Commit Graph

4 Commits

Author SHA1 Message Date
Jerry Zhang
7b4d080496 [quant][pt2e] Rename _pt2e to pt2e (#104668)
Summary:
X-link: https://github.com/pytorch/executorch/pull/3

att

Test Plan: Imported from OSS

Differential Revision: D47202807

Pull Request resolved: https://github.com/pytorch/pytorch/pull/104668
Approved by: https://github.com/andrewor14
2023-07-15 06:34:17 +00:00
leslie-fang-intel
945a257277 [Quant][PT2E] Supported customized _EQUIVALENT_TYPES in Module Partition API (#102516)
**Summary**
`Module Partition API` can simplify the pattern match process in Quantization annotation. However, current implementation of
`Module Partition API` has hardcoded `_EQUIVALENT_TYPES` 999bae0f54/torch/ao/quantization/_pt2e/graph_utils.py (L13-L20). So, PyTorch Extension Libraries such as [intel-extension-for-pytorch](https://github.com/intel/intel-extension-for-pytorch) can't use `Module Partition API` with customized `_EQUIVALENT_TYPES` . In this PR, we plan to enable customized `_EQUIVALENT_TYPES` by pass in parameter.

**Test Plan**
```
python -m pytest test_graph_utils.py -k test_customized_equivalet_types_dict
```

Pull Request resolved: https://github.com/pytorch/pytorch/pull/102516
Approved by: https://github.com/jgong5, https://github.com/kimishpatel
2023-06-28 00:20:25 +00:00
Andrew Or
a1142053f0 [reland][quant][test] Fix broken PT2 import, add warnings (#102819)
Summary:
We are currently silently skipping all PT2 quantization
tests due to a recent typo. This commit fixes this and also adds
warnings so it'll be easier to debug similar issues in the future.

Test Plan: python test/test_quantization.py

Differential Revision: D46383546

Pull Request resolved: https://github.com/pytorch/pytorch/pull/102819
Approved by: https://github.com/jerryzh168
2023-06-02 22:35:30 +00:00
Kimish Patel
4cb6add471 [PT2][Quant] Use module partition for fused patterns (#102394)
This diff introduces utility `find_sequential_partitions`.
This utility allows one to specify sequential pattern of
nn.Module/nn.functional and returns a list. Each item in the list contains a
List[SourcePartition] that represents sequentially connected partitions that
are of the pattern requested.
For example `find_sequential_partitions(model, [nn.Conv2d, nn.ReLU])` will find
all nn.Conv2d and nn.ReLU partitions that are sequentially connected.

Furthmore, move to using `find_sequential_partitions` for conv_bn/conv_bn_relu
for QAT.

Differential Revision: [D45948057](https://our.internmc.facebook.com/intern/diff/D45948057/)

**NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D45948057/)!
Pull Request resolved: https://github.com/pytorch/pytorch/pull/102394
Approved by: https://github.com/jerryzh168
2023-05-28 05:29:16 +00:00