Annotate linear node for `linear_dynamic_fp16` with `X86InductorQuantizer`
After `convert_pt2e`, the pattern will be
```
x
|
linear <- to_fp32 <- to_fp16 <- w
```
**Test plan**
```
pytest test/quantization/pt2e/test_x86inductor_quantizer.py -k test_linear_dynamic_fp16
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/141480
Approved by: https://github.com/jgong5, https://github.com/jerryzh168
Summary:
Populate nn.module.stack in _fuse_conv_bn_qat for replacement nodes that correspond to a `get_attr` node in the original graph.
In new training ir , `get_attr` nodes don't have `nn_module_stack` in node meta anymore (because the get_attr nodes are de-duplicated, so one get_attr node can potential have users in different module stacks).
We populate it by checking if "conv_input" or "conv_weight" replacement node has nn_module_stack. If not, we copy it from the conv node.
Test Plan:
CI
```
buck run fbcode//caffe2/test:quantization_pt2e -- -r test_preserve_nn_module_stack
```
Differential Revision: D66393517
Pull Request resolved: https://github.com/pytorch/pytorch/pull/141400
Approved by: https://github.com/angelayi
**About this PR**
This PR adds the following ops for `linear_dynamic_fp16` in onednn namespace. These ops are intended for PT2E quantization eager mode.
- `onednn::linear_prepack_fp16`: packs fp32 weight to an fp16 MkldnnCPU tensor.
- `onednn::linear_dynamic_fp16`: takes an fp32 CPU tensor and an fp16 MkldnnCPU tensor and compute linear in fp32
- `onednn::linear_relu_dynamic_fp16`: similar as the former and apply relu on output.
**Test plan**
`python test/test_quantization.py -k test_linear_dynamic_fp16_onednn`
**Implementation**
These ops call oneDNN lib under the hood. It's worth noting that oneDNN does not support f32 * f16 -> f32 computation, so we have to convert fp16 weight to fp32 before computation. And weight is still in plain format after packing.
**Correctness and performance**
Correctness is guaranteed by UT.
Performance of the new ops may be better than the FBGEMM implementation when weight shape is small but worse when weight shape is large. It's because weight dtype conversion and computation are not fused.
For example, I ran benchmarks on an Intel(R) Xeon(R) Platinum 8490H machine with different cores and shapes. When using 1 core per instance, the new implementation generally is faster for weight shape < 1024 * 1024. When using more cores, the threshold will increase.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/140376
Approved by: https://github.com/jerryzh168, https://github.com/jgong5
This PR adds C shim for `QConvPointWisePT2E` and `QConvPointWiseBinaryPT2E` similar to https://github.com/pytorch/pytorch/pull/138439. Besides that, we aligned the implementation of `qconv_pointwise` with `qlinear_pointwise` in the following aspects:
1. The parameter order of `qconv_pointwise` and `qlinear_pointwise` are quite different, we aligned the schema of `qconv_pointwise` to have similar parameter order as `qlinear_pointwise` to make it more consistent.
2. We always converted `x_scale` and `x_zero_point` to Tensors, just like in the lowering of `qlinear_pointwise`. This avoids the need to create two separate C APIs (one for `double x_scale` and `int64_t x_zero_point`, and another for `Tensor` versions). Instead, we only need one API for `Tensor`-based `x_scale` and `x_zero_point`. If we later add dynamic quantization for qconv (which will use `Tensor` for `x_scale` and `x_zero_point`), we can reuse the code from this PR and don't need to change the C shim layer API.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/138540
Approved by: https://github.com/jgong5, https://github.com/desertfire
ghstack dependencies: #138691, #138806
**MOTIVATION**
We recently integrated support for Intel Gaudi devices (identified as 'hpu') into the common_device_type framework via the pull request at https://github.com/pytorch/pytorch/pull/126970. This integration allows tests to be automatically instantiated for Gaudi devices upon loading the relevant library. Building on this development, the current pull request extends the utility of these hooks by adapting selected CUDA tests to operate on Gaudi devices. Additionally, we have confirmed that these modifications do not interfere with the existing tests on CUDA devices.
**CHANGES**
- Add support for HPU devices within the test_move_exported_model_bn using TEST_HPU flag
- Use instantiate_device_type_tests with targeted attributes to generate device-specific test instances.
- Apply skipIfHPU decorator to bypass tests that are not yet compatible with HPU devices.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/137863
Approved by: https://github.com/jerryzh168
Summary:
as title
also change it in `prepare_pt2e()` docstring
Test Plan:
```
buck2 run 'fbcode//mode/dev-nosan' fbcode//caffe2/test:quantization_pt2e_qat
buck2 run 'fbcode//mode/dev-nosan' fbcode//caffe2/test/quantization:test_quantization
```
Differential Revision: D63345059
Pull Request resolved: https://github.com/pytorch/pytorch/pull/137233
Approved by: https://github.com/tugsbayasgalan
Summary: Previously is_fbcode just checked whether the checkout was git or not. This is extremely error prone. Lets make it fool-proof.
Test Plan: unit tests
Reviewed By: masnesral
Differential Revision: D63545169
Pull Request resolved: https://github.com/pytorch/pytorch/pull/136871
Approved by: https://github.com/masnesral
Summary:
Add a customizable loss function callback to NodeAccuracySummary to
allow users to pass in their own loss function.
Also, fix some type errors and propagate better exception messages when
unexpected tensor comparisons occur. Finally, enhance the robustness of
`generate_numeric_debug_handle` in the case where it is called multiple
times on the same model, by avoiding reuse of the same IDs.
Test Plan: Added a test for this case in `test_numeric_debugger`.
Reviewed By: jerryzh168
Differential Revision: D62898297
Pull Request resolved: https://github.com/pytorch/pytorch/pull/136282
Approved by: https://github.com/jerryzh168
Summary:
Add a customizable loss function callback to NodeAccuracySummary to
allow users to pass in their own loss function.
Also, fix some type errors and propagate better exception messages when
unexpected tensor comparisons occur. Finally, enhance the robustness of
`generate_numeric_debug_handle` in the case where it is called multiple
times on the same model, by avoiding reuse of the same IDs.
Test Plan: Added a test for this case in `test_numeric_debugger`.
Reviewed By: jerryzh168
Differential Revision: D62898297
Pull Request resolved: https://github.com/pytorch/pytorch/pull/136282
Approved by: https://github.com/jerryzh168
Summary:
Previously we only checked dtype and is_dynamic to decide if two quantization spec are equivalent
this may not work in some cases, e.g. when people use different qscheme or quant_min/quant_max
This PR added checks for other fields as well
Test Plan:
regression tests
Reviewers:
Subscribers:
Tasks:
Tags:
Differential Revision: [D62530974](https://our.internmc.facebook.com/intern/diff/D62530974)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/135736
Approved by: https://github.com/sxu
Summary:
Fixes https://github.com/pytorch/pytorch/issues/134778
The previous D62304294 broke some executorch tests. It has already been reverted.
In this diff, `_collect_param_buffer_metadata()` is modified in a way that when a `call_function` node is encountered and its input nodes include `get_attr`. We skip the fields that have been collected previously and only collect rest of the fields. This prevents over-writing.
Test Plan:
```
buck2 test 'fbcode//mode/dev-nosan' fbcode//executorch/backends/xnnpack/test:test_xnnpack_ops
buck2 test 'fbcode//mode/dev-nosan' fbcode//caffe2/test/quantization:test_quantization -- -r test_re_export_preserve_handle
buck2 test 'fbcode//mode/dev-nosan' fbcode//caffe2/test/quantization:test_quantization -- -r test_run_decompositions_preserve_handle
```
Differential Revision: D62514208
Pull Request resolved: https://github.com/pytorch/pytorch/pull/135720
Approved by: https://github.com/zhxchen17, https://github.com/jerryzh168
Summary:
Skip test_prepare_qat_conv_bn_fusion_getitem_placeholder when we use training ir, since it's only for bn-getitem pattern, but the pattern doesn't exist in training ir.
Remove BLOCK_LIST since it's empty.
Now all internal unittests will use training ir.
Test Plan:
```
buck2 run 'fbcode//mode/dev-nosan' caffe2/test/quantization:test_quantization -- -r test_prepare_qat_conv_bn_fusion_getitem_placeholder
buck2 run 'fbcode//mode/dev-nosan' caffe2/test:quantization_pt2e_qat -- -r test_prepare_qat_conv_bn_fusion_getitem_placeholder
```
Differential Revision: D62387987
Pull Request resolved: https://github.com/pytorch/pytorch/pull/135729
Approved by: https://github.com/tugsbayasgalan
Summary: as title
Test Plan:
```
buck2 run 'fbcode//mode/dev-nosan' fbcode//caffe2/test:test_export -- -r test_conv_dynamic
buck2 run 'fbcode//mode/dev-nosan' fbcode//caffe2/test:fx -- -r matcher
buck2 run 'fbcode//mode/dev-nosan' fbcode//caffe2/test/quantization:test_quantization -- -r x86
```
CI
Differential Revision: D62448302
Pull Request resolved: https://github.com/pytorch/pytorch/pull/135623
Approved by: https://github.com/tugsbayasgalan
Summary: In new export_for_training, "stack_trace" does not exist in node meta anymore.
Test Plan:
```
buck run fbcode//mode/dev-nosan fbcode//caffe2/test:quantization_pt2e -- -r test_constant_prop_preserve_metadata
```
Reviewed By: angelayi
Differential Revision: D62219974
Pull Request resolved: https://github.com/pytorch/pytorch/pull/135268
Approved by: https://github.com/angelayi
Summary:
In graph of TestXNNPACKQuantizer.test_dynamic_linear_with_con test, some quantized_decomposed.quantize_per_tensor.default ops are becoming quantized_decomposed.dequantize_per_tensor.tensor ops when using the new training ir.
This is because we lift params/buffers before calling make_fx. So previously, for the graph that’s passed to make_fx,`graph.L__self___linear1.weight` is a tensor
now in training ir, graph.L__self___linear1.weight is a FakeTensor. This caused the node overload to be different.
Test Plan:
```
buck2 run 'fbcode//mode/dev-nosan' fbcode//caffe2/test/quantization:test_quantization -- -r test_dynamic_linear_with_conv
```
Differential Revision: D61364547
Pull Request resolved: https://github.com/pytorch/pytorch/pull/134525
Approved by: https://github.com/tugsbayasgalan, https://github.com/jerryzh168
Summary:
Fixed some quantization tests for new training ir:
Fix batch norm node pattern matcher. In training ir, we have `aten.batch_norm` node instead of `aten._native_batch_norm_legit` and `aten._native_batch_norm_legit_no_training`.
Test Plan:
```
buck run fbcode//mode/dev-nosan fbcode//caffe2/test:quantization_pt2e
```
Reviewed By: tugsbayasgalan
Differential Revision: D62209819
Pull Request resolved: https://github.com/pytorch/pytorch/pull/135184
Approved by: https://github.com/tugsbayasgalan
Summary:
This diff adds two new operators torch.ops._quantized.wrapped_linear_prepack and torch.ops._quantized.wrapped_quantized_linear_prepacked. It is a decomposition of the op torch.ops._quantized.wrapped_quantized_linear added in the previous diff.
We decomposed in this way as packed weight could be computed early so we don;t need to do it in every forward in AOTI
Reviewed By: jerryzh168
Differential Revision: D61395887
Pull Request resolved: https://github.com/pytorch/pytorch/pull/134232
Approved by: https://github.com/houseroad
Summary:
Make quantization tests compatible with the new training IR.
With the new batch norm node `torch.ops.aten.batch_norm.default`, we don't need an additional getitem node after the bn node, so tests need to be fixed to not check for the getitem node.
We added a capture_pre_autograd_graph_using_training_ir() function, which returns True when we are using the training ir, and False otherwise. This way, the code supports both training ir and the old ir.
For now, we are just rolling out the training ir for fbcode internal tests.
Test Plan:
```
buck2 run 'fbcode//mode/dev-nosan' fbcode//caffe2/test/quantization:test_quantization -- -r test_qat_preserve_source_fn_stack
buck2 run 'fbcode//mode/dev-nosan' fbcode//caffe2/test/quantization:test_quantization -- -r test_qat_update_shared_qspec
buck2 run 'fbcode//mode/dev-nosan' fbcode//caffe2/test/quantization:test_quantization -- -r test_conv2d
buck2 run 'fbcode//mode/dev-nosan' fbcode//caffe2/test/quantization:test_quantization -- -r test_qat_conv_bn_relu_fusion
buck2 run 'fbcode//mode/dev-nosan' fbcode//caffe2/test/quantization:test_quantization -- -r test_qat_conv_bn_fusion
buck2 run 'fbcode//mode/dev-nosan' fbcode//caffe2/test/quantization:test_quantization -- -r test_qat_conv_bn_fusion_literal_args
```
Reviewed By: andrewor14, tugsbayasgalan
Differential Revision: D61292102
Pull Request resolved: https://github.com/pytorch/pytorch/pull/134259
Approved by: https://github.com/tugsbayasgalan
Summary:
This diff adds a new operator wrapped_quantized_linear (torch.ops._quantized.wrapped_quantized_linear) and takes the following input argument: input (in fp32) , input_scale, input_zero_point, weight (in fp32), weight_scale, weight_zero_point, bias (in fp32), output_scale, output_zero_point, and out_channel. It does the following
1. Use quantize_per_tensor(input, input_scale, input_zero_point) to quantize the input tensor to int8
2. Use quantized::linear_prepack(weight, weight_scale, weight_zero_point, bias) to pack the weight and bias
3. Use quantized::linear to perform int8 quantized linear
4. dequantize
This new op is essentially a wrapper of mutiple ops. We do this as torch.export cannot handle models where it has old quantize apis.
Reviewed By: jerryzh168
Differential Revision: D61377266
Pull Request resolved: https://github.com/pytorch/pytorch/pull/134024
Approved by: https://github.com/houseroad
Summary:
This diff aims to fix the GPU Test skips in the quantization tests under the `caffe2/test/quantization` directory. The changes made in the `TARGETS` files include adding the `should_use_remote_gpu` flag to enable remote GPU testing. This should help to resolve the skipped tests and improve the overall test coverage.
[This diff] Fixed skip count: 4
[Running total] Fixed skip count: 4
Note: Creating separate diffs for each test-group.
Test Plan:
**281475054644766**: buck2 test 'fbcode//mode/opt' fbcode//caffe2/test/quantization:test_quantization -- --exact 'caffe2/test/quantization:test_quantization - test_compare_per_channel_device_numerics (caffe2.test.quantization.core.test_quantized_tensor.TestQuantizedTensor)'
https://www.internalfb.com/intern/testinfra/testrun/5629499773981783
**281475054644780**: buck2 test 'fbcode//mode/opt' fbcode//caffe2/test/quantization:test_quantization -- --exact 'caffe2/test/quantization:test_quantization - test_compare_per_tensor_device_numerics (caffe2.test.quantization.core.test_quantized_tensor.TestQuantizedTensor)'
https://www.internalfb.com/intern/testinfra/testrun/11540474087422107
**281475054644853**: buck2 test 'fbcode//mode/opt' fbcode//caffe2/test/quantization:test_quantization -- --exact 'caffe2/test/quantization:test_quantization - test_quant_pin_memory (caffe2.test.quantization.core.test_quantized_tensor.TestQuantizedTensor)'
https://www.internalfb.com/intern/testinfra/testrun/11540474087422477
**844425008078016**: buck2 test 'fbcode//mode/opt' fbcode//caffe2/test/quantization:test_quantization -- --exact 'caffe2/test/quantization:test_quantization - test_cuda_quantization_does_not_pin_memory (caffe2.test.quantization.core.test_quantized_tensor.TestQuantizedTensor)'
https://www.internalfb.com/intern/testinfra/testrun/1407375259845199
Differential Revision: D60055277
Pull Request resolved: https://github.com/pytorch/pytorch/pull/133158
Approved by: https://github.com/jovianjaison
Summary:
The Quantizer subclass can return a new model from `transform_for_annotation`,
and this is common if it uses any ExportPass subclass which does not mutate in-place.
Use the returned model instead of assuming its the same.
Differential Revision: D60869676
Pull Request resolved: https://github.com/pytorch/pytorch/pull/132893
Approved by: https://github.com/jerryzh168
Summary:
- make default DCE pass check schema,
- need to rebase onto https://github.com/pytorch/pytorch/pull/131651 after it's in phabricator (for now the change is manually added).
- mark Proxy dump as NotImplemented for better error msg
- Remove Proxy from tensors when dumping models, as Proxy cannot be dumped.
More details in https://docs.google.com/document/d/1G5vmTXjzxoyVGRI2kpA1gQukK_Glyg2NrE0Oh6Nlg9A/edit?usp=sharing.
Test Plan:
CI
```
- buck2 run 'fbcode//mode/dev-nosan' fbcode//caffe2/test/quantization:test_quantization -- -r qat_conv2d
- test_export.py
- buck2 run 'fbcode//mode/dev-nosan' fbcode//modai/test:test_modai -- -r test_qat_stinson_htp_export
- buck2 run 'fbcode//mode/dev-nosan' fbcode//vizard_projects/ml_depth/tests:test_model -- -r test_qat_model_et
- buck2 run 'fbcode//mode/dev-nosan' fbcode//caffe2/test:fx -- -r dce
- buck2 run 'fbcode//mode/dev-nosan' fbcode//bolt/nn/executorch/backends/tests:qnn_test -- -r test_qat_bias=False,use_3d_input=False
- buck2 run 'fbcode//mode/dev-nosan' fbcode//bolt/nn/executorch/backends/tests:qnn_test -- -r test_qat_bias=True,use_3d_input=False
- buck2 run 'fbcode//mode/dev-nosan' fbcode//caffe2/test/quantization:test_quantization -- -r test_fold_bn_erases_bn_node
```
Reviewed By: angelayi
Differential Revision: D60319175
Pull Request resolved: https://github.com/pytorch/pytorch/pull/132764
Approved by: https://github.com/angelayi
Summary: Previously, when folding BN into conv, we rely on DCE
to clean up the unused BN node from the graph. This works if
the model is already in eval mode, but fails if the model is
still in train mode because DCE doesn't remove nodes with
potential side effects (in this case `_native_batch_norm_legit`).
This required users to move the model to eval mode before calling
convert in order to get a properly DCE'd graph.
To solve this, we manually erase the BN node after folding
instead of relying on DCE. This relaxes the ordering constraints
between `move_exported_model_to_eval` and `convert_pt2e`.
Test Plan:
python test/test_quantization.py TestQuantizePT2EQAT_ConvBn1d.test_fold_bn_erases_bn_node
python test/test_quantization.py TestQuantizePT2EQAT_ConvBn2d.test_fold_bn_erases_bn_node
Reviewers: jerryzh168, yushangdi
Subscribers: jerryzh168, yushangdi, supriyar
Differential Revision: [D60520149](https://our.internmc.facebook.com/intern/diff/D60520149)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/131651
Approved by: https://github.com/yushangdi, https://github.com/leslie-fang-intel
Add similar semantics for creating a buffer object similar to creating a parameter. This is done by introducing a new Buffer class that can be used for type disambiguation. The underlying functionality of registering a buffer remains the same as the register_buffer method has not been changed. The persistent parameter in the Buffer type is to indicate whether a buffer object should be persistent or not. Other non-test changes have to do with getting the new Buffer type recognized by inductor and dynamo. Remaining changes are test changes to make sure that the Buffer type can be used as a drop in replacement for register_buffer as it just leads to register_buffer being called. The addition of this new functionality still allows for normal tensors to be used as buffers so these changes are intended to be backwards compatible.
Fixes#35735
Co-authored-by: Mikayla Gawarecki <mikaylagawarecki@gmail.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125971
Approved by: https://github.com/albanD, https://github.com/anijain2305, https://github.com/mlazos
Summary: Previously, when folding BN into conv, we rely on DCE
to clean up the unused BN node from the graph. This works if
the model is already in eval mode, but fails if the model is
still in train mode because DCE doesn't remove nodes with
potential side effects (in this case `_native_batch_norm_legit`).
This required users to move the model to eval mode before calling
convert in order to get a properly DCE'd graph.
To solve this, we manually erase the BN node after folding
instead of relying on DCE. This relaxes the ordering constraints
between `move_exported_model_to_eval` and `convert_pt2e`.
Test Plan:
python test/test_quantization.py TestQuantizePT2EQAT_ConvBn1d.test_fold_bn_erases_bn_node
python test/test_quantization.py TestQuantizePT2EQAT_ConvBn2d.test_fold_bn_erases_bn_node
Reviewers: jerryzh168, yushangdi
Subscribers: jerryzh168, yushangdi, supriyar
Pull Request resolved: https://github.com/pytorch/pytorch/pull/131651
Approved by: https://github.com/yushangdi
Summary:
This pr fixes all the places in strict export stack where the output node's meta is not preserved correctly. However, we're getting a new error for the test we intend to fix: `buck2 run caffe2/test/quantization:test_quantization -- -r "test_re_export_preserve_handle"`:
The `get_attr` nodes has wrong metadata. I guess there are more things need to be fixed to get it working but it's beyond the scope of this PR.
Test Plan: buck2 run caffe2/test/quantization:test_quantization -- -r "test_re_export_preserve_handle"
Differential Revision: D60198221
Pull Request resolved: https://github.com/pytorch/pytorch/pull/131706
Approved by: https://github.com/yushangdi
Summary:
This diff reverts D59561509
D59561509: [FX][export] DCE pass, check schema for node impurity (#130395) by yushangdi causes the following test failure:
Tests affected:
- [cogwheel:cogwheel_mtia_cmf_m5_shrunk_test#test_flow_with_verification](https://www.internalfb.com/intern/test/844425041436985/)
Here's the Multisect link:
https://www.internalfb.com/multisect/6533402
Here are the tasks that are relevant to this breakage:
T191383430: 10+ tests unhealthy for ads_mtia_inference
The backout may land if someone accepts it.
If this diff has been generated in error, you can Commandeer and Abandon it.
Test Plan: NA
Differential Revision: D60029318
Pull Request resolved: https://github.com/pytorch/pytorch/pull/131341
Approved by: https://github.com/angelayi
Earlier the signature of dequantize ops for decomposed quantized Tensor was changed for wider use-cases where the output dtype can be different from torch.float and needs to be passed during dequantization.
Please refer: https://github.com/pytorch/pytorch/pull/121450
However, setting of correct output dtype for dequantize ops was still missing in convert_pt2e flow.
This change enables the users to use PT2E quantization flow with non torch.float unquantized dtype, such as torch.bfloat16.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/128953
Approved by: https://github.com/jgong5, https://github.com/jerryzh168
Summary:
Add three top level APIs for numeric debugger in pt2e flow that can log intermediate output in the model
and calculate summary for metric comparisons between nodes in two graphs
* `prepare_for_propagation_comparison`
* `extract_results_from_loggers`
* `compare_results`
Test Plan:
python test/test_quantization.py -k test_prepare_for_propagation_comparison
python test/test_quantization.py -k test_extract_results_from_loggers
Reviewers:
Subscribers:
Tasks:
Tags:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/130643
Approved by: https://github.com/dulinriley, https://github.com/tarun292
Summary:
* Added support for preserving it during deepcopy, need to remap the args since _numeric_debug_handle refers
to the nodes in the graph
TODO: need to fully support re-export, currently the metadata for output node is not preserved
Test Plan:
python test/test_quantization.py -k test_deepcopy_preserve_handle
python test/test_quantization.py -k test_copy_preserve_handle
all related tests:
python test/test_quantization.py -k TestGenerateNumericDebugHandle
Reviewers:
Subscribers:
Tasks:
Tags:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129287
Approved by: https://github.com/zhxchen17
Summary:
Previously we store edge id in numeric_debug_handle to support operator fusion and operator decomposition throughout the stack,
but according to feedback from customers, people prefer the simpler per-node id, and they are fine with not having the additional
support for numerical debugging for inputs and willing to hack around to achieve this.
This PR changes the structure of numeric_debug_handle to store unique_id for each node instead.
e.g.
graph:
```
node = op(input_node, weight_node)
```
Before:
```
node.meta[NUMERIC_DEBUG_HANDLE_KEY] = {input_node: id1, weight_node: id2, "output": id3}
```
After:
```
node.meta[NUMERIC_DEBUG_HANDLE_KEY] = id1
```
Test Plan:
python test/test_quantization.py -k TestGenerateNumericDebugHandle
Reviewers:
Subscribers:
Tasks:
Tags:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129811
Approved by: https://github.com/tarun292
Summary:
Previously we store edge id in numeric_debug_handle to support operator fusion and operator decomposition throughout the stack,
but according to feedback from customers, people prefer the simpler per-node id, and they are fine with not having the additional
support for numerical debugging for inputs and willing to hack around to achieve this.
This PR changes the structure of numeric_debug_handle to store unique_id for each node instead.
e.g.
graph:
```
node = op(input_node, weight_node)
```
Before:
```
node.meta[NUMERIC_DEBUG_HANDLE_KEY] = {input_node: id1, weight_node: id2, "output": id3}
```
After:
```
node.meta[NUMERIC_DEBUG_HANDLE_KEY] = id1
```
Test Plan:
python test/test_quantization.py -k TestGenerateNumericDebugHandle
Reviewers:
Subscribers:
Tasks:
Tags:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129811
Approved by: https://github.com/tarun292
Summary:
There were two problems with the HistogramObserver:
1. It does not work when someone passes a batch_size 1, tensor_size 1 data-point.
2. The Histogram doesn't seem to actually update if the range of the new x falls within the old one
These issues were both fixed.
On top of this, I greatly simplified the logic for the histogram updating. Now, it doesn't do the downsampling anymore, which saves a ton of memory and code. The accuracy can still be controlled with the upsampling ratio. This ratio was also too high for the accuracy we generally need here, I reduced the default for this.
Also the code is cleaner now, much easier to follow what's happening.
test_histogram_observer_same_inputs was likely wrong - If I pass 0s and 1s to my histogramobserver, I want them to actually count! The current test now thinks it's good to discard and ignore these values.
Test Plan: You can run the included tests.
Differential Revision: D58931336
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129387
Approved by: https://github.com/jerryzh168
**Summary**
We change the schema of QLinear Binary, so it will be easier to enable the corresponding gemm template.
- Extra input of binary post-op is a tensor which needs to be an input node of autotuning, we need to move it at front of `output_scale` which is a scalar.
- We also move it at front of `bias`, since `bias` is optional tensor for this fusion, but `other` is a must to have for linear binary fusion.
**Test Plan**
```
python -u -m pytest -s -v test/quantization/core/test_quantized_op.py -k qlinear
python -u -m pytest -s -v test/inductor/test_mkldnn_pattern_matcher.py -k qlinear
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129049
Approved by: https://github.com/jgong5, https://github.com/jansel
ghstack dependencies: #128825, #129048
Changes by apply order:
1. Replace all `".."` and `os.pardir` usage with `os.path.dirname(...)`.
2. Replace nested `os.path.dirname(os.path.dirname(...))` call with `str(Path(...).parent.parent)`.
3. Reorder `.absolute()` ~/ `.resolve()`~ and `.parent`: always resolve the path first.
`.parent{...}.absolute()` -> `.absolute().parent{...}`
4. Replace chained `.parent x N` with `.parents[${N - 1}]`: the code is easier to read (see 5.)
`.parent.parent.parent.parent` -> `.parents[3]`
5. ~Replace `.parents[${N - 1}]` with `.parents[${N} - 1]`: the code is easier to read and does not introduce any runtime overhead.~
~`.parents[3]` -> `.parents[4 - 1]`~
6. ~Replace `.parents[2 - 1]` with `.parent.parent`: because the code is shorter and easier to read.~
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129374
Approved by: https://github.com/justinchuby, https://github.com/malfet
Changes by apply order:
1. Replace all `".."` and `os.pardir` usage with `os.path.dirname(...)`.
2. Replace nested `os.path.dirname(os.path.dirname(...))` call with `str(Path(...).parent.parent)`.
3. Reorder `.absolute()` ~/ `.resolve()`~ and `.parent`: always resolve the path first.
`.parent{...}.absolute()` -> `.absolute().parent{...}`
4. Replace chained `.parent x N` with `.parents[${N - 1}]`: the code is easier to read (see 5.)
`.parent.parent.parent.parent` -> `.parents[3]`
5. ~Replace `.parents[${N - 1}]` with `.parents[${N} - 1]`: the code is easier to read and does not introduce any runtime overhead.~
~`.parents[3]` -> `.parents[4 - 1]`~
6. ~Replace `.parents[2 - 1]` with `.parent.parent`: because the code is shorter and easier to read.~
Pull Request resolved: https://github.com/pytorch/pytorch/pull/129374
Approved by: https://github.com/justinchuby, https://github.com/malfet
Summary:
Added `set_module_name_qconfig` support to allow users to set configurations based on module name in `X86InductorQuantizer`.
For example, only quantize the `sub`:
```python
class M(torch.nn.Module):
def __init__(self):
super().__init__()
self.linear = torch.nn.Linear(5, 5)
self.sub = Sub()
def forward(self, x):
x = self.linear(x)
x = self.sub(x)
return x
m = M().eval()
example_inputs = (torch.randn(3, 5),)
# Set config for a specific submodule.
quantizer = X86InductorQuantizer()
quantizer.set_module_name_qconfig("sub", xiq.get_default_x86_inductor_quantization_config())
```
- Added `set_module_name_qconfig` to allow user set the configuration at the `module_name` level.
- Unified the annotation process to follow this order: `module_name_qconfig`, `operator_type_qconfig`, and `global_config`.
- Added `config_checker` to validate all user configurations and prevent mixing of static/dynamic or QAT/non-QAT configs.
- Moved `_get_module_name_filter` from `xnnpack_quantizer.py` into `utils.py` as it common for all quantizer.
Test Plan
```bash
python -m pytest quantization/pt2e/test_x86inductor_quantizer.py -k test_set_module_name
```
@Xia-Weiwen @leslie-fang-intel @jgong5
Pull Request resolved: https://github.com/pytorch/pytorch/pull/126044
Approved by: https://github.com/jgong5, https://github.com/leslie-fang-intel, https://github.com/jerryzh168
Summary:
For Speech sequential model, there could be a case where model(data) does not work correctly for feed forward,
Speech model uses a different type of Criterion (a.k.a loss function) to feed a data on individual components like encoder, predictor, joiner.
Hence we need extra parameter to pass feedforward wrapper
Differential Revision: D57680391
Pull Request resolved: https://github.com/pytorch/pytorch/pull/126891
Approved by: https://github.com/jerryzh168
Summary:
After QAT is completed or given pre-tuned weight observer via tunable PTQ algorithm, it should not over-write again with a given weight, at least for static QAT never.
Dynamic QAT also does not require to re-run weight observer again by design.
This is a fix
Test Plan: Signals
Differential Revision: D57747749
Pull Request resolved: https://github.com/pytorch/pytorch/pull/127309
Approved by: https://github.com/jerryzh168
Summary: The PT2E quantization flow does not support unquantized
outputs yet. To work around this, users may wish to remove the
output observer from their graphs. However, this fails currently
in some cases because the `PortNodeMetaForQDQ` pass is too
restrictive, for example:
```
conv -> obs -------> output0
\\-> add -> output1
```
Previously we expected conv to always have exactly 1 user,
which is the observer. When the observer is removed, however,
conv now has 2 users, and this fails the check.
```
conv -------> output0
\\-> add -> output1
```
This commit relaxes the error into a warning to enable
this workaround.
Test Plan:
python test/test_quantization.py TestQuantizePT2E.test_multi_users_without_output_observer
Reviewers: jerryzh168
Subscribers: jerryzh168, supriyar
Differential Revision: [D57472601](https://our.internmc.facebook.com/intern/diff/D57472601)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/126487
Approved by: https://github.com/tarun292
Summary: Follow-up to https://github.com/pytorch/ao/pull/229.
This resolves the difference between `input.div(scales)` and
`input.mul(1.0 / scales)`, which results in small numerical
discrepancies on some inputs.
Test Plan:
python test/test_quantization.py TestQuantizedTensor.test_decomposed_quantize_per_channel_group
python test/test_quantization.py TestQuantizedTensor.test_decomposed_quantize_per_token
Reviewers: jerryzh168
Subscribers: jerryzh168, supriyar
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125781
Approved by: https://github.com/jerryzh168
**Summary**
Per the discussion in https://github.com/pytorch/pytorch/pull/123444, the `decomposed quant/dequant` patterns changed after https://github.com/pytorch/pytorch/pull/123445, we can move the optimization of `decomposed quant/dequant` from inductor decomposition into lowering phase to avoid the changes. In this way, we can:
- Avoid the pattern matcher failure introduced in https://github.com/pytorch/pytorch/pull/123445
- Make the quantization pattern clearer in the pattern matcher phase, since the `quant/dequant` nodes have not been decomposed.
**Changes in this PR**
- Move optimization of `decomposed quant/dequant` from inductor decomposition into lowering phase.
- Corresponding changes in the quantization pattern matcher to ensure no bc-breaking.
**TestPlan**
```
python -u -m pytest -s -v test/inductor/test_mkldnn_pattern_matcher.py -k test_q
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124041
Approved by: https://github.com/peterbell10, https://github.com/jgong5
**Summary**
Per the discussion in https://github.com/pytorch/pytorch/pull/123444, the `decomposed quant/dequant` patterns changed after https://github.com/pytorch/pytorch/pull/123445, we can move the optimization of `decomposed quant/dequant` from inductor decomposition into lowering phase to avoid the changes. In this way, we can:
- Avoid the pattern matcher failure introduced in https://github.com/pytorch/pytorch/pull/123445
- Make the quantization pattern clearer in the pattern matcher phase, since the `quant/dequant` nodes have not been decomposed.
**Changes in this PR**
- Move optimization of `decomposed quant/dequant` from inductor decomposition into lowering phase.
- Corresponding changes in the quantization pattern matcher to ensure no bc-breaking.
**TestPlan**
```
python -u -m pytest -s -v test/inductor/test_mkldnn_pattern_matcher.py -k test_q
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124041
Approved by: https://github.com/peterbell10, https://github.com/jgong5
Summary: This commit fixes the pattern matching for conv-bn
during QAT fusion where both weight and bias are quantized per
channel. Previously this failed because weights and biases used
the same example kwargs for their scales and zero points,
causing these qparams to be tied during pattern matching.
Test Plan:
python test/test_quantization.py TestQuantizePT2EQAT_ConvBn1d.test_qat_conv_bn_per_channel_weight_bias
python test/test_quantization.py TestQuantizePT2EQAT_ConvBn2d.test_qat_conv_bn_per_channel_weight_bias
Reviewers: jerryzh168, angelayi
Subscribers: jerryzh168, angelayi, supriyar
Differential Revision: [D56740694](https://our.internmc.facebook.com/intern/diff/D56740694)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/125208
Approved by: https://github.com/angelayi
Summary: Before in `move_exported_model_to_train/eval`, we only
switched the CPU versions of the batch norm op. This commit adds
support for the cuda versions of the op too. Note that this fix
is temporary; we won't have to differentiate between these two
cases once we have batch norm consolidation.
Test Plan:
python test/test_quantization.py -k test_move_exported_model_bn
Reviewers: jerryzh168
Subscribers: jerryzh168, leslie-fang-intel, supriyar
Differential Revision: [D56070054](https://our.internmc.facebook.com/intern/diff/D56070054)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/123957
Approved by: https://github.com/jerryzh168
Summary: Avoid situation where the graph traversal finds a matmul node with a `get_attr` as its `args[0]`, and incorrectly propagate the `get_attr`'s meta to everything downstream.
Test Plan: CI
Differential Revision: D56219120
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124415
Approved by: https://github.com/jerryzh168
Summary: Before in `move_exported_model_to_train/eval`, we only
switched the CPU versions of the batch norm op. This commit adds
support for the cuda versions of the op too. Note that this fix
is temporary; we won't have to differentiate between these two
cases once we have batch norm consolidation.
Test Plan:
python test/test_quantization.py -k test_move_exported_model_bn
Reviewers: jerryzh168
Subscribers: jerryzh168, leslie-fang-intel, supriyar
Differential Revision: [D56070054](https://our.internmc.facebook.com/intern/diff/D56070054)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/123957
Approved by: https://github.com/jerryzh168
This commit enables float8_e5m2 and float8_e4m3fn dtypes in fx quantization and PT2E.
Motivation for using fp8 quantization instead of int8:
- it works better to run inference with the same datatype the model was trained with,
- fp8 can handle outliers better, which is one of the problems in LLMs activations.
The numerical recipe we want to use it for is fp8 inference:
- bgemms/gemms running in float8_e4m3fn,
- Per-Tensor-Quantization/Scaling,
- amax observer for measurement with input_backoff and weight_backoff.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/123161
Approved by: https://github.com/jgong5, https://github.com/jerryzh168
Adds the following to allowed globals for the `weights_only` unpickler
- [x] `torch._utils._rebuild_qtensor` and qtensor related types
- [x] `torch._utils._rebuild_parameter_with_state` (used deserializing a parameter that has user-defined attributes like `Param.foo`)
The remaining rebuild functions that have not been allowlisted are
- [x] `torch._utils._rebuild_wrapper_subclass` (allowlisted in above PR)
- [ ] `torch._utils._rebuild_device_tensor_from_numpy`
- [ ] `torch._utils._rebuild_xla_tensor` (legacy)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124330
Approved by: https://github.com/albanD
Update ruff to 0.4.1 .
This version fixes a lot false negatives/false positives, is 20-40% faster, and has various other bug fixes.
Below is a before and after table showing the execution time of ruff lint and ruff format in milliseconds courtesy of https://astral.sh/blog/ruff-v0.4.0
| Repository | Linter (v0.3) | Linter (v0.4) | Formatter (v0.3) | Formatter (v0.4) |
|----------------------------------------------------|---------------|---------------|------------------|------------------|
| [pytorch/pytorch](https://github.com/pytorch/pytorch) | 328.7 | 251.8 | 351.1 | 274.9 |
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124549
Approved by: https://github.com/ezyang
Summary: https://github.com/pytorch/pytorch/pull/123452 added
backward support to this op by turning it into
CompositeImplicitAutograd, which meant it gets decomposed during
export/compile. However, this is not desirable behavior for the
PTQ case when we try to lower the model. This commit enables
QAT without breaking PTQ by refactoring the impl into a separate
op that does have backward support.
Test Plan:
python test/test_quantization.py -k test_decomposed_choose_qparams_per_token_asymmetric_backward
Reviewers: jerryzh168, digantdesai, zou3519
Subscribers: jerryzh168, digantdesai, zou3519, supriyar
Differential Revision: [D56192116](https://our.internmc.facebook.com/intern/diff/D56192116)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/124178
Approved by: https://github.com/digantdesai
Automatic fixes that replaces certain list comprehensions with generator ones where appropriate so that they are immediately consumed. This is preview functionality in ruff for rule C419 and it was automatically applied.
Co-authored-by: Nikita Shulga <2453524+malfet@users.noreply.github.com>
Pull Request resolved: https://github.com/pytorch/pytorch/pull/123960
Approved by: https://github.com/malfet
Summary: When running the backward for this op, we get the error:
```
RuntimeError: derivative for aten::aminmax is not implemented
```
This commit replaces this call with separate amin and amax
calls instead, which do have implemented derivatives.
Test Plan:
python test/test_quantization.py -k test_decomposed_choose_qparams_per_token_asymmetric_backward
Reviewers: jerryzh168, digantdesai
Subscribers: jerryzh168, digantdesai, supriyar
Differential Revision: [D55805170](https://our.internmc.facebook.com/intern/diff/D55805170)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/123452
Approved by: https://github.com/digantdesai, https://github.com/jerryzh168
**Summary**
Add `matmul` in the quantization recipes, noting that it's not a general recipe but tailored to meet accuracy criteria for specific models. `matmul` recipe is disabled by default.
**Test Plan**
```
python -m pytest quantization/pt2e/test_x86inductor_quantizer.py -k test_attention_block
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/122776
Approved by: https://github.com/jgong5, https://github.com/jerryzh168
ghstack dependencies: #122775
**Summary**
Default recipes are enabled in `X86InductorQuantizer` and request comes to customize recipes based on these defaults.
- Avoid annotation propagation and restrict annotation only to annotate `conv`/`linear`.
- Add `matmul` in the quantization recipes, noting that it's not a general recipe but tailored to meet accuracy criteria for specific models.
To meet these requests, we made changes in this PR by introducing interface as `set_function_type_qconfig` and `set_module_type_qconfig`
- `set_function_type_qconfig` accepts functional input as `torch.nn.functional.linear` or `torch.matmul`; `set_module_type_qconfig` accepts nn.Module input as `torch.nn.Conv2d`.
- To disable the recipe for this operator, user can simply exclude it from the list of operations as `quantizer.set_function_type_qconfig(op, None)`.
- To modify or extend the recipe for this operator with default recipe, user can customize as `quantizer.set_function_type_qconfig(op, config)`.
**Test Plan**
```
python -m pytest quantization/pt2e/test_x86inductor_quantizer.py -k test_filter_conv2d_recipe
python -m pytest quantization/pt2e/test_x86inductor_quantizer.py -k test_filter_linear_recipe
python -m pytest quantization/pt2e/test_x86inductor_quantizer.py -k test_filter_maxpool2d_recipe
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/122775
Approved by: https://github.com/jgong5, https://github.com/jerryzh168
**Note**: This is a reopen of https://github.com/pytorch/pytorch/pull/122288, which was merged by `ghstack land` to its base (not main) by mistake.
**Description**
Add qlinear_binary op for X86Inductor backend of quantization PT2E. It only supports `add` and `add_relu` now.
It will use post op sum if the extra input has the same dtype as output. Otherwise, it uses binary add.
```
+-------------------+--------------+---------------+
| Extra input dtype | Output dtype | Post op |
+-------------------+--------------+---------------+
| Fp32/bf16 | fp32/bf16 | sum or add* |
+-------------------+--------------+---------------+
| Fp32/bf16 | int8 | add |
+-------------------+--------------+---------------+
| int8 | fp32/bf16 | not supported |
+-------------------+--------------+---------------+
| int8 | int8 | sum |
+-------------------+--------------+---------------+
*Use sum if extra input and output have the same dtype; otherwise use add.
```
**Test plan**
python test_quantization.py -k test_qlinear_add_pt2e
python test_quantization.py -k test_qlinear_add_relu_pt2e
Pull Request resolved: https://github.com/pytorch/pytorch/pull/123144
Approved by: https://github.com/leslie-fang-intel, https://github.com/jgong5, https://github.com/jerryzh168
Summary:
Even with changes in D55347133, it is still possible to OOM in histogram observer, because the size of allocated tensor also depends on *downsample_rate*.
For example, I still see OOM due to the attempt of allocating a 10GB+ histogram tensor in multi-task model.
To fix OOM issue better, we use *try-catch* clause to avoid OOM.
Empirically, we set the max size of a single histogram tensor size to 1 GB.
Test Plan: Test the change for Multi-Task model (depth + segmentation)
Differential Revision: D55567292
Pull Request resolved: https://github.com/pytorch/pytorch/pull/123309
Approved by: https://github.com/jerryzh168
Summary: After we migrate to torch.export, we won't see ops like add_ and mul_ due to functionalization. We are rolling out pre dispatch export, so for now we just skip those mutating ops in tests.
Test Plan: buck run mode/opt caffe2/test/quantization:test_quantization
Reviewed By: tugsbayasgalan
Differential Revision: D55442019
Pull Request resolved: https://github.com/pytorch/pytorch/pull/122863
Approved by: https://github.com/clee2000
Summary:
We add a new op quantized.linear_unpacked_dynamic_fp16, which is essentially linear_dynamic_fp16 with different (unpacked) weight/bias format.
This op does packing on the fly for each call with standard at::Tensor weight & bias.
Test Plan:
Included in commit.
test_quantized_op::test_unpacked_qlinear_dynamic_fp16
Differential Revision: [D55433203](https://our.internmc.facebook.com/intern/diff/D55433203)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/122762
Approved by: https://github.com/jerryzh168
Summary:
Right now we don't insert additional observers (share observers) if qspec.dtype and qspec.is_dynamic matches exactly,
since fixed qparams quantization spec and derived quantization spec do have have is_dynamic field curerntly, observer sharing does not happen between them and quantization spec, in this PR we fixed the issue by
adding is_dynamic to all quantization specs.
Note: SharedQuantizationSpec should probably be its own type in the future
TODO later:
(1). move all these fields (dtype, is_dynamic, quant_min, quant_max etc.) to QuantizationSpecBase,
(2). make SharedQuantizationSpec a separate type
(3). add quant_min/quant_max in observer sharing checking in pt2e/prepare.py
Test Plan:
python test/test_quantization.py -k test_fixed_qparams_qspec_observer_dedup
Reviewers:
Subscribers:
Tasks:
Tags:
Differential Revision: [D55396546](https://our.internmc.facebook.com/intern/diff/D55396546)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/122734
Approved by: https://github.com/andrewor14
Summary:
also added some utils in xnnpack_quantizer_utils.py
* annotate_conv_tranpsose_bn_relu and annotate_conv_transpose_bn -> this is for QAT
* annotate_conv_transpose_relu
conv_transpose + bn weights fusion is performed automatically and can not be disabled currently
we can add support to allow disable this fusion later if needed
Test Plan:
python test/test_quantization.py -k test_conv_transpose_bn_fusion
Reviewers:
Subscribers:
Tasks:
Tags:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/122046
Approved by: https://github.com/andrewor14
