Link PyQnnManagerAdaptor with quantized_ops_aot_lib.

Author: shewu-quic

backends/qualcomm/CMakeLists.txt

@@ -320,6 +320,7 @@ if(${CMAKE_SYSTEM_PROCESSOR} MATCHES "x86_64")
             qnn_backend_options
             wrappers
             qnn_executorch_logging
+            quantized_ops_aot_lib
   )
 
   pybind11_extension(PyQnnManagerAdaptor)

backends/qualcomm/aot/python/targets.bzl

@@ -26,6 +26,8 @@ def define_common_targets():
             "-DEXECUTORCH_PYTHON_MODULE_NAME={}".format(PYTHON_MODULE_NAME),
         ],
         deps = [
+            "//executorch/kernels/quantized:aot_lib",
+            "//executorch/kernels/quantized:quantized_ops_lib",
             "//executorch/runtime/core:core",
             "//executorch/backends/qualcomm/aot/python:python_lib",
             "//executorch/backends/qualcomm/aot/wrappers:wrappers",

backends/qualcomm/quantizer/quantizer.py

@@ -8,16 +8,6 @@
 from functools import partial
 from typing import Callable, Dict, List, Optional, Sequence, Set, Tuple
 
-# To support quantize op lowering in AOT
-try:
-    import executorch.kernels.quantized  # noqa[F401]
-except:
-    import logging
-
-    logging.info(
-        "Failed to load quantized_aot_lib. To run on LPAI backend, please make sure that quantized_aot_lib is accessible."
-    )
-    del logging
 import torch
 from executorch.backends.qualcomm._passes.qnn_pass_manager import QnnPassManager
 

backends/qualcomm/quantizer/targets.bzl

@@ -15,7 +15,5 @@ def define_common_targets():
         visibility = ["PUBLIC"],
         deps = [
             "//executorch/backends/transforms:decompose_sdpa",
-            "//executorch/kernels/quantized:aot_lib",
-            "//executorch/kernels/quantized:quantized_ops_lib",
         ],
     )