Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/26666
Changes:
- Introduce a `ConcreteModuleType` concept. This acts both as the key into the type
cache, and as the source of truth for `ModuleValue::attr` queries. It needs
to do both jobs because that's how we ensure correctness (if the types are
different, it's because `ModuleValue::attr` would return different things).
- Now `recursive_script` will first construct a `ConcreteModuleType` and search for a
pre-existing type before starting compilation.
- All previous paths to creating a `ScriptModule` (including inheriting from
`ScriptModule`) are now rewritten to go through `create_script_module`, so
that we have only a single place where construction happens.
Behavioral changes:
- Big change to `torch.jit.ScriptModule` inheritance: all attributes are now
recursively scripted if possible, matching recursive scripting semantics.
This makes it hard to keep something from being scripted (for example, a
Python submodule). Possibly we'll need an `ignore()` type thing for
attributes. In particular, this adds `self.training` to *every* ScriptModule, since
it's present on every `nn.Module`.
- I believe this change to be transparent to existing users of the inheritance API, since if you had an attribute that is unscriptable that you never used, there is no error. In some cases, we will create new attributes (even if they are unused), which will increase serialized model size from before.
Test Plan: Imported from OSS
Differential Revision: D17551196
Pulled By: suo
fbshipit-source-id: b476d1c9feb3ddfd63406d90989aaf9dfe890591
Summary:
Most of this was old cruft left over from special handling of `training` before we had a `bool` type. This makes all modules have a `training` attribute that is true by default and removes all other special handling.
Fixes#26884
](https://our.intern.facebook.com/intern/diff/17728129/)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/27109
Pulled By: driazati
Differential Revision: D17728129
fbshipit-source-id: 8ddc9fbb07a953dd05529538bfdd01ed88b5cb57
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/26457
Enhancement to fuse module to support sequentials, fuse list can now be just like the state dict.
Also add support for Conv-Relu and linear-relu fusion
Also support inplace and out of place fusion of models.
ghstack-source-id: 91076386
Test Plan:
buck test caffe2/test:quantization -- 'test_fusion_sequential_model_train \(test_quantization\.FusionTest\)' --print-passing-details
buck test caffe2/test:quantization -- 'test_fusion_sequential_model_eval \(test_quantization\.FusionTest\)' --print-passing-details
Differential Revision: D17466382
fbshipit-source-id: 0a548f8f4c366f3ecc59db693bac725ccd62328e
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/26516
ghstack-source-id: 90982010
Test Plan:
Integrate per-channel support into conv and linear modules.
The following tests pass:
buck test caffe2/test:quantized -- 'test_linear_api \(test_quantized_nn_mods\.ModuleAPITest\)' --print-passing-details
buck test caffe2/test:quantized -- 'test_conv_api \(test_quantized_nn_mods\.ModuleAPITest\)' --print-passing-details
buck test caffe2/test:quantized -- 'test_float_quant_compare_per_channel \(test_quantized_models\.ModelNumerics\)' --print-passing-details
Differential Revision: D17342622
fbshipit-source-id: f0d618928e3d9348672c589a6b7a47049c372a2e
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/26828
Pickle serialization for quantized modules is currently broken by https://github.com/pytorch/pytorch/issues/24045, so let's be loud and fail if the user tries to do it
Test Plan: Imported from OSS
Differential Revision: D17579127
Pulled By: jamesr66a
fbshipit-source-id: 3deccac7e4590c6f648f22bb79c57badf3bf0487
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/26574
Since we also have `quantized::linear`, `quantize_linear` sounds
confusing, so we plan to rename it before the branch cut
Test Plan:
ci
Imported from OSS
Differential Revision: D17514876
fbshipit-source-id: 01d9005e6ec8cb9950b9d8bba122109c389641d3
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/25428
Added bias as an optional param to the quantized_linear_prepack function.
Bias is quantized during runtime using input scale and weight scale.
ghstack-source-id: 89601399
Test Plan: python test/run_test.py --exclude nn --verbose --bring-to-front quantization quantized quantized_tensor quantized_nn_mods quantizer
Differential Revision: D17121304
fbshipit-source-id: 8adb0e55e4aed0a5430aaa2c8639c8ad1639c85a
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/25678
As an effort to unify fbgemm and qnnpack at the dispatcher level, we need to have a generic name for the quantized backed ops.
Currently FBGEMM is guarded by the USE_FBGEMM macro and QNNPACK uses USE_QNNPACK.
ghstack-source-id: 89518961
Test Plan: buck test caffe2/test:quantized
Differential Revision: D17194364
fbshipit-source-id: 5960aedff6b8cb89eb3872c39b74caf54c0fbf20
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/25338
As an effort to unify fbgemm and qnnpack at the dispatcher level, we need to have a generic name for the quantized backed ops.
Currently FBGEMM is guarded by the USE_FBGEMM macro and QNNPACK uses USE_QNNPACK.
TBD: Use compile time macro or run_time to switch between fbgemm and qnnpack.
ghstack-source-id: 89454244
Test Plan: buck test caffe2/test:quantized
Differential Revision: D17097735
fbshipit-source-id: 447112a7a421387724d3e29b8fd8412dfb1c373a
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/25212
In eager mode, all modules need to work with input tensors that can change qparams dynamically. This issue https://github.com/pytorch/pytorch/issues/23874 will address this via FBGEMM modifications. This is a work around before that.
ghstack-source-id: 89118038
Test Plan:
buck test caffe2/test:quantized -- 'test_conv_api \(test_quantized_nn_mods\.ModuleAPITest\)' --print-passing-details
Summary (total time 65.86s):
PASS: 1
FAIL: 0
SKIP: 0
FATAL: 0
TIMEOUT: 0
OMIT: 0
Differential Revision: D17064471
fbshipit-source-id: 3c192442b19bf2d9d88d4e52de6c24dc134a846f
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/24789
In eager mode, all modules need to work with input tensors that can change qparams dynamically. This issue https://github.com/pytorch/pytorch/issues/23874 will address this via FBGEMM modifications. This is a work around before that.
ghstack-source-id: 89003798
Test Plan:
buck test caffe2/test:quantized -- 'test_conv_api \(test_quantized_nn_mods\.ModuleAPITest\)' --print-passing-details
Summary (total time 65.86s):
PASS: 1
FAIL: 0
SKIP: 0
FATAL: 0
TIMEOUT: 0
OMIT: 0
Differential Revision: D16852280
fbshipit-source-id: 988f8ff91616eddf511e71926aa7d2d0f1938188
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/23753
Add intrinsic(fused) module mappings in quantize.py to enable mapping fused modules
in both QAT and post PTQ
Differential Revision: D16820749
fbshipit-source-id: 07de76a4f09b44bde8b193c103eac02c22b875b6
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/24048
Add `__{g,s}etstate__ methods on `nnq.Linear` for JIT (and torch.{save,load} serialization).
Unfortunately, this unearthed a bug in serialization documented in https://github.com/pytorch/pytorch/issues/24045. The check that triggered the bug has been disabled pending a fix
Test Plan: Imported from OSS
Reviewed By: driazati
Differential Revision: D16728347
Pulled By: jamesr66a
fbshipit-source-id: c3b850be3b831f4c77cec3c2df626151b2af8b34
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/24047
Add `_{save_to,load_from}_state_dict` methods to `nnq.Linear` that explicitly deal with conversions from the Python attributes to the serialized state dict form
Test Plan: Imported from OSS
Reviewed By: driazati
Differential Revision: D16728346
Pulled By: jamesr66a
fbshipit-source-id: 182c9f5069d509147dc9020b341b6cb87505fe7f
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/24046
`nnq.Linear` was a confusing mess of buffers/attributes and Tensor/not tensor members. This PR reworks it to consistently have only Python attributes, with the conversions handled explicitly by state_dict or __{get,set}state__ methods (added in PRs further up the stack
Test Plan: Imported from OSS
Reviewed By: driazati
Differential Revision: D16728345
Pulled By: jamesr66a
fbshipit-source-id: 47468b776b428fca2409bb55c8b161afb68a3379
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/23891
This adds an initial set of testing coverage for quantization that checks if the modules can be scripted. Testing for tracing and serialization is forthcoming
Test Plan: Imported from OSS
Differential Revision: D16698045
Pulled By: jamesr66a
fbshipit-source-id: 96d80d938b816220af72359165a7b96d998a30c9
Summary:
Add support for quantization aware training in eager mode
Modifications to Post training flow:
## Prepare
* Fusion: e.g. (Conv, Bn) → ConvBn (float)
* Swapping: To insert fake_quant to weight, we need to swap the float modules that has weight with different qat modules, e.g. Conv → torch.nn.qat.Conv , ConvBn → torch.nn._intrinsic.qat.ConvBn
```
* previously we were thinking about modify the weight in forward_pre hook and change it back in forward_hook:
* def forward_pre_hook(self, input):
self.float_weight = self.weight
self.weight = self.fake_quantize(self.float_weight)
def forward_hook(self, input):
self.weight = self.float_weight
```
* Assignments to self.weight are needed because we can’t change forward function and in forward function they are using self.weight.
* But we will need to keep two copies of weight in this case, so it’s probably better to just swap the module
* So we want to just swap Conv to torch.nn.qat.Conv and Linear to torch.nn.qat.Linear
* qat modules will have fake_quant for output and weights inserted in forward function
## Convert
* flow should be identical to ptq, but the swapping dictionary is slightly different since modules are changed in prepare step.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/23082
ghstack-source-id: 86824650
Differential Revision: D16379374
fbshipit-source-id: 7d16d1acd87025065a24942ff92abf18e9fc8070
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/22732
Add support for quantization aware training in eager mode
Modifications to Post training flow:
## Prepare
* Fusion: e.g. (Conv, Bn) → ConvBn (float)
* Swapping: To insert fake_quant to weight, we need to swap the float modules that has weight with different qat modules, e.g. Conv → torch.nn.qat.Conv , ConvBn → torch.nn._intrinsic.qat.ConvBn
```
* previously we were thinking about modify the weight in forward_pre hook and change it back in forward_hook:
* def forward_pre_hook(self, input):
self.float_weight = self.weight
self.weight = self.fake_quantize(self.float_weight)
def forward_hook(self, input):
self.weight = self.float_weight
```
* Assignments to self.weight are needed because we can’t change forward function and in forward function they are using self.weight.
* But we will need to keep two copies of weight in this case, so it’s probably better to just swap the module
* So we want to just swap Conv to torch.nn.qat.Conv and Linear to torch.nn.qat.Linear
* qat modules will have fake_quant for output and weights inserted in forward function
## Convert
* flow should be identical to ptq, but the swapping dictionary is slightly different since modules are changed in prepare step.
Reviewed By: zafartahirov
Differential Revision: D16199356
fbshipit-source-id: 62aeaf47c12c62a87d9cac208f25f7592e245d6c
Summary:
* Deletes all weak script decorators / associated data structures / methods
* In order to keep supporting the standard library in script, this enables recursive script on any function defined in `torch.nn`
* Most changes in `torch/nn` are the result of `ag -Q "weak" torch/nn/ -l | xargs sed -i '/weak/d'`, only `rnn.py` needed manual editing to use the `ignore` and `export` to continue supporting the overloaded `forward` methods
* `Sequential`/`ModuleList` no longer need to be added to constants since they are compiled on demand
This should also fix https://github.com/pytorch/pytorch/issues/22212
Pull Request resolved: https://github.com/pytorch/pytorch/pull/22212
Differential Revision: D15988346
Pulled By: driazati
fbshipit-source-id: af223e3ad0580be895377312949997a70e988e4f
Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/21921
Call FBGEMM kernels to implement quantized linear operator. This operator is used only for inference.
Differential Revision: D15375695
fbshipit-source-id: b9ca6c156fd60481fea83e55603b2897f7bfc3eb
