Move XNNPACKQuantizer from PyTorch to ExecuTorch (#144940)

digantdesai · facebook-github-bot · commit 5a7bfc98913f · 2025-01-23T16:55:00.000-08:00
Summary: X-link: pytorch/executorch#7804 Pull Request resolved: #144940 This migrates XNNPACKQuantizer from PyTorch to ExecuTorch. Rationale: Main motivation is to avoid pytorch pin update in OSS after updating XNNPACKQuantizer, which can be rather frequent. Other impact and considerations: - PT2e flow (which lives in PyTorch) relies havily on XNNPACKQuantizer for a "example" implementation for quantizer and more importantly tests. Fow now, we will keep the torch.ao.quantization.xnnpack_quantizer as is but mark is as not BC, and deprecated to discourace future new dependencies on it. - Other OSS repository using XNNPACKQuantizer from PyTorch now have to take an additional dependency on ExecuTorch. Test Plan: CI Reviewed By: mcr229 Differential Revision: D68191752
diff --git a/torch/ao/quantization/quantizer/xnnpack_quantizer.py b/torch/ao/quantization/quantizer/xnnpack_quantizer.py
@@ -3,6 +3,7 @@
 
 import copy
 import functools
+import warnings
 from typing import Any, Callable, Optional, TYPE_CHECKING
 
 import torch
@@ -30,6 +31,7 @@
     propagate_annotation,
     QuantizationConfig,
 )
+from torch.fx._compatibility import compatibility
 
 
 if TYPE_CHECKING:
@@ -235,7 +237,15 @@ def not_module_type_or_name_filter(n: Node) -> bool:
     return not_module_type_or_name_filter
 
 
+@compatibility(is_backward_compatible=False)
 class XNNPACKQuantizer(Quantizer):
+    """
+    !!! DEPRECATED !!!
+    XNNPACKQuantizer is a marked as deprected. It will be removed in the future.
+    It has been moved to executorch.backends.xnnpack.quantizer.xnnpack_quantizer.XNNPACKQuantizer.
+    Please use the new quantizer instead.
+    """
+
     supported_config_and_operators = _get_supported_config_and_operators()
     STATIC_QAT_ONLY_OPS = [
         "conv_bn_relu",
@@ -268,6 +278,7 @@ class XNNPACKQuantizer(Quantizer):
 
     def __init__(self) -> None:
         super().__init__()
+        warnings.warn(f"{self.__class__.__name__} is deprecated!")
         self.global_config: Optional[QuantizationConfig] = None
         self.operator_type_config: dict[
             torch._ops.OpOverloadPacket, Optional[QuantizationConfig]
