-id: 5ef5274 Pull Request resolved: #88620
pytorchmergebot
pushed a commit
that referenced
this pull request
Nov 16, 2022
Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should not set the observer constructors themselves, but should instead specify only requirements on the observers. Later, #80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers from the BackendConfig, since we expect the user to pass in the correct observers for these ops. This commit removes these overwrite observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. Note that, however, to be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo Pull Request resolved: #88620 Approved by: https://github.com/jerryzh168
kulinseth
pushed a commit
to kulinseth/pytorch
that referenced
this pull request
Dec 10, 2022
Summary: When the BackendConfig was first introduced, `overwrite_output_observer` and `overwrite_output_fake_quantize` were added to ensure fixed qparams ops like `torch.nn.Sigmoid` and `torch.nn.Tanh` used the correct observers and fake quantizes. However, this is hacky because the BackendConfig should not set the observer constructors themselves, but should instead specify only requirements on the observers. Later, pytorch#80184 added the correct observers to `get_default_qconfig_mapping` along with validation logic that throws an error if incorrect observers were specified. With this change, we no longer need to overwrite the observers from the BackendConfig, since we expect the user to pass in the correct observers for these ops. This commit removes these overwrite observer settings in the BackendConfig. Instead, we represent the observer constraints for fixed qparams ops through the existing DTypeWithConstraints mechanism. Note that, however, to be consistent with other DTypeWithConstraints checks, we no longer throw an error if an incorrect observer is specified, but simply ignore the offending QConfig and log a warning instead. This is the BC-breaking part of the change. BC-breaking notes: ``` from torch.ao.quantization.qconfig import default_qconfig from torch.ao.quantization.quantize_fx import prepare_fx model = ModelWithFixedQParamsOps() qconfig_mapping = QConfigMapping().set_global(default_qconfig) example_inputs = ... prepare_fx(model, qconfig_mapping, example_inputs) ``` Before this commit, running the above leads to an exception because the wrong observers are used for fixed qparams ops. After this commit, the above will only encounter a warning, and the fixed qparams ops will not be quantized. In both cases, switching to `get_default_qconfig_mapping` will cause the fixed qparams ops to be quantized. Test Plan: python test/test_quantization.py TestQuantizeFx python test/test_quantization.py TestQuantizeFxOps Reviewers: jerryzh168, vkuzo Subscribers: jerryzh168, vkuzo Pull Request resolved: pytorch#88620 Approved by: https://github.com/jerryzh168
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Stack from ghstack (oldest at bottom):
Summary: This commit adds qconfigs with special observers for fixed
qparams ops in get_default_qconfig_mapping and
get_default_qat_qconfig_mapping. For correctness, we also require
users to use these special observers if we detect these fixed
qparams ops in prepare.
BC-breaking notes:
Before (fails after this PR):
After:
Test Plan:
python test/test_quantization.py TestQuantizeFx
python test/test_quantization.py TestQuantizeFxOps
Reviewers: jerryzh168, vkuzo
Subscribers: jerryzh168, vkuzo
Differential Revision: D37396379