Update

[ghstack-poisoned]

Author: Github Executorch

backends/cadence/aot/BUCK

@@ -81,9 +81,9 @@ fbcode_target(_kind = runtime.python_library,
         "pass_utils.py",
     ],
     deps = [
-        "fbsource//third-party/pypi/beartype:beartype",
         ":utils",
         "//caffe2:torch",
+        "//executorch/backends/transforms:permute_pass_utils",
         "//executorch/exir:pass_base",
         "//executorch/exir/dialects:lib",
         "//executorch/exir/passes:lib",
@@ -188,7 +188,6 @@ fbcode_target(_kind = python_unittest,
     ],
     typing = True,
     deps = [
-        "fbsource//third-party/pypi/beartype:beartype",
         ":pass_utils",
         "//caffe2:torch",
     ],
@@ -267,6 +266,10 @@ fbcode_target(_kind = runtime.python_library,
         "//caffe2:torch",
         "//executorch/backends/cadence/aot:pass_utils",
         "//executorch/backends/cadence/aot:utils",
+        "//executorch/backends/transforms:fuse_cascaded_transpose_or_permute_ops",
+        "//executorch/backends/transforms:fuse_cascaded_view_ops",
+        "//executorch/backends/transforms:fuse_transpose_or_permute_op_pairs_pass",
+        "//executorch/backends/transforms:permute_pass_utils",
         "//executorch/exir:pass_base",
         "//executorch/exir/dialects:lib",
         "//executorch/exir/dialects/edge:lib",
@@ -304,6 +307,7 @@ fbcode_target(_kind = runtime.python_library,
         "//executorch/backends/cadence/aot:pass_utils",
         "//executorch/backends/cadence/aot:simplify_ops",
         "//executorch/backends/transforms:remove_clone_ops",
+        "//executorch/backends/transforms:remove_permutes_around_elementwise_ops",
         "//executorch/exir:pass_base",
         "//executorch/exir/dialects:lib",
         "//executorch/exir/dialects/edge:lib",
@@ -322,6 +326,7 @@ fbcode_target(_kind = runtime.python_library,
         "//executorch/backends/cadence/aot:compiler_utils",
         "//executorch/backends/cadence/aot:pass_utils",
         "//executorch/backends/cadence/aot:utils",
+        "//executorch/backends/transforms:postpone_permute_below_squeeze_view",
         "//executorch/exir:pass_base",
         "//executorch/exir:tensor",
         "//executorch/exir/dialects:lib",
@@ -343,6 +348,7 @@ fbcode_target(_kind = runtime.python_library,
         "//executorch/backends/cadence/aot:pass_utils",
         "//executorch/backends/cadence/aot:remove_ops",
         "//executorch/backends/cadence/aot:utils",
+        "//executorch/backends/transforms:replace_nop_transpose_or_permute_with_view",
         "//executorch/backends/transforms:replace_scalar_with_tensor",
         "//executorch/exir:pass_base",
         "//executorch/exir/dialects:lib",

backends/cadence/aot/functions.yaml

@@ -384,7 +384,7 @@
     - arg_meta: null
       kernel_name: impl::generic::quantized_conv2d_nchw_per_tensor_out
 
-- func: cadence::quantized_conv2d_nhwc.per_tensor_out(Tensor input, Tensor weight, Tensor bias, int[] stride, SymInt[] padding, int[] dilation, int groups, int input_zero_point, int weight_zero_point, float bias_scale, float out_scale, int out_zero_point, int out_multiplier, int out_shift, *, Tensor(a!) out) -> Tensor(a!)
+- func: cadence::quantized_conv2d_nhwc.per_tensor_out(Tensor input, Tensor weight, Tensor bias, int[] stride, SymInt[] padding, int[] dilation, int groups, int input_zero_point, int weight_zero_point, float bias_scale, float out_scale, int out_zero_point, int out_multiplier, int out_shift, Tensor? offset=None, *, Tensor(a!) out) -> Tensor(a!)
   kernels:
     - arg_meta: null
       kernel_name: impl::generic::quantized_conv2d_nhwc_per_tensor_out

backends/cadence/aot/functions_hifi.yaml

@@ -395,7 +395,7 @@
     - arg_meta: null
       kernel_name: impl::HiFi::quantized_conv2d_nchw_per_tensor_out
 
-- func: cadence::quantized_conv2d_nhwc.per_tensor_out(Tensor input, Tensor weight, Tensor bias, int[] stride, SymInt[] padding, int[] dilation, int groups, int input_zero_point, int weight_zero_point, float bias_scale, float out_scale, int out_zero_point, int out_multiplier, int out_shift, *, Tensor(a!) out) -> Tensor(a!)
+- func: cadence::quantized_conv2d_nhwc.per_tensor_out(Tensor input, Tensor weight, Tensor bias, int[] stride, SymInt[] padding, int[] dilation, int groups, int input_zero_point, int weight_zero_point, float bias_scale, float out_scale, int out_zero_point, int out_multiplier, int out_shift, Tensor? offset=None, *, Tensor(a!) out) -> Tensor(a!)
   kernels:
     - arg_meta: null
       kernel_name: impl::HiFi::quantized_conv2d_nhwc_per_tensor_out