From 5c57cd1a3c38f6a6fc86b5bc4426f75f30ae2551 Mon Sep 17 00:00:00 2001
From: cccclai <chenlai@meta.com>
Date: Thu, 25 Sep 2025 10:29:37 -0700
Subject: [PATCH] Documentation polishing related to delegates (#14551)

Summary:
1. backend-delegates-dependencies.md - fix typos
2. backend-delegates-integration.md - a bit polishing, use bullet point
instead of plain text
3. debug-backend-delegate.md - use a real model so users can copy paste,
also adapt to the API changes

Differential Revision: D83186517

(cherry picked from commit a0e1395d6da8130cf3853da311a3866b658545a3)
---
 docs/source/backend-delegates-dependencies.md |   2 +-
 docs/source/backend-delegates-integration.md  |  21 +-
 docs/source/debug-backend-delegate.md         | 603 +++++++++++++++++-
 3 files changed, 586 insertions(+), 40 deletions(-)

diff --git a/docs/source/backend-delegates-dependencies.md b/docs/source/backend-delegates-dependencies.md
index f2068989bd2..06f23ca36bc 100644
--- a/docs/source/backend-delegates-dependencies.md
+++ b/docs/source/backend-delegates-dependencies.md
@@ -49,7 +49,7 @@ for these third-party dependencies.
   `executorch/third-party` then try to use that if possible. This
   helps with reducing the binary size when the delegate is enabled.
 * The rest of the ExecuTorch code, outside of the delegate, should not depend on
-  this. And it should should build and run correctly without this dependency
+  this. And it should build and run correctly without this dependency
   when the delegate is disabled at build time.
 
 More details in the section [below](#runtime-dependencies).
diff --git a/docs/source/backend-delegates-integration.md b/docs/source/backend-delegates-integration.md
index 0179ceff872..130da0d3225 100644
--- a/docs/source/backend-delegates-integration.md
+++ b/docs/source/backend-delegates-integration.md
@@ -23,12 +23,13 @@ the top level ExecuTorch package. For third-party dependencies, please refer to
 At a minimum, a delegate must provide CMake support for building its C++
 sources.
 
-For the CMake setup, the delegate dir should be included by the
-top level `CMakeLists.txt` file using `add_subdirectory` CMake command, and
-should be built conditionally with an ExecuTorch build flag like
-`EXECUTORCH_BUILD_<DELEGATE_NAME>`, see `EXECUTORCH_BUILD_XNNPACK` for example.
-For third-party dependencies, please refer to
-[this](backend-delegates-dependencies.md).
+For the CMake setup:
+
+- The delegate directory should be included by the top-level `CMakeLists.txt` file using the `add_subdirectory` command.
+- It should be built conditionally using an ExecuTorch build flag like `EXECUTORCH_BUILD_<DELEGATE_NAME>`.
+(See `EXECUTORCH_BUILD_XNNPACK` for an example.)
+
+For third-party dependencies, please refer to [this](backend-delegates-dependencies.md).
 
 <!---
 TODO: Add more details. Need to insert a CMake layer in `executorch/backends` to
@@ -49,9 +50,7 @@ Common test types:
 
 ## Documentation
 
-A delegate must contain a `executorch/backends/<delegate_name>/README.md`
-explaining the basics of the delegate, directory structure, features, and known
-issues if any.
+A delegate must include:
 
-Any extra setup steps beyond the ones listed above should be documented in
-`executorch/backends/<delegate_name>/setup.md`
+- `executorch/backends/<delegate_name>/README.md` – covering the basics of the delegate, its directory structure, features, and any known issues.
+- `executorch/backends/<delegate_name>/setup.md` – documenting any additional setup steps beyond the ones listed above.
diff --git a/docs/source/debug-backend-delegate.md b/docs/source/debug-backend-delegate.md
index 86dddd75868..efb4653a994 100644
--- a/docs/source/debug-backend-delegate.md
+++ b/docs/source/debug-backend-delegate.md
@@ -6,60 +6,607 @@ We provide a list of util functions to give users insights on what happened to t
 The `get_delegation_info()` method provides a summary of what happened to the model after the `to_backend()` call:
 
 ```python
+import torch
+from executorch.backends.xnnpack.partition.xnnpack_partitioner import XnnpackPartitioner
+from executorch.exir import to_edge_transform_and_lower
+from torch.export import Dim, export
+from torchvision.models.mobilenetv2 import MobileNet_V2_Weights
+import torchvision.models as models
+
+# Dependency needed for debugging delegates
 from executorch.devtools.backend_debug import get_delegation_info
 from tabulate import tabulate
 
-# ... After call to to_backend(), but before to_executorch()
-graph_module = edge_manager.exported_program().graph_module
+
+model = models.mobilenetv2.mobilenet_v2(weights=MobileNet_V2_Weights.DEFAULT).eval()
+sample_inputs = (torch.randn(1, 3, 224, 224), )
+
+et_program = to_edge_transform_and_lower(
+    torch.export.export(model, sample_inputs),
+    partitioner=[XnnpackPartitioner()]
+)
+graph_module = et_program.exported_program().graph_module
 delegation_info = get_delegation_info(graph_module)
+# print the summary like the number of delegated nodes, non-delegated nodes, etc
 print(delegation_info.get_summary())
 df = delegation_info.get_operator_delegation_dataframe()
+# print the table including op_type, occurrences_in_delegated_graphs, occurrences_in_non_delegated_graphs
 print(tabulate(df, headers="keys", tablefmt="fancy_grid"))
 ```
 
 Example printout:
 ```
-Total  delegated  subgraphs:  86
-Number  of  delegated  nodes:  473
-Number  of  non-delegated  nodes:  430
+Total delegated subgraphs: 2
+Number of delegated nodes: 203
+Number of non-delegated nodes: 4
 ```
 
+|    | op_type                                           | occurrences_in_delegated_graphs | occurrences_in_non_delegated_graphs |
+|----|---------------------------------------------------|---------------------------------|-------------------------------------|
+|  0 | aten__native_batch_norm_legit_no_training_default | 52                              | 0                                   |
+|  1 | aten_add_tensor                                   | 10                              | 0                                   |
+|  2 | aten_convolution_default                          | 52                              | 0                                   |
+|  3 | aten_hardtanh_default                             | 35                              | 0                                   |
+|  4 | aten_linear_default                               | 1                               | 0                                   |
+|  5 | aten_mean_dim                                     | 1                               | 0                                   |
+|  6 | aten_view_copy_default                            | 0                               | 1                                   |
+|  7 | dim_order_ops__clone_dim_order_default            | 0                               | 1                                   |
+|  8 | getitem                                           | 52                              | 2                                   |
+|  9 | **Total**                                         | **203**                         | **4**                               |
 
-|    |  op_type                                 |  occurrences_in_delegated_graphs  |  occurrences_in_non_delegated_graphs  |
-|----|---------------------------------|------- |-----|
-|  0  |  aten__softmax_default  |  12  |  0  |
-|  1  |  aten_add_tensor  |  37  |  0  |
-|  2  |  aten_addmm_default  |  48  |  0  |
-|  3  |  aten_arange_start_step  |  0  |  25  |
-|      |  ...  |    |    |
-|  23  |  aten_view_copy_default  |  170  |  48  |
-|      |  ...  |    |    |
-|  26  |  Total  |  473  |  430  |
 
-From the table, the operator `aten_view_copy_default` appears 170 times in delegate graphs and 48 times in non-delegated graphs. Users can use information like this to debug.
+From the table, the operator `aten_view_copy_default` appears 0 times in delegate graphs and 1 times in non-delegated graphs. Users can use information like this to debug. `get_item node` is a special case, it means getting the output from the delegate subgraph.
 
 ## Visualize delegated graph
-To see a more detailed view, use the `format_delegated_graph()` method to get a str of printout of the whole graph or use `print_delegated_graph()` to print directly:
+To see a more detailed view, use the `format_delegated_graph()` method to get a string representation of the entire graph or use `print_delegated_graph()` to print directly:
 
 ```python
 from executorch.exir.backend.utils import format_delegated_graph
-graph_module = edge_manager.exported_program().graph_module
+graph_module = et_program.exported_program().graph_module
 print(format_delegated_graph(graph_module)) # or call print_delegated_graph(graph_module)
 ```
-It will print the whole model as well as the subgraph consumed by the backend. The generic debug function provided by fx like `print_tabular()` or `print_readable()` will only show `call_delegate` but hide the the subgraph consumes by the backend, while this function exposes the contents inside the subgraph.
+It will print the whole model as well as the subgraph consumed by the backend. The generic debug function provided by fx like `print_tabular()` or `print_readable()` will only show `call_delegate` and hide the subgraph consumed by the backend, while this function exposes the contents inside the subgraph.
 
-In the example printout below, observe that `embedding` and `add` operators are delegated to `XNNPACK` while the `sub` operator is not.
+In the example printout below, observe that there are two subgraphs, `aten_view_copy_default` is not delegated, while most of the others ops are delegated.
 
+<details>
 ```
-%aten_unsqueeze_copy_default_22 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.unsqueeze_copy.default](args = (%aten_arange_start_step_23, -2), kwargs = {})
-  %aten_unsqueeze_copy_default_23 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.unsqueeze_copy.default](args = (%aten_arange_start_step_24, -1), kwargs = {})
+graph():
+  %b_features_0_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_0_1_num_batches_tracked]
+  %b_features_1_conv_0_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_1_conv_0_1_num_batches_tracked]
+  %b_features_1_conv_2_num_batches_tracked : [num_users=0] = placeholder[target=b_features_1_conv_2_num_batches_tracked]
+  %b_features_2_conv_0_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_2_conv_0_1_num_batches_tracked]
+  %b_features_2_conv_1_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_2_conv_1_1_num_batches_tracked]
+  %b_features_2_conv_3_num_batches_tracked : [num_users=0] = placeholder[target=b_features_2_conv_3_num_batches_tracked]
+  %b_features_3_conv_0_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_3_conv_0_1_num_batches_tracked]
+  %b_features_3_conv_1_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_3_conv_1_1_num_batches_tracked]
+  %b_features_3_conv_3_num_batches_tracked : [num_users=0] = placeholder[target=b_features_3_conv_3_num_batches_tracked]
+  %b_features_4_conv_0_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_4_conv_0_1_num_batches_tracked]
+  %b_features_4_conv_1_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_4_conv_1_1_num_batches_tracked]
+  %b_features_4_conv_3_num_batches_tracked : [num_users=0] = placeholder[target=b_features_4_conv_3_num_batches_tracked]
+  %b_features_5_conv_0_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_5_conv_0_1_num_batches_tracked]
+  %b_features_5_conv_1_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_5_conv_1_1_num_batches_tracked]
+  %b_features_5_conv_3_num_batches_tracked : [num_users=0] = placeholder[target=b_features_5_conv_3_num_batches_tracked]
+  %b_features_6_conv_0_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_6_conv_0_1_num_batches_tracked]
+  %b_features_6_conv_1_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_6_conv_1_1_num_batches_tracked]
+  %b_features_6_conv_3_num_batches_tracked : [num_users=0] = placeholder[target=b_features_6_conv_3_num_batches_tracked]
+  %b_features_7_conv_0_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_7_conv_0_1_num_batches_tracked]
+  %b_features_7_conv_1_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_7_conv_1_1_num_batches_tracked]
+  %b_features_7_conv_3_num_batches_tracked : [num_users=0] = placeholder[target=b_features_7_conv_3_num_batches_tracked]
+  %b_features_8_conv_0_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_8_conv_0_1_num_batches_tracked]
+  %b_features_8_conv_1_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_8_conv_1_1_num_batches_tracked]
+  %b_features_8_conv_3_num_batches_tracked : [num_users=0] = placeholder[target=b_features_8_conv_3_num_batches_tracked]
+  %b_features_9_conv_0_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_9_conv_0_1_num_batches_tracked]
+  %b_features_9_conv_1_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_9_conv_1_1_num_batches_tracked]
+  %b_features_9_conv_3_num_batches_tracked : [num_users=0] = placeholder[target=b_features_9_conv_3_num_batches_tracked]
+  %b_features_10_conv_0_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_10_conv_0_1_num_batches_tracked]
+  %b_features_10_conv_1_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_10_conv_1_1_num_batches_tracked]
+  %b_features_10_conv_3_num_batches_tracked : [num_users=0] = placeholder[target=b_features_10_conv_3_num_batches_tracked]
+  %b_features_11_conv_0_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_11_conv_0_1_num_batches_tracked]
+  %b_features_11_conv_1_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_11_conv_1_1_num_batches_tracked]
+  %b_features_11_conv_3_num_batches_tracked : [num_users=0] = placeholder[target=b_features_11_conv_3_num_batches_tracked]
+  %b_features_12_conv_0_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_12_conv_0_1_num_batches_tracked]
+  %b_features_12_conv_1_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_12_conv_1_1_num_batches_tracked]
+  %b_features_12_conv_3_num_batches_tracked : [num_users=0] = placeholder[target=b_features_12_conv_3_num_batches_tracked]
+  %b_features_13_conv_0_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_13_conv_0_1_num_batches_tracked]
+  %b_features_13_conv_1_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_13_conv_1_1_num_batches_tracked]
+  %b_features_13_conv_3_num_batches_tracked : [num_users=0] = placeholder[target=b_features_13_conv_3_num_batches_tracked]
+  %b_features_14_conv_0_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_14_conv_0_1_num_batches_tracked]
+  %b_features_14_conv_1_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_14_conv_1_1_num_batches_tracked]
+  %b_features_14_conv_3_num_batches_tracked : [num_users=0] = placeholder[target=b_features_14_conv_3_num_batches_tracked]
+  %b_features_15_conv_0_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_15_conv_0_1_num_batches_tracked]
+  %b_features_15_conv_1_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_15_conv_1_1_num_batches_tracked]
+  %b_features_15_conv_3_num_batches_tracked : [num_users=0] = placeholder[target=b_features_15_conv_3_num_batches_tracked]
+  %b_features_16_conv_0_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_16_conv_0_1_num_batches_tracked]
+  %b_features_16_conv_1_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_16_conv_1_1_num_batches_tracked]
+  %b_features_16_conv_3_num_batches_tracked : [num_users=0] = placeholder[target=b_features_16_conv_3_num_batches_tracked]
+  %b_features_17_conv_0_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_17_conv_0_1_num_batches_tracked]
+  %b_features_17_conv_1_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_17_conv_1_1_num_batches_tracked]
+  %b_features_17_conv_3_num_batches_tracked : [num_users=0] = placeholder[target=b_features_17_conv_3_num_batches_tracked]
+  %b_features_18_1_num_batches_tracked : [num_users=0] = placeholder[target=b_features_18_1_num_batches_tracked]
+  %x : [num_users=1] = placeholder[target=x]
   %lowered_module_0 : [num_users=1] = get_attr[target=lowered_module_0]
     backend_id: XnnpackBackend
     lowered graph():
-      %aten_embedding_default : [num_users=1] = placeholder[target=aten_embedding_default]
-      %aten_embedding_default_1 : [num_users=1] = placeholder[target=aten_embedding_default_1]
-      %aten_add_tensor : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.add.Tensor](args = (%aten_embedding_default, %aten_embedding_default_1), kwargs = {})
-      return (aten_add_tensor,)
-  %executorch_call_delegate : [num_users=1] = call_function[target=torch.ops.higher_order.executorch_call_delegate](args = (%lowered_module_0, %aten_embedding_default, %aten_embedding_default_1), kwargs = {})
-  %aten_sub_tensor : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.sub.Tensor](args = (%aten_unsqueeze_copy_default, %aten_unsqueeze_copy_default_1), kwargs = {})
+      %p_features_0_0_weight : [num_users=1] = placeholder[target=p_features_0_0_weight]
+      %p_features_0_1_weight : [num_users=1] = placeholder[target=p_features_0_1_weight]
+      %p_features_0_1_bias : [num_users=1] = placeholder[target=p_features_0_1_bias]
+      %p_features_1_conv_0_0_weight : [num_users=1] = placeholder[target=p_features_1_conv_0_0_weight]
+      %p_features_1_conv_0_1_weight : [num_users=1] = placeholder[target=p_features_1_conv_0_1_weight]
+      %p_features_1_conv_0_1_bias : [num_users=1] = placeholder[target=p_features_1_conv_0_1_bias]
+      %p_features_1_conv_1_weight : [num_users=1] = placeholder[target=p_features_1_conv_1_weight]
+      %p_features_1_conv_2_weight : [num_users=1] = placeholder[target=p_features_1_conv_2_weight]
+      %p_features_1_conv_2_bias : [num_users=1] = placeholder[target=p_features_1_conv_2_bias]
+      %p_features_2_conv_0_0_weight : [num_users=1] = placeholder[target=p_features_2_conv_0_0_weight]
+      %p_features_2_conv_0_1_weight : [num_users=1] = placeholder[target=p_features_2_conv_0_1_weight]
+      %p_features_2_conv_0_1_bias : [num_users=1] = placeholder[target=p_features_2_conv_0_1_bias]
+      %p_features_2_conv_1_0_weight : [num_users=1] = placeholder[target=p_features_2_conv_1_0_weight]
+      %p_features_2_conv_1_1_weight : [num_users=1] = placeholder[target=p_features_2_conv_1_1_weight]
+      %p_features_2_conv_1_1_bias : [num_users=1] = placeholder[target=p_features_2_conv_1_1_bias]
+      %p_features_2_conv_2_weight : [num_users=1] = placeholder[target=p_features_2_conv_2_weight]
+      %p_features_2_conv_3_weight : [num_users=1] = placeholder[target=p_features_2_conv_3_weight]
+      %p_features_2_conv_3_bias : [num_users=1] = placeholder[target=p_features_2_conv_3_bias]
+      %p_features_3_conv_0_0_weight : [num_users=1] = placeholder[target=p_features_3_conv_0_0_weight]
+      %p_features_3_conv_0_1_weight : [num_users=1] = placeholder[target=p_features_3_conv_0_1_weight]
+      %p_features_3_conv_0_1_bias : [num_users=1] = placeholder[target=p_features_3_conv_0_1_bias]
+      %p_features_3_conv_1_0_weight : [num_users=1] = placeholder[target=p_features_3_conv_1_0_weight]
+      %p_features_3_conv_1_1_weight : [num_users=1] = placeholder[target=p_features_3_conv_1_1_weight]
+      %p_features_3_conv_1_1_bias : [num_users=1] = placeholder[target=p_features_3_conv_1_1_bias]
+      %p_features_3_conv_2_weight : [num_users=1] = placeholder[target=p_features_3_conv_2_weight]
+      %p_features_3_conv_3_weight : [num_users=1] = placeholder[target=p_features_3_conv_3_weight]
+      %p_features_3_conv_3_bias : [num_users=1] = placeholder[target=p_features_3_conv_3_bias]
+      %p_features_4_conv_0_0_weight : [num_users=1] = placeholder[target=p_features_4_conv_0_0_weight]
+      %p_features_4_conv_0_1_weight : [num_users=1] = placeholder[target=p_features_4_conv_0_1_weight]
+      %p_features_4_conv_0_1_bias : [num_users=1] = placeholder[target=p_features_4_conv_0_1_bias]
+      %p_features_4_conv_1_0_weight : [num_users=1] = placeholder[target=p_features_4_conv_1_0_weight]
+      %p_features_4_conv_1_1_weight : [num_users=1] = placeholder[target=p_features_4_conv_1_1_weight]
+      %p_features_4_conv_1_1_bias : [num_users=1] = placeholder[target=p_features_4_conv_1_1_bias]
+      %p_features_4_conv_2_weight : [num_users=1] = placeholder[target=p_features_4_conv_2_weight]
+      %p_features_4_conv_3_weight : [num_users=1] = placeholder[target=p_features_4_conv_3_weight]
+      %p_features_4_conv_3_bias : [num_users=1] = placeholder[target=p_features_4_conv_3_bias]
+      %p_features_5_conv_0_0_weight : [num_users=1] = placeholder[target=p_features_5_conv_0_0_weight]
+      %p_features_5_conv_0_1_weight : [num_users=1] = placeholder[target=p_features_5_conv_0_1_weight]
+      %p_features_5_conv_0_1_bias : [num_users=1] = placeholder[target=p_features_5_conv_0_1_bias]
+      %p_features_5_conv_1_0_weight : [num_users=1] = placeholder[target=p_features_5_conv_1_0_weight]
+      %p_features_5_conv_1_1_weight : [num_users=1] = placeholder[target=p_features_5_conv_1_1_weight]
+      %p_features_5_conv_1_1_bias : [num_users=1] = placeholder[target=p_features_5_conv_1_1_bias]
+      %p_features_5_conv_2_weight : [num_users=1] = placeholder[target=p_features_5_conv_2_weight]
+      %p_features_5_conv_3_weight : [num_users=1] = placeholder[target=p_features_5_conv_3_weight]
+      %p_features_5_conv_3_bias : [num_users=1] = placeholder[target=p_features_5_conv_3_bias]
+      %p_features_6_conv_0_0_weight : [num_users=1] = placeholder[target=p_features_6_conv_0_0_weight]
+      %p_features_6_conv_0_1_weight : [num_users=1] = placeholder[target=p_features_6_conv_0_1_weight]
+      %p_features_6_conv_0_1_bias : [num_users=1] = placeholder[target=p_features_6_conv_0_1_bias]
+      %p_features_6_conv_1_0_weight : [num_users=1] = placeholder[target=p_features_6_conv_1_0_weight]
+      %p_features_6_conv_1_1_weight : [num_users=1] = placeholder[target=p_features_6_conv_1_1_weight]
+      %p_features_6_conv_1_1_bias : [num_users=1] = placeholder[target=p_features_6_conv_1_1_bias]
+      %p_features_6_conv_2_weight : [num_users=1] = placeholder[target=p_features_6_conv_2_weight]
+      %p_features_6_conv_3_weight : [num_users=1] = placeholder[target=p_features_6_conv_3_weight]
+      %p_features_6_conv_3_bias : [num_users=1] = placeholder[target=p_features_6_conv_3_bias]
+      %p_features_7_conv_0_0_weight : [num_users=1] = placeholder[target=p_features_7_conv_0_0_weight]
+      %p_features_7_conv_0_1_weight : [num_users=1] = placeholder[target=p_features_7_conv_0_1_weight]
+      %p_features_7_conv_0_1_bias : [num_users=1] = placeholder[target=p_features_7_conv_0_1_bias]
+      %p_features_7_conv_1_0_weight : [num_users=1] = placeholder[target=p_features_7_conv_1_0_weight]
+      %p_features_7_conv_1_1_weight : [num_users=1] = placeholder[target=p_features_7_conv_1_1_weight]
+      %p_features_7_conv_1_1_bias : [num_users=1] = placeholder[target=p_features_7_conv_1_1_bias]
+      %p_features_7_conv_2_weight : [num_users=1] = placeholder[target=p_features_7_conv_2_weight]
+      %p_features_7_conv_3_weight : [num_users=1] = placeholder[target=p_features_7_conv_3_weight]
+      %p_features_7_conv_3_bias : [num_users=1] = placeholder[target=p_features_7_conv_3_bias]
+      %p_features_8_conv_0_0_weight : [num_users=1] = placeholder[target=p_features_8_conv_0_0_weight]
+      %p_features_8_conv_0_1_weight : [num_users=1] = placeholder[target=p_features_8_conv_0_1_weight]
+      %p_features_8_conv_0_1_bias : [num_users=1] = placeholder[target=p_features_8_conv_0_1_bias]
+      %p_features_8_conv_1_0_weight : [num_users=1] = placeholder[target=p_features_8_conv_1_0_weight]
+      %p_features_8_conv_1_1_weight : [num_users=1] = placeholder[target=p_features_8_conv_1_1_weight]
+      %p_features_8_conv_1_1_bias : [num_users=1] = placeholder[target=p_features_8_conv_1_1_bias]
+      %p_features_8_conv_2_weight : [num_users=1] = placeholder[target=p_features_8_conv_2_weight]
+      %p_features_8_conv_3_weight : [num_users=1] = placeholder[target=p_features_8_conv_3_weight]
+      %p_features_8_conv_3_bias : [num_users=1] = placeholder[target=p_features_8_conv_3_bias]
+      %p_features_9_conv_0_0_weight : [num_users=1] = placeholder[target=p_features_9_conv_0_0_weight]
+      %p_features_9_conv_0_1_weight : [num_users=1] = placeholder[target=p_features_9_conv_0_1_weight]
+      %p_features_9_conv_0_1_bias : [num_users=1] = placeholder[target=p_features_9_conv_0_1_bias]
+      %p_features_9_conv_1_0_weight : [num_users=1] = placeholder[target=p_features_9_conv_1_0_weight]
+      %p_features_9_conv_1_1_weight : [num_users=1] = placeholder[target=p_features_9_conv_1_1_weight]
+      %p_features_9_conv_1_1_bias : [num_users=1] = placeholder[target=p_features_9_conv_1_1_bias]
+      %p_features_9_conv_2_weight : [num_users=1] = placeholder[target=p_features_9_conv_2_weight]
+      %p_features_9_conv_3_weight : [num_users=1] = placeholder[target=p_features_9_conv_3_weight]
+      %p_features_9_conv_3_bias : [num_users=1] = placeholder[target=p_features_9_conv_3_bias]
+      %p_features_10_conv_0_0_weight : [num_users=1] = placeholder[target=p_features_10_conv_0_0_weight]
+      %p_features_10_conv_0_1_weight : [num_users=1] = placeholder[target=p_features_10_conv_0_1_weight]
+      %p_features_10_conv_0_1_bias : [num_users=1] = placeholder[target=p_features_10_conv_0_1_bias]
+      %p_features_10_conv_1_0_weight : [num_users=1] = placeholder[target=p_features_10_conv_1_0_weight]
+      %p_features_10_conv_1_1_weight : [num_users=1] = placeholder[target=p_features_10_conv_1_1_weight]
+      %p_features_10_conv_1_1_bias : [num_users=1] = placeholder[target=p_features_10_conv_1_1_bias]
+      %p_features_10_conv_2_weight : [num_users=1] = placeholder[target=p_features_10_conv_2_weight]
+      %p_features_10_conv_3_weight : [num_users=1] = placeholder[target=p_features_10_conv_3_weight]
+      %p_features_10_conv_3_bias : [num_users=1] = placeholder[target=p_features_10_conv_3_bias]
+      %p_features_11_conv_0_0_weight : [num_users=1] = placeholder[target=p_features_11_conv_0_0_weight]
+      %p_features_11_conv_0_1_weight : [num_users=1] = placeholder[target=p_features_11_conv_0_1_weight]
+      %p_features_11_conv_0_1_bias : [num_users=1] = placeholder[target=p_features_11_conv_0_1_bias]
+      %p_features_11_conv_1_0_weight : [num_users=1] = placeholder[target=p_features_11_conv_1_0_weight]
+      %p_features_11_conv_1_1_weight : [num_users=1] = placeholder[target=p_features_11_conv_1_1_weight]
+      %p_features_11_conv_1_1_bias : [num_users=1] = placeholder[target=p_features_11_conv_1_1_bias]
+      %p_features_11_conv_2_weight : [num_users=1] = placeholder[target=p_features_11_conv_2_weight]
+      %p_features_11_conv_3_weight : [num_users=1] = placeholder[target=p_features_11_conv_3_weight]
+      %p_features_11_conv_3_bias : [num_users=1] = placeholder[target=p_features_11_conv_3_bias]
+      %p_features_12_conv_0_0_weight : [num_users=1] = placeholder[target=p_features_12_conv_0_0_weight]
+      %p_features_12_conv_0_1_weight : [num_users=1] = placeholder[target=p_features_12_conv_0_1_weight]
+      %p_features_12_conv_0_1_bias : [num_users=1] = placeholder[target=p_features_12_conv_0_1_bias]
+      %p_features_12_conv_1_0_weight : [num_users=1] = placeholder[target=p_features_12_conv_1_0_weight]
+      %p_features_12_conv_1_1_weight : [num_users=1] = placeholder[target=p_features_12_conv_1_1_weight]
+      %p_features_12_conv_1_1_bias : [num_users=1] = placeholder[target=p_features_12_conv_1_1_bias]
+      %p_features_12_conv_2_weight : [num_users=1] = placeholder[target=p_features_12_conv_2_weight]
+      %p_features_12_conv_3_weight : [num_users=1] = placeholder[target=p_features_12_conv_3_weight]
+      %p_features_12_conv_3_bias : [num_users=1] = placeholder[target=p_features_12_conv_3_bias]
+      %p_features_13_conv_0_0_weight : [num_users=1] = placeholder[target=p_features_13_conv_0_0_weight]
+      %p_features_13_conv_0_1_weight : [num_users=1] = placeholder[target=p_features_13_conv_0_1_weight]
+      %p_features_13_conv_0_1_bias : [num_users=1] = placeholder[target=p_features_13_conv_0_1_bias]
+      %p_features_13_conv_1_0_weight : [num_users=1] = placeholder[target=p_features_13_conv_1_0_weight]
+      %p_features_13_conv_1_1_weight : [num_users=1] = placeholder[target=p_features_13_conv_1_1_weight]
+      %p_features_13_conv_1_1_bias : [num_users=1] = placeholder[target=p_features_13_conv_1_1_bias]
+      %p_features_13_conv_2_weight : [num_users=1] = placeholder[target=p_features_13_conv_2_weight]
+      %p_features_13_conv_3_weight : [num_users=1] = placeholder[target=p_features_13_conv_3_weight]
+      %p_features_13_conv_3_bias : [num_users=1] = placeholder[target=p_features_13_conv_3_bias]
+      %p_features_14_conv_0_0_weight : [num_users=1] = placeholder[target=p_features_14_conv_0_0_weight]
+      %p_features_14_conv_0_1_weight : [num_users=1] = placeholder[target=p_features_14_conv_0_1_weight]
+      %p_features_14_conv_0_1_bias : [num_users=1] = placeholder[target=p_features_14_conv_0_1_bias]
+      %p_features_14_conv_1_0_weight : [num_users=1] = placeholder[target=p_features_14_conv_1_0_weight]
+      %p_features_14_conv_1_1_weight : [num_users=1] = placeholder[target=p_features_14_conv_1_1_weight]
+      %p_features_14_conv_1_1_bias : [num_users=1] = placeholder[target=p_features_14_conv_1_1_bias]
+      %p_features_14_conv_2_weight : [num_users=1] = placeholder[target=p_features_14_conv_2_weight]
+      %p_features_14_conv_3_weight : [num_users=1] = placeholder[target=p_features_14_conv_3_weight]
+      %p_features_14_conv_3_bias : [num_users=1] = placeholder[target=p_features_14_conv_3_bias]
+      %p_features_15_conv_0_0_weight : [num_users=1] = placeholder[target=p_features_15_conv_0_0_weight]
+      %p_features_15_conv_0_1_weight : [num_users=1] = placeholder[target=p_features_15_conv_0_1_weight]
+      %p_features_15_conv_0_1_bias : [num_users=1] = placeholder[target=p_features_15_conv_0_1_bias]
+      %p_features_15_conv_1_0_weight : [num_users=1] = placeholder[target=p_features_15_conv_1_0_weight]
+      %p_features_15_conv_1_1_weight : [num_users=1] = placeholder[target=p_features_15_conv_1_1_weight]
+      %p_features_15_conv_1_1_bias : [num_users=1] = placeholder[target=p_features_15_conv_1_1_bias]
+      %p_features_15_conv_2_weight : [num_users=1] = placeholder[target=p_features_15_conv_2_weight]
+      %p_features_15_conv_3_weight : [num_users=1] = placeholder[target=p_features_15_conv_3_weight]
+      %p_features_15_conv_3_bias : [num_users=1] = placeholder[target=p_features_15_conv_3_bias]
+      %p_features_16_conv_0_0_weight : [num_users=1] = placeholder[target=p_features_16_conv_0_0_weight]
+      %p_features_16_conv_0_1_weight : [num_users=1] = placeholder[target=p_features_16_conv_0_1_weight]
+      %p_features_16_conv_0_1_bias : [num_users=1] = placeholder[target=p_features_16_conv_0_1_bias]
+      %p_features_16_conv_1_0_weight : [num_users=1] = placeholder[target=p_features_16_conv_1_0_weight]
+      %p_features_16_conv_1_1_weight : [num_users=1] = placeholder[target=p_features_16_conv_1_1_weight]
+      %p_features_16_conv_1_1_bias : [num_users=1] = placeholder[target=p_features_16_conv_1_1_bias]
+      %p_features_16_conv_2_weight : [num_users=1] = placeholder[target=p_features_16_conv_2_weight]
+      %p_features_16_conv_3_weight : [num_users=1] = placeholder[target=p_features_16_conv_3_weight]
+      %p_features_16_conv_3_bias : [num_users=1] = placeholder[target=p_features_16_conv_3_bias]
+      %p_features_17_conv_0_0_weight : [num_users=1] = placeholder[target=p_features_17_conv_0_0_weight]
+      %p_features_17_conv_0_1_weight : [num_users=1] = placeholder[target=p_features_17_conv_0_1_weight]
+      %p_features_17_conv_0_1_bias : [num_users=1] = placeholder[target=p_features_17_conv_0_1_bias]
+      %p_features_17_conv_1_0_weight : [num_users=1] = placeholder[target=p_features_17_conv_1_0_weight]
+      %p_features_17_conv_1_1_weight : [num_users=1] = placeholder[target=p_features_17_conv_1_1_weight]
+      %p_features_17_conv_1_1_bias : [num_users=1] = placeholder[target=p_features_17_conv_1_1_bias]
+      %p_features_17_conv_2_weight : [num_users=1] = placeholder[target=p_features_17_conv_2_weight]
+      %p_features_17_conv_3_weight : [num_users=1] = placeholder[target=p_features_17_conv_3_weight]
+      %p_features_17_conv_3_bias : [num_users=1] = placeholder[target=p_features_17_conv_3_bias]
+      %p_features_18_0_weight : [num_users=1] = placeholder[target=p_features_18_0_weight]
+      %p_features_18_1_weight : [num_users=1] = placeholder[target=p_features_18_1_weight]
+      %p_features_18_1_bias : [num_users=1] = placeholder[target=p_features_18_1_bias]
+      %b_features_0_1_running_mean : [num_users=1] = placeholder[target=b_features_0_1_running_mean]
+      %b_features_0_1_running_var : [num_users=1] = placeholder[target=b_features_0_1_running_var]
+      %b_features_1_conv_0_1_running_mean : [num_users=1] = placeholder[target=b_features_1_conv_0_1_running_mean]
+      %b_features_1_conv_0_1_running_var : [num_users=1] = placeholder[target=b_features_1_conv_0_1_running_var]
+      %b_features_1_conv_2_running_mean : [num_users=1] = placeholder[target=b_features_1_conv_2_running_mean]
+      %b_features_1_conv_2_running_var : [num_users=1] = placeholder[target=b_features_1_conv_2_running_var]
+      %b_features_2_conv_0_1_running_mean : [num_users=1] = placeholder[target=b_features_2_conv_0_1_running_mean]
+      %b_features_2_conv_0_1_running_var : [num_users=1] = placeholder[target=b_features_2_conv_0_1_running_var]
+      %b_features_2_conv_1_1_running_mean : [num_users=1] = placeholder[target=b_features_2_conv_1_1_running_mean]
+      %b_features_2_conv_1_1_running_var : [num_users=1] = placeholder[target=b_features_2_conv_1_1_running_var]
+      %b_features_2_conv_3_running_mean : [num_users=1] = placeholder[target=b_features_2_conv_3_running_mean]
+      %b_features_2_conv_3_running_var : [num_users=1] = placeholder[target=b_features_2_conv_3_running_var]
+      %b_features_3_conv_0_1_running_mean : [num_users=1] = placeholder[target=b_features_3_conv_0_1_running_mean]
+      %b_features_3_conv_0_1_running_var : [num_users=1] = placeholder[target=b_features_3_conv_0_1_running_var]
+      %b_features_3_conv_1_1_running_mean : [num_users=1] = placeholder[target=b_features_3_conv_1_1_running_mean]
+      %b_features_3_conv_1_1_running_var : [num_users=1] = placeholder[target=b_features_3_conv_1_1_running_var]
+      %b_features_3_conv_3_running_mean : [num_users=1] = placeholder[target=b_features_3_conv_3_running_mean]
+      %b_features_3_conv_3_running_var : [num_users=1] = placeholder[target=b_features_3_conv_3_running_var]
+      %b_features_4_conv_0_1_running_mean : [num_users=1] = placeholder[target=b_features_4_conv_0_1_running_mean]
+      %b_features_4_conv_0_1_running_var : [num_users=1] = placeholder[target=b_features_4_conv_0_1_running_var]
+      %b_features_4_conv_1_1_running_mean : [num_users=1] = placeholder[target=b_features_4_conv_1_1_running_mean]
+      %b_features_4_conv_1_1_running_var : [num_users=1] = placeholder[target=b_features_4_conv_1_1_running_var]
+      %b_features_4_conv_3_running_mean : [num_users=1] = placeholder[target=b_features_4_conv_3_running_mean]
+      %b_features_4_conv_3_running_var : [num_users=1] = placeholder[target=b_features_4_conv_3_running_var]
+      %b_features_5_conv_0_1_running_mean : [num_users=1] = placeholder[target=b_features_5_conv_0_1_running_mean]
+      %b_features_5_conv_0_1_running_var : [num_users=1] = placeholder[target=b_features_5_conv_0_1_running_var]
+      %b_features_5_conv_1_1_running_mean : [num_users=1] = placeholder[target=b_features_5_conv_1_1_running_mean]
+      %b_features_5_conv_1_1_running_var : [num_users=1] = placeholder[target=b_features_5_conv_1_1_running_var]
+      %b_features_5_conv_3_running_mean : [num_users=1] = placeholder[target=b_features_5_conv_3_running_mean]
+      %b_features_5_conv_3_running_var : [num_users=1] = placeholder[target=b_features_5_conv_3_running_var]
+      %b_features_6_conv_0_1_running_mean : [num_users=1] = placeholder[target=b_features_6_conv_0_1_running_mean]
+      %b_features_6_conv_0_1_running_var : [num_users=1] = placeholder[target=b_features_6_conv_0_1_running_var]
+      %b_features_6_conv_1_1_running_mean : [num_users=1] = placeholder[target=b_features_6_conv_1_1_running_mean]
+      %b_features_6_conv_1_1_running_var : [num_users=1] = placeholder[target=b_features_6_conv_1_1_running_var]
+      %b_features_6_conv_3_running_mean : [num_users=1] = placeholder[target=b_features_6_conv_3_running_mean]
+      %b_features_6_conv_3_running_var : [num_users=1] = placeholder[target=b_features_6_conv_3_running_var]
+      %b_features_7_conv_0_1_running_mean : [num_users=1] = placeholder[target=b_features_7_conv_0_1_running_mean]
+      %b_features_7_conv_0_1_running_var : [num_users=1] = placeholder[target=b_features_7_conv_0_1_running_var]
+      %b_features_7_conv_1_1_running_mean : [num_users=1] = placeholder[target=b_features_7_conv_1_1_running_mean]
+      %b_features_7_conv_1_1_running_var : [num_users=1] = placeholder[target=b_features_7_conv_1_1_running_var]
+      %b_features_7_conv_3_running_mean : [num_users=1] = placeholder[target=b_features_7_conv_3_running_mean]
+      %b_features_7_conv_3_running_var : [num_users=1] = placeholder[target=b_features_7_conv_3_running_var]
+      %b_features_8_conv_0_1_running_mean : [num_users=1] = placeholder[target=b_features_8_conv_0_1_running_mean]
+      %b_features_8_conv_0_1_running_var : [num_users=1] = placeholder[target=b_features_8_conv_0_1_running_var]
+      %b_features_8_conv_1_1_running_mean : [num_users=1] = placeholder[target=b_features_8_conv_1_1_running_mean]
+      %b_features_8_conv_1_1_running_var : [num_users=1] = placeholder[target=b_features_8_conv_1_1_running_var]
+      %b_features_8_conv_3_running_mean : [num_users=1] = placeholder[target=b_features_8_conv_3_running_mean]
+      %b_features_8_conv_3_running_var : [num_users=1] = placeholder[target=b_features_8_conv_3_running_var]
+      %b_features_9_conv_0_1_running_mean : [num_users=1] = placeholder[target=b_features_9_conv_0_1_running_mean]
+      %b_features_9_conv_0_1_running_var : [num_users=1] = placeholder[target=b_features_9_conv_0_1_running_var]
+      %b_features_9_conv_1_1_running_mean : [num_users=1] = placeholder[target=b_features_9_conv_1_1_running_mean]
+      %b_features_9_conv_1_1_running_var : [num_users=1] = placeholder[target=b_features_9_conv_1_1_running_var]
+      %b_features_9_conv_3_running_mean : [num_users=1] = placeholder[target=b_features_9_conv_3_running_mean]
+      %b_features_9_conv_3_running_var : [num_users=1] = placeholder[target=b_features_9_conv_3_running_var]
+      %b_features_10_conv_0_1_running_mean : [num_users=1] = placeholder[target=b_features_10_conv_0_1_running_mean]
+      %b_features_10_conv_0_1_running_var : [num_users=1] = placeholder[target=b_features_10_conv_0_1_running_var]
+      %b_features_10_conv_1_1_running_mean : [num_users=1] = placeholder[target=b_features_10_conv_1_1_running_mean]
+      %b_features_10_conv_1_1_running_var : [num_users=1] = placeholder[target=b_features_10_conv_1_1_running_var]
+      %b_features_10_conv_3_running_mean : [num_users=1] = placeholder[target=b_features_10_conv_3_running_mean]
+      %b_features_10_conv_3_running_var : [num_users=1] = placeholder[target=b_features_10_conv_3_running_var]
+      %b_features_11_conv_0_1_running_mean : [num_users=1] = placeholder[target=b_features_11_conv_0_1_running_mean]
+      %b_features_11_conv_0_1_running_var : [num_users=1] = placeholder[target=b_features_11_conv_0_1_running_var]
+      %b_features_11_conv_1_1_running_mean : [num_users=1] = placeholder[target=b_features_11_conv_1_1_running_mean]
+      %b_features_11_conv_1_1_running_var : [num_users=1] = placeholder[target=b_features_11_conv_1_1_running_var]
+      %b_features_11_conv_3_running_mean : [num_users=1] = placeholder[target=b_features_11_conv_3_running_mean]
+      %b_features_11_conv_3_running_var : [num_users=1] = placeholder[target=b_features_11_conv_3_running_var]
+      %b_features_12_conv_0_1_running_mean : [num_users=1] = placeholder[target=b_features_12_conv_0_1_running_mean]
+      %b_features_12_conv_0_1_running_var : [num_users=1] = placeholder[target=b_features_12_conv_0_1_running_var]
+      %b_features_12_conv_1_1_running_mean : [num_users=1] = placeholder[target=b_features_12_conv_1_1_running_mean]
+      %b_features_12_conv_1_1_running_var : [num_users=1] = placeholder[target=b_features_12_conv_1_1_running_var]
+      %b_features_12_conv_3_running_mean : [num_users=1] = placeholder[target=b_features_12_conv_3_running_mean]
+      %b_features_12_conv_3_running_var : [num_users=1] = placeholder[target=b_features_12_conv_3_running_var]
+      %b_features_13_conv_0_1_running_mean : [num_users=1] = placeholder[target=b_features_13_conv_0_1_running_mean]
+      %b_features_13_conv_0_1_running_var : [num_users=1] = placeholder[target=b_features_13_conv_0_1_running_var]
+      %b_features_13_conv_1_1_running_mean : [num_users=1] = placeholder[target=b_features_13_conv_1_1_running_mean]
+      %b_features_13_conv_1_1_running_var : [num_users=1] = placeholder[target=b_features_13_conv_1_1_running_var]
+      %b_features_13_conv_3_running_mean : [num_users=1] = placeholder[target=b_features_13_conv_3_running_mean]
+      %b_features_13_conv_3_running_var : [num_users=1] = placeholder[target=b_features_13_conv_3_running_var]
+      %b_features_14_conv_0_1_running_mean : [num_users=1] = placeholder[target=b_features_14_conv_0_1_running_mean]
+      %b_features_14_conv_0_1_running_var : [num_users=1] = placeholder[target=b_features_14_conv_0_1_running_var]
+      %b_features_14_conv_1_1_running_mean : [num_users=1] = placeholder[target=b_features_14_conv_1_1_running_mean]
+      %b_features_14_conv_1_1_running_var : [num_users=1] = placeholder[target=b_features_14_conv_1_1_running_var]
+      %b_features_14_conv_3_running_mean : [num_users=1] = placeholder[target=b_features_14_conv_3_running_mean]
+      %b_features_14_conv_3_running_var : [num_users=1] = placeholder[target=b_features_14_conv_3_running_var]
+      %b_features_15_conv_0_1_running_mean : [num_users=1] = placeholder[target=b_features_15_conv_0_1_running_mean]
+      %b_features_15_conv_0_1_running_var : [num_users=1] = placeholder[target=b_features_15_conv_0_1_running_var]
+      %b_features_15_conv_1_1_running_mean : [num_users=1] = placeholder[target=b_features_15_conv_1_1_running_mean]
+      %b_features_15_conv_1_1_running_var : [num_users=1] = placeholder[target=b_features_15_conv_1_1_running_var]
+      %b_features_15_conv_3_running_mean : [num_users=1] = placeholder[target=b_features_15_conv_3_running_mean]
+      %b_features_15_conv_3_running_var : [num_users=1] = placeholder[target=b_features_15_conv_3_running_var]
+      %b_features_16_conv_0_1_running_mean : [num_users=1] = placeholder[target=b_features_16_conv_0_1_running_mean]
+      %b_features_16_conv_0_1_running_var : [num_users=1] = placeholder[target=b_features_16_conv_0_1_running_var]
+      %b_features_16_conv_1_1_running_mean : [num_users=1] = placeholder[target=b_features_16_conv_1_1_running_mean]
+      %b_features_16_conv_1_1_running_var : [num_users=1] = placeholder[target=b_features_16_conv_1_1_running_var]
+      %b_features_16_conv_3_running_mean : [num_users=1] = placeholder[target=b_features_16_conv_3_running_mean]
+      %b_features_16_conv_3_running_var : [num_users=1] = placeholder[target=b_features_16_conv_3_running_var]
+      %b_features_17_conv_0_1_running_mean : [num_users=1] = placeholder[target=b_features_17_conv_0_1_running_mean]
+      %b_features_17_conv_0_1_running_var : [num_users=1] = placeholder[target=b_features_17_conv_0_1_running_var]
+      %b_features_17_conv_1_1_running_mean : [num_users=1] = placeholder[target=b_features_17_conv_1_1_running_mean]
+      %b_features_17_conv_1_1_running_var : [num_users=1] = placeholder[target=b_features_17_conv_1_1_running_var]
+      %b_features_17_conv_3_running_mean : [num_users=1] = placeholder[target=b_features_17_conv_3_running_mean]
+      %b_features_17_conv_3_running_var : [num_users=1] = placeholder[target=b_features_17_conv_3_running_var]
+      %b_features_18_1_running_mean : [num_users=1] = placeholder[target=b_features_18_1_running_mean]
+      %b_features_18_1_running_var : [num_users=1] = placeholder[target=b_features_18_1_running_var]
+      %x : [num_users=1] = placeholder[target=x]
+      %aten_convolution_default : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%x, %p_features_0_0_weight, None, [2, 2], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default, %p_features_0_1_weight, %p_features_0_1_bias, %b_features_0_1_running_mean, %b_features_0_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default, 0), kwargs = {})
+      %aten_hardtanh_default : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_1 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default, %p_features_1_conv_0_0_weight, None, [1, 1], [1, 1], [1, 1], False, [0, 0], 32), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_1 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_1, %p_features_1_conv_0_1_weight, %p_features_1_conv_0_1_bias, %b_features_1_conv_0_1_running_mean, %b_features_1_conv_0_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_1, 0), kwargs = {})
+      %aten_hardtanh_default_1 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_1, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_2 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_1, %p_features_1_conv_1_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_2 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_2, %p_features_1_conv_2_weight, %p_features_1_conv_2_bias, %b_features_1_conv_2_running_mean, %b_features_1_conv_2_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_2 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_2, 0), kwargs = {})
+      %aten_convolution_default_3 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%getitem_2, %p_features_2_conv_0_0_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_3 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_3, %p_features_2_conv_0_1_weight, %p_features_2_conv_0_1_bias, %b_features_2_conv_0_1_running_mean, %b_features_2_conv_0_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_3 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_3, 0), kwargs = {})
+      %aten_hardtanh_default_2 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_3, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_4 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_2, %p_features_2_conv_1_0_weight, None, [2, 2], [1, 1], [1, 1], False, [0, 0], 96), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_4 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_4, %p_features_2_conv_1_1_weight, %p_features_2_conv_1_1_bias, %b_features_2_conv_1_1_running_mean, %b_features_2_conv_1_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_4 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_4, 0), kwargs = {})
+      %aten_hardtanh_default_3 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_4, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_5 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_3, %p_features_2_conv_2_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_5 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_5, %p_features_2_conv_3_weight, %p_features_2_conv_3_bias, %b_features_2_conv_3_running_mean, %b_features_2_conv_3_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_5 : [num_users=2] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_5, 0), kwargs = {})
+      %aten_convolution_default_6 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%getitem_5, %p_features_3_conv_0_0_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_6 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_6, %p_features_3_conv_0_1_weight, %p_features_3_conv_0_1_bias, %b_features_3_conv_0_1_running_mean, %b_features_3_conv_0_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_6 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_6, 0), kwargs = {})
+      %aten_hardtanh_default_4 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_6, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_7 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_4, %p_features_3_conv_1_0_weight, None, [1, 1], [1, 1], [1, 1], False, [0, 0], 144), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_7 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_7, %p_features_3_conv_1_1_weight, %p_features_3_conv_1_1_bias, %b_features_3_conv_1_1_running_mean, %b_features_3_conv_1_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_7 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_7, 0), kwargs = {})
+      %aten_hardtanh_default_5 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_7, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_8 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_5, %p_features_3_conv_2_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_8 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_8, %p_features_3_conv_3_weight, %p_features_3_conv_3_bias, %b_features_3_conv_3_running_mean, %b_features_3_conv_3_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_8 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_8, 0), kwargs = {})
+      %aten_add_tensor : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.add.Tensor](args = (%getitem_5, %getitem_8), kwargs = {})
+      %aten_convolution_default_9 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_add_tensor, %p_features_4_conv_0_0_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_9 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_9, %p_features_4_conv_0_1_weight, %p_features_4_conv_0_1_bias, %b_features_4_conv_0_1_running_mean, %b_features_4_conv_0_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_9 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_9, 0), kwargs = {})
+      %aten_hardtanh_default_6 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_9, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_10 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_6, %p_features_4_conv_1_0_weight, None, [2, 2], [1, 1], [1, 1], False, [0, 0], 144), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_10 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_10, %p_features_4_conv_1_1_weight, %p_features_4_conv_1_1_bias, %b_features_4_conv_1_1_running_mean, %b_features_4_conv_1_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_10 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_10, 0), kwargs = {})
+      %aten_hardtanh_default_7 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_10, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_11 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_7, %p_features_4_conv_2_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_11 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_11, %p_features_4_conv_3_weight, %p_features_4_conv_3_bias, %b_features_4_conv_3_running_mean, %b_features_4_conv_3_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_11 : [num_users=2] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_11, 0), kwargs = {})
+      %aten_convolution_default_12 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%getitem_11, %p_features_5_conv_0_0_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_12 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_12, %p_features_5_conv_0_1_weight, %p_features_5_conv_0_1_bias, %b_features_5_conv_0_1_running_mean, %b_features_5_conv_0_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_12 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_12, 0), kwargs = {})
+      %aten_hardtanh_default_8 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_12, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_13 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_8, %p_features_5_conv_1_0_weight, None, [1, 1], [1, 1], [1, 1], False, [0, 0], 192), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_13 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_13, %p_features_5_conv_1_1_weight, %p_features_5_conv_1_1_bias, %b_features_5_conv_1_1_running_mean, %b_features_5_conv_1_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_13 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_13, 0), kwargs = {})
+      %aten_hardtanh_default_9 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_13, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_14 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_9, %p_features_5_conv_2_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_14 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_14, %p_features_5_conv_3_weight, %p_features_5_conv_3_bias, %b_features_5_conv_3_running_mean, %b_features_5_conv_3_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_14 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_14, 0), kwargs = {})
+      %aten_add_tensor_1 : [num_users=2] = call_function[target=executorch.exir.dialects.edge._ops.aten.add.Tensor](args = (%getitem_11, %getitem_14), kwargs = {})
+      %aten_convolution_default_15 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_add_tensor_1, %p_features_6_conv_0_0_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_15 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_15, %p_features_6_conv_0_1_weight, %p_features_6_conv_0_1_bias, %b_features_6_conv_0_1_running_mean, %b_features_6_conv_0_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_15 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_15, 0), kwargs = {})
+      %aten_hardtanh_default_10 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_15, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_16 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_10, %p_features_6_conv_1_0_weight, None, [1, 1], [1, 1], [1, 1], False, [0, 0], 192), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_16 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_16, %p_features_6_conv_1_1_weight, %p_features_6_conv_1_1_bias, %b_features_6_conv_1_1_running_mean, %b_features_6_conv_1_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_16 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_16, 0), kwargs = {})
+      %aten_hardtanh_default_11 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_16, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_17 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_11, %p_features_6_conv_2_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_17 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_17, %p_features_6_conv_3_weight, %p_features_6_conv_3_bias, %b_features_6_conv_3_running_mean, %b_features_6_conv_3_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_17 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_17, 0), kwargs = {})
+      %aten_add_tensor_2 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.add.Tensor](args = (%aten_add_tensor_1, %getitem_17), kwargs = {})
+      %aten_convolution_default_18 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_add_tensor_2, %p_features_7_conv_0_0_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_18 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_18, %p_features_7_conv_0_1_weight, %p_features_7_conv_0_1_bias, %b_features_7_conv_0_1_running_mean, %b_features_7_conv_0_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_18 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_18, 0), kwargs = {})
+      %aten_hardtanh_default_12 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_18, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_19 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_12, %p_features_7_conv_1_0_weight, None, [2, 2], [1, 1], [1, 1], False, [0, 0], 192), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_19 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_19, %p_features_7_conv_1_1_weight, %p_features_7_conv_1_1_bias, %b_features_7_conv_1_1_running_mean, %b_features_7_conv_1_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_19 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_19, 0), kwargs = {})
+      %aten_hardtanh_default_13 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_19, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_20 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_13, %p_features_7_conv_2_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_20 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_20, %p_features_7_conv_3_weight, %p_features_7_conv_3_bias, %b_features_7_conv_3_running_mean, %b_features_7_conv_3_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_20 : [num_users=2] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_20, 0), kwargs = {})
+      %aten_convolution_default_21 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%getitem_20, %p_features_8_conv_0_0_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_21 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_21, %p_features_8_conv_0_1_weight, %p_features_8_conv_0_1_bias, %b_features_8_conv_0_1_running_mean, %b_features_8_conv_0_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_21 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_21, 0), kwargs = {})
+      %aten_hardtanh_default_14 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_21, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_22 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_14, %p_features_8_conv_1_0_weight, None, [1, 1], [1, 1], [1, 1], False, [0, 0], 384), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_22 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_22, %p_features_8_conv_1_1_weight, %p_features_8_conv_1_1_bias, %b_features_8_conv_1_1_running_mean, %b_features_8_conv_1_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_22 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_22, 0), kwargs = {})
+      %aten_hardtanh_default_15 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_22, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_23 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_15, %p_features_8_conv_2_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_23 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_23, %p_features_8_conv_3_weight, %p_features_8_conv_3_bias, %b_features_8_conv_3_running_mean, %b_features_8_conv_3_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_23 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_23, 0), kwargs = {})
+      %aten_add_tensor_3 : [num_users=2] = call_function[target=executorch.exir.dialects.edge._ops.aten.add.Tensor](args = (%getitem_20, %getitem_23), kwargs = {})
+      %aten_convolution_default_24 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_add_tensor_3, %p_features_9_conv_0_0_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_24 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_24, %p_features_9_conv_0_1_weight, %p_features_9_conv_0_1_bias, %b_features_9_conv_0_1_running_mean, %b_features_9_conv_0_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_24 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_24, 0), kwargs = {})
+      %aten_hardtanh_default_16 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_24, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_25 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_16, %p_features_9_conv_1_0_weight, None, [1, 1], [1, 1], [1, 1], False, [0, 0], 384), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_25 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_25, %p_features_9_conv_1_1_weight, %p_features_9_conv_1_1_bias, %b_features_9_conv_1_1_running_mean, %b_features_9_conv_1_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_25 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_25, 0), kwargs = {})
+      %aten_hardtanh_default_17 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_25, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_26 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_17, %p_features_9_conv_2_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_26 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_26, %p_features_9_conv_3_weight, %p_features_9_conv_3_bias, %b_features_9_conv_3_running_mean, %b_features_9_conv_3_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_26 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_26, 0), kwargs = {})
+      %aten_add_tensor_4 : [num_users=2] = call_function[target=executorch.exir.dialects.edge._ops.aten.add.Tensor](args = (%aten_add_tensor_3, %getitem_26), kwargs = {})
+      %aten_convolution_default_27 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_add_tensor_4, %p_features_10_conv_0_0_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_27 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_27, %p_features_10_conv_0_1_weight, %p_features_10_conv_0_1_bias, %b_features_10_conv_0_1_running_mean, %b_features_10_conv_0_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_27 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_27, 0), kwargs = {})
+      %aten_hardtanh_default_18 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_27, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_28 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_18, %p_features_10_conv_1_0_weight, None, [1, 1], [1, 1], [1, 1], False, [0, 0], 384), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_28 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_28, %p_features_10_conv_1_1_weight, %p_features_10_conv_1_1_bias, %b_features_10_conv_1_1_running_mean, %b_features_10_conv_1_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_28 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_28, 0), kwargs = {})
+      %aten_hardtanh_default_19 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_28, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_29 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_19, %p_features_10_conv_2_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_29 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_29, %p_features_10_conv_3_weight, %p_features_10_conv_3_bias, %b_features_10_conv_3_running_mean, %b_features_10_conv_3_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_29 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_29, 0), kwargs = {})
+      %aten_add_tensor_5 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.add.Tensor](args = (%aten_add_tensor_4, %getitem_29), kwargs = {})
+      %aten_convolution_default_30 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_add_tensor_5, %p_features_11_conv_0_0_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_30 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_30, %p_features_11_conv_0_1_weight, %p_features_11_conv_0_1_bias, %b_features_11_conv_0_1_running_mean, %b_features_11_conv_0_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_30 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_30, 0), kwargs = {})
+      %aten_hardtanh_default_20 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_30, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_31 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_20, %p_features_11_conv_1_0_weight, None, [1, 1], [1, 1], [1, 1], False, [0, 0], 384), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_31 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_31, %p_features_11_conv_1_1_weight, %p_features_11_conv_1_1_bias, %b_features_11_conv_1_1_running_mean, %b_features_11_conv_1_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_31 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_31, 0), kwargs = {})
+      %aten_hardtanh_default_21 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_31, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_32 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_21, %p_features_11_conv_2_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_32 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_32, %p_features_11_conv_3_weight, %p_features_11_conv_3_bias, %b_features_11_conv_3_running_mean, %b_features_11_conv_3_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_32 : [num_users=2] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_32, 0), kwargs = {})
+      %aten_convolution_default_33 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%getitem_32, %p_features_12_conv_0_0_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_33 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_33, %p_features_12_conv_0_1_weight, %p_features_12_conv_0_1_bias, %b_features_12_conv_0_1_running_mean, %b_features_12_conv_0_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_33 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_33, 0), kwargs = {})
+      %aten_hardtanh_default_22 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_33, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_34 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_22, %p_features_12_conv_1_0_weight, None, [1, 1], [1, 1], [1, 1], False, [0, 0], 576), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_34 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_34, %p_features_12_conv_1_1_weight, %p_features_12_conv_1_1_bias, %b_features_12_conv_1_1_running_mean, %b_features_12_conv_1_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_34 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_34, 0), kwargs = {})
+      %aten_hardtanh_default_23 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_34, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_35 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_23, %p_features_12_conv_2_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_35 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_35, %p_features_12_conv_3_weight, %p_features_12_conv_3_bias, %b_features_12_conv_3_running_mean, %b_features_12_conv_3_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_35 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_35, 0), kwargs = {})
+      %aten_add_tensor_6 : [num_users=2] = call_function[target=executorch.exir.dialects.edge._ops.aten.add.Tensor](args = (%getitem_32, %getitem_35), kwargs = {})
+      %aten_convolution_default_36 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_add_tensor_6, %p_features_13_conv_0_0_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_36 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_36, %p_features_13_conv_0_1_weight, %p_features_13_conv_0_1_bias, %b_features_13_conv_0_1_running_mean, %b_features_13_conv_0_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_36 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_36, 0), kwargs = {})
+      %aten_hardtanh_default_24 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_36, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_37 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_24, %p_features_13_conv_1_0_weight, None, [1, 1], [1, 1], [1, 1], False, [0, 0], 576), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_37 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_37, %p_features_13_conv_1_1_weight, %p_features_13_conv_1_1_bias, %b_features_13_conv_1_1_running_mean, %b_features_13_conv_1_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_37 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_37, 0), kwargs = {})
+      %aten_hardtanh_default_25 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_37, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_38 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_25, %p_features_13_conv_2_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_38 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_38, %p_features_13_conv_3_weight, %p_features_13_conv_3_bias, %b_features_13_conv_3_running_mean, %b_features_13_conv_3_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_38 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_38, 0), kwargs = {})
+      %aten_add_tensor_7 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.add.Tensor](args = (%aten_add_tensor_6, %getitem_38), kwargs = {})
+      %aten_convolution_default_39 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_add_tensor_7, %p_features_14_conv_0_0_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_39 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_39, %p_features_14_conv_0_1_weight, %p_features_14_conv_0_1_bias, %b_features_14_conv_0_1_running_mean, %b_features_14_conv_0_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_39 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_39, 0), kwargs = {})
+      %aten_hardtanh_default_26 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_39, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_40 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_26, %p_features_14_conv_1_0_weight, None, [2, 2], [1, 1], [1, 1], False, [0, 0], 576), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_40 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_40, %p_features_14_conv_1_1_weight, %p_features_14_conv_1_1_bias, %b_features_14_conv_1_1_running_mean, %b_features_14_conv_1_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_40 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_40, 0), kwargs = {})
+      %aten_hardtanh_default_27 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_40, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_41 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_27, %p_features_14_conv_2_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_41 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_41, %p_features_14_conv_3_weight, %p_features_14_conv_3_bias, %b_features_14_conv_3_running_mean, %b_features_14_conv_3_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_41 : [num_users=2] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_41, 0), kwargs = {})
+      %aten_convolution_default_42 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%getitem_41, %p_features_15_conv_0_0_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_42 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_42, %p_features_15_conv_0_1_weight, %p_features_15_conv_0_1_bias, %b_features_15_conv_0_1_running_mean, %b_features_15_conv_0_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_42 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_42, 0), kwargs = {})
+      %aten_hardtanh_default_28 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_42, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_43 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_28, %p_features_15_conv_1_0_weight, None, [1, 1], [1, 1], [1, 1], False, [0, 0], 960), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_43 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_43, %p_features_15_conv_1_1_weight, %p_features_15_conv_1_1_bias, %b_features_15_conv_1_1_running_mean, %b_features_15_conv_1_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_43 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_43, 0), kwargs = {})
+      %aten_hardtanh_default_29 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_43, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_44 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_29, %p_features_15_conv_2_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_44 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_44, %p_features_15_conv_3_weight, %p_features_15_conv_3_bias, %b_features_15_conv_3_running_mean, %b_features_15_conv_3_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_44 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_44, 0), kwargs = {})
+      %aten_add_tensor_8 : [num_users=2] = call_function[target=executorch.exir.dialects.edge._ops.aten.add.Tensor](args = (%getitem_41, %getitem_44), kwargs = {})
+      %aten_convolution_default_45 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_add_tensor_8, %p_features_16_conv_0_0_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_45 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_45, %p_features_16_conv_0_1_weight, %p_features_16_conv_0_1_bias, %b_features_16_conv_0_1_running_mean, %b_features_16_conv_0_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_45 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_45, 0), kwargs = {})
+      %aten_hardtanh_default_30 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_45, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_46 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_30, %p_features_16_conv_1_0_weight, None, [1, 1], [1, 1], [1, 1], False, [0, 0], 960), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_46 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_46, %p_features_16_conv_1_1_weight, %p_features_16_conv_1_1_bias, %b_features_16_conv_1_1_running_mean, %b_features_16_conv_1_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_46 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_46, 0), kwargs = {})
+      %aten_hardtanh_default_31 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_46, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_47 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_31, %p_features_16_conv_2_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_47 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_47, %p_features_16_conv_3_weight, %p_features_16_conv_3_bias, %b_features_16_conv_3_running_mean, %b_features_16_conv_3_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_47 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_47, 0), kwargs = {})
+      %aten_add_tensor_9 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.add.Tensor](args = (%aten_add_tensor_8, %getitem_47), kwargs = {})
+      %aten_convolution_default_48 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_add_tensor_9, %p_features_17_conv_0_0_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_48 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_48, %p_features_17_conv_0_1_weight, %p_features_17_conv_0_1_bias, %b_features_17_conv_0_1_running_mean, %b_features_17_conv_0_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_48 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_48, 0), kwargs = {})
+      %aten_hardtanh_default_32 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_48, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_49 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_32, %p_features_17_conv_1_0_weight, None, [1, 1], [1, 1], [1, 1], False, [0, 0], 960), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_49 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_49, %p_features_17_conv_1_1_weight, %p_features_17_conv_1_1_bias, %b_features_17_conv_1_1_running_mean, %b_features_17_conv_1_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_49 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_49, 0), kwargs = {})
+      %aten_hardtanh_default_33 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_49, 0.0, 6.0), kwargs = {})
+      %aten_convolution_default_50 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%aten_hardtanh_default_33, %p_features_17_conv_2_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_50 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_50, %p_features_17_conv_3_weight, %p_features_17_conv_3_bias, %b_features_17_conv_3_running_mean, %b_features_17_conv_3_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_50 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_50, 0), kwargs = {})
+      %aten_convolution_default_51 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.convolution.default](args = (%getitem_50, %p_features_18_0_weight, None, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
+      %aten__native_batch_norm_legit_no_training_default_51 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten._native_batch_norm_legit_no_training.default](args = (%aten_convolution_default_51, %p_features_18_1_weight, %p_features_18_1_bias, %b_features_18_1_running_mean, %b_features_18_1_running_var, 0.1, 1e-05), kwargs = {})
+      %getitem_51 : [num_users=1] = call_function[target=operator.getitem](args = (%aten__native_batch_norm_legit_no_training_default_51, 0), kwargs = {})
+      %aten_hardtanh_default_34 : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.hardtanh.default](args = (%getitem_51, 0.0, 6.0), kwargs = {})
+      %aten_mean_dim : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.mean.dim](args = (%aten_hardtanh_default_34, [-1, -2], True), kwargs = {})
+      return (aten_mean_dim,)
+  %executorch_call_delegate : [num_users=1] = call_function[target=torch.ops.higher_order.executorch_call_delegate](args = (%lowered_module_0, %x), kwargs = {})
+  %getitem : [num_users=1] = call_function[target=operator.getitem](args = (%executorch_call_delegate, 0), kwargs = {})
+  %aten_view_copy_default : [num_users=1] = call_function[target=executorch.exir.memory.view](args = (%getitem, [1, 1280]), kwargs = {})
+  %alloc : [num_users=1] = call_function[target=executorch.exir.memory.alloc](args = (((1, 1280), torch.float32),), kwargs = {})
+  %dim_order_ops__clone_dim_order_default : [num_users=1] = call_function[target=torch.ops.dim_order_ops._clone_dim_order.out](args = (%aten_view_copy_default,), kwargs = {dim_order: [0, 1], out: %alloc})
+  %lowered_module_1 : [num_users=1] = get_attr[target=lowered_module_1]
+    backend_id: XnnpackBackend
+    lowered graph():
+      %p_classifier_1_weight : [num_users=1] = placeholder[target=p_classifier_1_weight]
+      %p_classifier_1_bias : [num_users=1] = placeholder[target=p_classifier_1_bias]
+      %dim_order_ops__clone_dim_order_default : [num_users=1] = placeholder[target=dim_order_ops__clone_dim_order_default]
+      %aten_linear_default : [num_users=1] = call_function[target=executorch.exir.dialects.edge._ops.aten.linear.default](args = (%dim_order_ops__clone_dim_order_default, %p_classifier_1_weight, %p_classifier_1_bias), kwargs = {})
+      return (aten_linear_default,)
+  %executorch_call_delegate_1 : [num_users=1] = call_function[target=torch.ops.higher_order.executorch_call_delegate](args = (%lowered_module_1, %dim_order_ops__clone_dim_order_default), kwargs = {})
+  %getitem_1 : [num_users=1] = call_function[target=operator.getitem](args = (%executorch_call_delegate_1, 0), kwargs = {})
+  return (getitem_1,)
 ```
+</details>