Skip to content

Instantly share code, notes, and snippets.

@pashu123
Created January 25, 2023 15:54
Show Gist options
  • Save pashu123/ddd747b716838a6b472083d05e2c0154 to your computer and use it in GitHub Desktop.
Save pashu123/ddd747b716838a6b472083d05e2c0154 to your computer and use it in GitHub Desktop.
graph():
%arg0_1 : [#users=1] = placeholder[target=arg0_1]
%mul : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%arg0_1, 5.489980785067252), kwargs = {})
%_param_constant0 : [#users=1] = get_attr[target=_param_constant0]
%_param_constant1 : [#users=1] = get_attr[target=_param_constant1]
%convolution : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%mul, %_param_constant0, %_param_constant1, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
%_param_constant2 : [#users=1] = get_attr[target=_param_constant2]
%_param_constant3 : [#users=1] = get_attr[target=_param_constant3]
%convolution_1 : [#users=2] = call_function[target=torch.ops.aten.convolution](args = (%convolution, %_param_constant2, %_param_constant3, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_1, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view, torch.float32), kwargs = {})
%var_mean : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem : [#users=1] = call_function[target=operator.getitem](args = (%var_mean, 0), kwargs = {})
%_to_copy : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem,), kwargs = {dtype: torch.float16})
%getitem_1 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean, 1), kwargs = {})
%_to_copy_1 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_1,), kwargs = {dtype: torch.float16})
%add : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy, 1e-06), kwargs = {})
%rsqrt : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add,), kwargs = {})
%sub : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view, %_to_copy_1), kwargs = {})
%mul_1 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub, %rsqrt), kwargs = {})
%view_1 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_1, [1, 512, 64, 64]), kwargs = {})
%_param_constant4 : [#users=1] = get_attr[target=_param_constant4]
%unsqueeze : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant4, 0), kwargs = {})
%unsqueeze_1 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze, 2), kwargs = {})
%unsqueeze_2 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_1, 3), kwargs = {})
%_param_constant5 : [#users=1] = get_attr[target=_param_constant5]
%unsqueeze_3 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant5, 0), kwargs = {})
%unsqueeze_4 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_3, 2), kwargs = {})
%unsqueeze_5 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_4, 3), kwargs = {})
%mul_2 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_1, %unsqueeze_5), kwargs = {})
%add_1 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_2, %unsqueeze_2), kwargs = {})
%squeeze : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_1, 3), kwargs = {})
%squeeze_1 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze, 2), kwargs = {})
%squeeze_2 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt, 3), kwargs = {})
%squeeze_3 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_2, 2), kwargs = {})
%detach : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_1,), kwargs = {})
%detach_1 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_3,), kwargs = {})
%silu : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_1,), kwargs = {})
%_param_constant6 : [#users=1] = get_attr[target=_param_constant6]
%_param_constant7 : [#users=1] = get_attr[target=_param_constant7]
%convolution_2 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu, %_param_constant6, %_param_constant7, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_2 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_2, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type_1 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_2, torch.float32), kwargs = {})
%var_mean_1 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_1, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_2 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_1, 0), kwargs = {})
%_to_copy_2 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_2,), kwargs = {dtype: torch.float16})
%getitem_3 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_1, 1), kwargs = {})
%_to_copy_3 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_3,), kwargs = {dtype: torch.float16})
%add_2 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_2, 1e-06), kwargs = {})
%rsqrt_1 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_2,), kwargs = {})
%sub_1 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_2, %_to_copy_3), kwargs = {})
%mul_3 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_1, %rsqrt_1), kwargs = {})
%view_3 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_3, [1, 512, 64, 64]), kwargs = {})
%_param_constant8 : [#users=1] = get_attr[target=_param_constant8]
%unsqueeze_6 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant8, 0), kwargs = {})
%unsqueeze_7 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_6, 2), kwargs = {})
%unsqueeze_8 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_7, 3), kwargs = {})
%_param_constant9 : [#users=1] = get_attr[target=_param_constant9]
%unsqueeze_9 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant9, 0), kwargs = {})
%unsqueeze_10 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_9, 2), kwargs = {})
%unsqueeze_11 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_10, 3), kwargs = {})
%mul_4 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_3, %unsqueeze_11), kwargs = {})
%add_3 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_4, %unsqueeze_8), kwargs = {})
%squeeze_4 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_3, 3), kwargs = {})
%squeeze_5 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_4, 2), kwargs = {})
%squeeze_6 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_1, 3), kwargs = {})
%squeeze_7 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_6, 2), kwargs = {})
%detach_2 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_5,), kwargs = {})
%detach_3 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_7,), kwargs = {})
%silu_1 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_3,), kwargs = {})
%_param_constant10 : [#users=1] = get_attr[target=_param_constant10]
%_param_constant11 : [#users=1] = get_attr[target=_param_constant11]
%convolution_3 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_1, %_param_constant10, %_param_constant11, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_4 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%convolution_1, %convolution_3), kwargs = {})
%div : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_4, 1), kwargs = {})
%view_4 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%div, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type_2 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_4, torch.float32), kwargs = {})
%var_mean_2 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_2, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_4 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_2, 0), kwargs = {})
%_to_copy_4 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_4,), kwargs = {dtype: torch.float16})
%getitem_5 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_2, 1), kwargs = {})
%_to_copy_5 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_5,), kwargs = {dtype: torch.float16})
%add_5 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_4, 1e-06), kwargs = {})
%rsqrt_2 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_5,), kwargs = {})
%sub_2 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_4, %_to_copy_5), kwargs = {})
%mul_5 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_2, %rsqrt_2), kwargs = {})
%view_5 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_5, [1, 512, 64, 64]), kwargs = {})
%_param_constant12 : [#users=1] = get_attr[target=_param_constant12]
%unsqueeze_12 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant12, 0), kwargs = {})
%unsqueeze_13 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_12, 2), kwargs = {})
%unsqueeze_14 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_13, 3), kwargs = {})
%_param_constant13 : [#users=1] = get_attr[target=_param_constant13]
%unsqueeze_15 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant13, 0), kwargs = {})
%unsqueeze_16 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_15, 2), kwargs = {})
%unsqueeze_17 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_16, 3), kwargs = {})
%mul_6 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_5, %unsqueeze_17), kwargs = {})
%add_6 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_6, %unsqueeze_14), kwargs = {})
%squeeze_8 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_5, 3), kwargs = {})
%squeeze_9 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_8, 2), kwargs = {})
%squeeze_10 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_2, 3), kwargs = {})
%squeeze_11 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_10, 2), kwargs = {})
%detach_4 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_9,), kwargs = {})
%detach_5 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_11,), kwargs = {})
%view_6 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%add_6, [1, 512, 4096]), kwargs = {})
%transpose : [#users=3] = call_function[target=torch.ops.aten.transpose](args = (%view_6, 1, 2), kwargs = {})
%_param_constant14 : [#users=1] = get_attr[target=_param_constant14]
%t : [#users=1] = call_function[target=torch.ops.aten.t](args = (%_param_constant14,), kwargs = {})
%expand : [#users=1] = call_function[target=torch.ops.aten.expand](args = (%transpose, [1, 4096, 512]), kwargs = {})
%view_7 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%expand, [1, 4096, 512]), kwargs = {})
%expand_1 : [#users=1] = call_function[target=torch.ops.aten.expand](args = (%t, [1, 512, 512]), kwargs = {})
%view_8 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%expand_1, [1, 512, 512]), kwargs = {})
%bmm : [#users=1] = call_function[target=torch.ops.aten.bmm](args = (%view_7, %view_8), kwargs = {})
%_unsafe_view : [#users=1] = call_function[target=torch.ops.aten._unsafe_view](args = (%bmm, [1, 4096, 512]), kwargs = {})
%_param_constant15 : [#users=1] = get_attr[target=_param_constant15]
%add_7 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_unsafe_view, %_param_constant15), kwargs = {})
%_param_constant16 : [#users=1] = get_attr[target=_param_constant16]
%t_1 : [#users=1] = call_function[target=torch.ops.aten.t](args = (%_param_constant16,), kwargs = {})
%expand_2 : [#users=1] = call_function[target=torch.ops.aten.expand](args = (%transpose, [1, 4096, 512]), kwargs = {})
%view_9 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%expand_2, [1, 4096, 512]), kwargs = {})
%expand_3 : [#users=1] = call_function[target=torch.ops.aten.expand](args = (%t_1, [1, 512, 512]), kwargs = {})
%view_10 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%expand_3, [1, 512, 512]), kwargs = {})
%bmm_1 : [#users=1] = call_function[target=torch.ops.aten.bmm](args = (%view_9, %view_10), kwargs = {})
%_unsafe_view_1 : [#users=1] = call_function[target=torch.ops.aten._unsafe_view](args = (%bmm_1, [1, 4096, 512]), kwargs = {})
%_param_constant17 : [#users=1] = get_attr[target=_param_constant17]
%add_8 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_unsafe_view_1, %_param_constant17), kwargs = {})
%_param_constant18 : [#users=1] = get_attr[target=_param_constant18]
%t_2 : [#users=1] = call_function[target=torch.ops.aten.t](args = (%_param_constant18,), kwargs = {})
%expand_4 : [#users=1] = call_function[target=torch.ops.aten.expand](args = (%transpose, [1, 4096, 512]), kwargs = {})
%view_11 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%expand_4, [1, 4096, 512]), kwargs = {})
%expand_5 : [#users=1] = call_function[target=torch.ops.aten.expand](args = (%t_2, [1, 512, 512]), kwargs = {})
%view_12 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%expand_5, [1, 512, 512]), kwargs = {})
%bmm_2 : [#users=1] = call_function[target=torch.ops.aten.bmm](args = (%view_11, %view_12), kwargs = {})
%_unsafe_view_2 : [#users=1] = call_function[target=torch.ops.aten._unsafe_view](args = (%bmm_2, [1, 4096, 512]), kwargs = {})
%_param_constant19 : [#users=1] = get_attr[target=_param_constant19]
%add_9 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_unsafe_view_2, %_param_constant19), kwargs = {})
%view_13 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%add_7, [1, 4096, 1, 512]), kwargs = {})
%permute : [#users=1] = call_function[target=torch.ops.aten.permute](args = (%view_13, [0, 2, 1, 3]), kwargs = {})
%view_14 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%permute, [1, 4096, 512]), kwargs = {})
%view_15 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%add_8, [1, 4096, 1, 512]), kwargs = {})
%permute_1 : [#users=1] = call_function[target=torch.ops.aten.permute](args = (%view_15, [0, 2, 1, 3]), kwargs = {})
%view_16 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%permute_1, [1, 4096, 512]), kwargs = {})
%view_17 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%add_9, [1, 4096, 1, 512]), kwargs = {})
%permute_2 : [#users=1] = call_function[target=torch.ops.aten.permute](args = (%view_17, [0, 2, 1, 3]), kwargs = {})
%view_18 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%permute_2, [1, 4096, 512]), kwargs = {})
%empty : [#users=1] = call_function[target=torch.ops.aten.empty](args = ([1, 4096, 4096],), kwargs = {dtype: torch.float16, device: cpu, pin_memory: False})
%zero_ : [#users=1] = call_function[target=torch.ops.aten.zero_](args = (%empty,), kwargs = {})
%transpose_1 : [#users=1] = call_function[target=torch.ops.aten.transpose](args = (%view_16, -1, -2), kwargs = {})
%baddbmm : [#users=1] = call_function[target=torch.ops.aten.baddbmm](args = (%zero_, %view_14, %transpose_1), kwargs = {beta: 0, alpha: 0.044194173824159216})
%_softmax : [#users=2] = call_function[target=torch.ops.aten._softmax](args = (%baddbmm, -1, False), kwargs = {})
%detach_6 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%_softmax,), kwargs = {})
%bmm_3 : [#users=1] = call_function[target=torch.ops.aten.bmm](args = (%_softmax, %view_18), kwargs = {})
%view_19 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%bmm_3, [1, 1, 4096, 512]), kwargs = {})
%permute_3 : [#users=1] = call_function[target=torch.ops.aten.permute](args = (%view_19, [0, 2, 1, 3]), kwargs = {})
%view_20 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%permute_3, [1, 4096, 512]), kwargs = {})
%_param_constant20 : [#users=1] = get_attr[target=_param_constant20]
%t_3 : [#users=1] = call_function[target=torch.ops.aten.t](args = (%_param_constant20,), kwargs = {})
%view_21 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%view_20, [4096, 512]), kwargs = {})
%_param_constant21 : [#users=1] = get_attr[target=_param_constant21]
%addmm : [#users=1] = call_function[target=torch.ops.aten.addmm](args = (%_param_constant21, %view_21, %t_3), kwargs = {})
%view_22 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%addmm, [1, 4096, 512]), kwargs = {})
%transpose_2 : [#users=1] = call_function[target=torch.ops.aten.transpose](args = (%view_22, -1, -2), kwargs = {})
%view_23 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%transpose_2, [1, 512, 64, 64]), kwargs = {})
%add_10 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%view_23, %div), kwargs = {})
%div_1 : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_10, 1), kwargs = {})
%clone : [#users=1] = call_function[target=torch.ops.aten.clone](args = (%div_1,), kwargs = {memory_format: torch.contiguous_format})
%view_24 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%clone, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type_3 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_24, torch.float32), kwargs = {})
%var_mean_3 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_3, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_6 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_3, 0), kwargs = {})
%_to_copy_6 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_6,), kwargs = {dtype: torch.float16})
%getitem_7 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_3, 1), kwargs = {})
%_to_copy_7 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_7,), kwargs = {dtype: torch.float16})
%add_11 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_6, 1e-06), kwargs = {})
%rsqrt_3 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_11,), kwargs = {})
%sub_3 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_24, %_to_copy_7), kwargs = {})
%mul_7 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_3, %rsqrt_3), kwargs = {})
%view_25 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_7, [1, 512, 64, 64]), kwargs = {})
%_param_constant22 : [#users=1] = get_attr[target=_param_constant22]
%unsqueeze_18 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant22, 0), kwargs = {})
%unsqueeze_19 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_18, 2), kwargs = {})
%unsqueeze_20 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_19, 3), kwargs = {})
%_param_constant23 : [#users=1] = get_attr[target=_param_constant23]
%unsqueeze_21 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant23, 0), kwargs = {})
%unsqueeze_22 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_21, 2), kwargs = {})
%unsqueeze_23 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_22, 3), kwargs = {})
%mul_8 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_25, %unsqueeze_23), kwargs = {})
%add_12 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_8, %unsqueeze_20), kwargs = {})
%squeeze_12 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_7, 3), kwargs = {})
%squeeze_13 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_12, 2), kwargs = {})
%squeeze_14 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_3, 3), kwargs = {})
%squeeze_15 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_14, 2), kwargs = {})
%detach_7 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_13,), kwargs = {})
%detach_8 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_15,), kwargs = {})
%silu_2 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_12,), kwargs = {})
%_param_constant24 : [#users=1] = get_attr[target=_param_constant24]
%_param_constant25 : [#users=1] = get_attr[target=_param_constant25]
%convolution_4 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_2, %_param_constant24, %_param_constant25, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_26 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_4, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type_4 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_26, torch.float32), kwargs = {})
%var_mean_4 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_4, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_8 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_4, 0), kwargs = {})
%_to_copy_8 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_8,), kwargs = {dtype: torch.float16})
%getitem_9 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_4, 1), kwargs = {})
%_to_copy_9 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_9,), kwargs = {dtype: torch.float16})
%add_13 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_8, 1e-06), kwargs = {})
%rsqrt_4 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_13,), kwargs = {})
%sub_4 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_26, %_to_copy_9), kwargs = {})
%mul_9 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_4, %rsqrt_4), kwargs = {})
%view_27 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_9, [1, 512, 64, 64]), kwargs = {})
%_param_constant26 : [#users=1] = get_attr[target=_param_constant26]
%unsqueeze_24 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant26, 0), kwargs = {})
%unsqueeze_25 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_24, 2), kwargs = {})
%unsqueeze_26 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_25, 3), kwargs = {})
%_param_constant27 : [#users=1] = get_attr[target=_param_constant27]
%unsqueeze_27 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant27, 0), kwargs = {})
%unsqueeze_28 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_27, 2), kwargs = {})
%unsqueeze_29 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_28, 3), kwargs = {})
%mul_10 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_27, %unsqueeze_29), kwargs = {})
%add_14 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_10, %unsqueeze_26), kwargs = {})
%squeeze_16 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_9, 3), kwargs = {})
%squeeze_17 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_16, 2), kwargs = {})
%squeeze_18 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_4, 3), kwargs = {})
%squeeze_19 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_18, 2), kwargs = {})
%detach_9 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_17,), kwargs = {})
%detach_10 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_19,), kwargs = {})
%silu_3 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_14,), kwargs = {})
%_param_constant28 : [#users=1] = get_attr[target=_param_constant28]
%_param_constant29 : [#users=1] = get_attr[target=_param_constant29]
%convolution_5 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_3, %_param_constant28, %_param_constant29, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_15 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_1, %convolution_5), kwargs = {})
%div_2 : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_15, 1), kwargs = {})
%clone_1 : [#users=1] = call_function[target=torch.ops.aten.clone](args = (%div_2,), kwargs = {memory_format: torch.contiguous_format})
%view_28 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%clone_1, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type_5 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_28, torch.float32), kwargs = {})
%var_mean_5 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_5, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_10 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_5, 0), kwargs = {})
%_to_copy_10 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_10,), kwargs = {dtype: torch.float16})
%getitem_11 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_5, 1), kwargs = {})
%_to_copy_11 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_11,), kwargs = {dtype: torch.float16})
%add_16 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_10, 1e-06), kwargs = {})
%rsqrt_5 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_16,), kwargs = {})
%sub_5 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_28, %_to_copy_11), kwargs = {})
%mul_11 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_5, %rsqrt_5), kwargs = {})
%view_29 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_11, [1, 512, 64, 64]), kwargs = {})
%_param_constant30 : [#users=1] = get_attr[target=_param_constant30]
%unsqueeze_30 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant30, 0), kwargs = {})
%unsqueeze_31 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_30, 2), kwargs = {})
%unsqueeze_32 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_31, 3), kwargs = {})
%_param_constant31 : [#users=1] = get_attr[target=_param_constant31]
%unsqueeze_33 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant31, 0), kwargs = {})
%unsqueeze_34 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_33, 2), kwargs = {})
%unsqueeze_35 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_34, 3), kwargs = {})
%mul_12 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_29, %unsqueeze_35), kwargs = {})
%add_17 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_12, %unsqueeze_32), kwargs = {})
%squeeze_20 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_11, 3), kwargs = {})
%squeeze_21 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_20, 2), kwargs = {})
%squeeze_22 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_5, 3), kwargs = {})
%squeeze_23 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_22, 2), kwargs = {})
%detach_11 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_21,), kwargs = {})
%detach_12 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_23,), kwargs = {})
%silu_4 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_17,), kwargs = {})
%_param_constant32 : [#users=1] = get_attr[target=_param_constant32]
%_param_constant33 : [#users=1] = get_attr[target=_param_constant33]
%convolution_6 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_4, %_param_constant32, %_param_constant33, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_30 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_6, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type_6 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_30, torch.float32), kwargs = {})
%var_mean_6 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_6, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_12 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_6, 0), kwargs = {})
%_to_copy_12 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_12,), kwargs = {dtype: torch.float16})
%getitem_13 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_6, 1), kwargs = {})
%_to_copy_13 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_13,), kwargs = {dtype: torch.float16})
%add_18 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_12, 1e-06), kwargs = {})
%rsqrt_6 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_18,), kwargs = {})
%sub_6 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_30, %_to_copy_13), kwargs = {})
%mul_13 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_6, %rsqrt_6), kwargs = {})
%view_31 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_13, [1, 512, 64, 64]), kwargs = {})
%_param_constant34 : [#users=1] = get_attr[target=_param_constant34]
%unsqueeze_36 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant34, 0), kwargs = {})
%unsqueeze_37 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_36, 2), kwargs = {})
%unsqueeze_38 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_37, 3), kwargs = {})
%_param_constant35 : [#users=1] = get_attr[target=_param_constant35]
%unsqueeze_39 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant35, 0), kwargs = {})
%unsqueeze_40 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_39, 2), kwargs = {})
%unsqueeze_41 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_40, 3), kwargs = {})
%mul_14 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_31, %unsqueeze_41), kwargs = {})
%add_19 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_14, %unsqueeze_38), kwargs = {})
%squeeze_24 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_13, 3), kwargs = {})
%squeeze_25 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_24, 2), kwargs = {})
%squeeze_26 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_6, 3), kwargs = {})
%squeeze_27 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_26, 2), kwargs = {})
%detach_13 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_25,), kwargs = {})
%detach_14 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_27,), kwargs = {})
%silu_5 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_19,), kwargs = {})
%_param_constant36 : [#users=1] = get_attr[target=_param_constant36]
%_param_constant37 : [#users=1] = get_attr[target=_param_constant37]
%convolution_7 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_5, %_param_constant36, %_param_constant37, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_20 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_2, %convolution_7), kwargs = {})
%div_3 : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_20, 1.0), kwargs = {})
%clone_2 : [#users=1] = call_function[target=torch.ops.aten.clone](args = (%div_3,), kwargs = {memory_format: torch.contiguous_format})
%view_32 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%clone_2, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type_7 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_32, torch.float32), kwargs = {})
%var_mean_7 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_7, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_14 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_7, 0), kwargs = {})
%_to_copy_14 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_14,), kwargs = {dtype: torch.float16})
%getitem_15 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_7, 1), kwargs = {})
%_to_copy_15 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_15,), kwargs = {dtype: torch.float16})
%add_21 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_14, 1e-06), kwargs = {})
%rsqrt_7 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_21,), kwargs = {})
%sub_7 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_32, %_to_copy_15), kwargs = {})
%mul_15 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_7, %rsqrt_7), kwargs = {})
%view_33 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_15, [1, 512, 64, 64]), kwargs = {})
%_param_constant38 : [#users=1] = get_attr[target=_param_constant38]
%unsqueeze_42 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant38, 0), kwargs = {})
%unsqueeze_43 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_42, 2), kwargs = {})
%unsqueeze_44 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_43, 3), kwargs = {})
%_param_constant39 : [#users=1] = get_attr[target=_param_constant39]
%unsqueeze_45 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant39, 0), kwargs = {})
%unsqueeze_46 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_45, 2), kwargs = {})
%unsqueeze_47 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_46, 3), kwargs = {})
%mul_16 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_33, %unsqueeze_47), kwargs = {})
%add_22 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_16, %unsqueeze_44), kwargs = {})
%squeeze_28 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_15, 3), kwargs = {})
%squeeze_29 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_28, 2), kwargs = {})
%squeeze_30 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_7, 3), kwargs = {})
%squeeze_31 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_30, 2), kwargs = {})
%detach_15 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_29,), kwargs = {})
%detach_16 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_31,), kwargs = {})
%silu_6 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_22,), kwargs = {})
%_param_constant40 : [#users=1] = get_attr[target=_param_constant40]
%_param_constant41 : [#users=1] = get_attr[target=_param_constant41]
%convolution_8 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_6, %_param_constant40, %_param_constant41, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_34 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_8, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type_8 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_34, torch.float32), kwargs = {})
%var_mean_8 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_8, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_16 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_8, 0), kwargs = {})
%_to_copy_16 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_16,), kwargs = {dtype: torch.float16})
%getitem_17 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_8, 1), kwargs = {})
%_to_copy_17 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_17,), kwargs = {dtype: torch.float16})
%add_23 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_16, 1e-06), kwargs = {})
%rsqrt_8 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_23,), kwargs = {})
%sub_8 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_34, %_to_copy_17), kwargs = {})
%mul_17 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_8, %rsqrt_8), kwargs = {})
%view_35 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_17, [1, 512, 64, 64]), kwargs = {})
%_param_constant42 : [#users=1] = get_attr[target=_param_constant42]
%unsqueeze_48 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant42, 0), kwargs = {})
%unsqueeze_49 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_48, 2), kwargs = {})
%unsqueeze_50 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_49, 3), kwargs = {})
%_param_constant43 : [#users=1] = get_attr[target=_param_constant43]
%unsqueeze_51 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant43, 0), kwargs = {})
%unsqueeze_52 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_51, 2), kwargs = {})
%unsqueeze_53 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_52, 3), kwargs = {})
%mul_18 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_35, %unsqueeze_53), kwargs = {})
%add_24 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_18, %unsqueeze_50), kwargs = {})
%squeeze_32 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_17, 3), kwargs = {})
%squeeze_33 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_32, 2), kwargs = {})
%squeeze_34 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_8, 3), kwargs = {})
%squeeze_35 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_34, 2), kwargs = {})
%detach_17 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_33,), kwargs = {})
%detach_18 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_35,), kwargs = {})
%silu_7 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_24,), kwargs = {})
%_param_constant44 : [#users=1] = get_attr[target=_param_constant44]
%_param_constant45 : [#users=1] = get_attr[target=_param_constant45]
%convolution_9 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_7, %_param_constant44, %_param_constant45, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_25 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_3, %convolution_9), kwargs = {})
%div_4 : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_25, 1.0), kwargs = {})
%clone_3 : [#users=1] = call_function[target=torch.ops.aten.clone](args = (%div_4,), kwargs = {memory_format: torch.contiguous_format})
%view_36 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%clone_3, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type_9 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_36, torch.float32), kwargs = {})
%var_mean_9 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_9, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_18 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_9, 0), kwargs = {})
%_to_copy_18 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_18,), kwargs = {dtype: torch.float16})
%getitem_19 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_9, 1), kwargs = {})
%_to_copy_19 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_19,), kwargs = {dtype: torch.float16})
%add_26 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_18, 1e-06), kwargs = {})
%rsqrt_9 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_26,), kwargs = {})
%sub_9 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_36, %_to_copy_19), kwargs = {})
%mul_19 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_9, %rsqrt_9), kwargs = {})
%view_37 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_19, [1, 512, 64, 64]), kwargs = {})
%_param_constant46 : [#users=1] = get_attr[target=_param_constant46]
%unsqueeze_54 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant46, 0), kwargs = {})
%unsqueeze_55 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_54, 2), kwargs = {})
%unsqueeze_56 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_55, 3), kwargs = {})
%_param_constant47 : [#users=1] = get_attr[target=_param_constant47]
%unsqueeze_57 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant47, 0), kwargs = {})
%unsqueeze_58 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_57, 2), kwargs = {})
%unsqueeze_59 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_58, 3), kwargs = {})
%mul_20 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_37, %unsqueeze_59), kwargs = {})
%add_27 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_20, %unsqueeze_56), kwargs = {})
%squeeze_36 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_19, 3), kwargs = {})
%squeeze_37 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_36, 2), kwargs = {})
%squeeze_38 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_9, 3), kwargs = {})
%squeeze_39 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_38, 2), kwargs = {})
%detach_19 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_37,), kwargs = {})
%detach_20 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_39,), kwargs = {})
%silu_8 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_27,), kwargs = {})
%_param_constant48 : [#users=1] = get_attr[target=_param_constant48]
%_param_constant49 : [#users=1] = get_attr[target=_param_constant49]
%convolution_10 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_8, %_param_constant48, %_param_constant49, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_38 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_10, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type_10 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_38, torch.float32), kwargs = {})
%var_mean_10 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_10, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_20 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_10, 0), kwargs = {})
%_to_copy_20 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_20,), kwargs = {dtype: torch.float16})
%getitem_21 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_10, 1), kwargs = {})
%_to_copy_21 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_21,), kwargs = {dtype: torch.float16})
%add_28 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_20, 1e-06), kwargs = {})
%rsqrt_10 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_28,), kwargs = {})
%sub_10 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_38, %_to_copy_21), kwargs = {})
%mul_21 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_10, %rsqrt_10), kwargs = {})
%view_39 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_21, [1, 512, 64, 64]), kwargs = {})
%_param_constant50 : [#users=1] = get_attr[target=_param_constant50]
%unsqueeze_60 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant50, 0), kwargs = {})
%unsqueeze_61 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_60, 2), kwargs = {})
%unsqueeze_62 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_61, 3), kwargs = {})
%_param_constant51 : [#users=1] = get_attr[target=_param_constant51]
%unsqueeze_63 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant51, 0), kwargs = {})
%unsqueeze_64 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_63, 2), kwargs = {})
%unsqueeze_65 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_64, 3), kwargs = {})
%mul_22 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_39, %unsqueeze_65), kwargs = {})
%add_29 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_22, %unsqueeze_62), kwargs = {})
%squeeze_40 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_21, 3), kwargs = {})
%squeeze_41 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_40, 2), kwargs = {})
%squeeze_42 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_10, 3), kwargs = {})
%squeeze_43 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_42, 2), kwargs = {})
%detach_21 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_41,), kwargs = {})
%detach_22 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_43,), kwargs = {})
%silu_9 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_29,), kwargs = {})
%_param_constant52 : [#users=1] = get_attr[target=_param_constant52]
%_param_constant53 : [#users=1] = get_attr[target=_param_constant53]
%convolution_11 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_9, %_param_constant52, %_param_constant53, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_30 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_4, %convolution_11), kwargs = {})
%div_5 : [#users=1] = call_function[target=torch.ops.aten.div](args = (%add_30, 1.0), kwargs = {})
%upsample_nearest2d : [#users=1] = call_function[target=torch.ops.aten.upsample_nearest2d](args = (%div_5, [128, 128], 2.0, 2.0), kwargs = {})
%_param_constant54 : [#users=1] = get_attr[target=_param_constant54]
%_param_constant55 : [#users=1] = get_attr[target=_param_constant55]
%convolution_12 : [#users=2] = call_function[target=torch.ops.aten.convolution](args = (%upsample_nearest2d, %_param_constant54, %_param_constant55, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_40 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_12, [1, 32, 16, 16384]), kwargs = {})
%convert_element_type_11 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_40, torch.float32), kwargs = {})
%var_mean_11 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_11, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_22 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_11, 0), kwargs = {})
%_to_copy_22 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_22,), kwargs = {dtype: torch.float16})
%getitem_23 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_11, 1), kwargs = {})
%_to_copy_23 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_23,), kwargs = {dtype: torch.float16})
%add_31 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_22, 1e-06), kwargs = {})
%rsqrt_11 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_31,), kwargs = {})
%sub_11 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_40, %_to_copy_23), kwargs = {})
%mul_23 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_11, %rsqrt_11), kwargs = {})
%view_41 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_23, [1, 512, 128, 128]), kwargs = {})
%_param_constant56 : [#users=1] = get_attr[target=_param_constant56]
%unsqueeze_66 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant56, 0), kwargs = {})
%unsqueeze_67 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_66, 2), kwargs = {})
%unsqueeze_68 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_67, 3), kwargs = {})
%_param_constant57 : [#users=1] = get_attr[target=_param_constant57]
%unsqueeze_69 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant57, 0), kwargs = {})
%unsqueeze_70 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_69, 2), kwargs = {})
%unsqueeze_71 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_70, 3), kwargs = {})
%mul_24 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_41, %unsqueeze_71), kwargs = {})
%add_32 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_24, %unsqueeze_68), kwargs = {})
%squeeze_44 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_23, 3), kwargs = {})
%squeeze_45 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_44, 2), kwargs = {})
%squeeze_46 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_11, 3), kwargs = {})
%squeeze_47 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_46, 2), kwargs = {})
%detach_23 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_45,), kwargs = {})
%detach_24 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_47,), kwargs = {})
%silu_10 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_32,), kwargs = {})
%_param_constant58 : [#users=1] = get_attr[target=_param_constant58]
%_param_constant59 : [#users=1] = get_attr[target=_param_constant59]
%convolution_13 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_10, %_param_constant58, %_param_constant59, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_42 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_13, [1, 32, 16, 16384]), kwargs = {})
%convert_element_type_12 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_42, torch.float32), kwargs = {})
%var_mean_12 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_12, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_24 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_12, 0), kwargs = {})
%_to_copy_24 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_24,), kwargs = {dtype: torch.float16})
%getitem_25 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_12, 1), kwargs = {})
%_to_copy_25 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_25,), kwargs = {dtype: torch.float16})
%add_33 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_24, 1e-06), kwargs = {})
%rsqrt_12 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_33,), kwargs = {})
%sub_12 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_42, %_to_copy_25), kwargs = {})
%mul_25 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_12, %rsqrt_12), kwargs = {})
%view_43 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_25, [1, 512, 128, 128]), kwargs = {})
%_param_constant60 : [#users=1] = get_attr[target=_param_constant60]
%unsqueeze_72 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant60, 0), kwargs = {})
%unsqueeze_73 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_72, 2), kwargs = {})
%unsqueeze_74 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_73, 3), kwargs = {})
%_param_constant61 : [#users=1] = get_attr[target=_param_constant61]
%unsqueeze_75 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant61, 0), kwargs = {})
%unsqueeze_76 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_75, 2), kwargs = {})
%unsqueeze_77 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_76, 3), kwargs = {})
%mul_26 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_43, %unsqueeze_77), kwargs = {})
%add_34 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_26, %unsqueeze_74), kwargs = {})
%squeeze_48 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_25, 3), kwargs = {})
%squeeze_49 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_48, 2), kwargs = {})
%squeeze_50 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_12, 3), kwargs = {})
%squeeze_51 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_50, 2), kwargs = {})
%detach_25 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_49,), kwargs = {})
%detach_26 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_51,), kwargs = {})
%silu_11 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_34,), kwargs = {})
%_param_constant62 : [#users=1] = get_attr[target=_param_constant62]
%_param_constant63 : [#users=1] = get_attr[target=_param_constant63]
%convolution_14 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_11, %_param_constant62, %_param_constant63, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_35 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%convolution_12, %convolution_14), kwargs = {})
%div_6 : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_35, 1.0), kwargs = {})
%view_44 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%div_6, [1, 32, 16, 16384]), kwargs = {})
%convert_element_type_13 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_44, torch.float32), kwargs = {})
%var_mean_13 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_13, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_26 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_13, 0), kwargs = {})
%_to_copy_26 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_26,), kwargs = {dtype: torch.float16})
%getitem_27 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_13, 1), kwargs = {})
%_to_copy_27 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_27,), kwargs = {dtype: torch.float16})
%add_36 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_26, 1e-06), kwargs = {})
%rsqrt_13 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_36,), kwargs = {})
%sub_13 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_44, %_to_copy_27), kwargs = {})
%mul_27 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_13, %rsqrt_13), kwargs = {})
%view_45 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_27, [1, 512, 128, 128]), kwargs = {})
%_param_constant64 : [#users=1] = get_attr[target=_param_constant64]
%unsqueeze_78 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant64, 0), kwargs = {})
%unsqueeze_79 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_78, 2), kwargs = {})
%unsqueeze_80 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_79, 3), kwargs = {})
%_param_constant65 : [#users=1] = get_attr[target=_param_constant65]
%unsqueeze_81 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant65, 0), kwargs = {})
%unsqueeze_82 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_81, 2), kwargs = {})
%unsqueeze_83 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_82, 3), kwargs = {})
%mul_28 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_45, %unsqueeze_83), kwargs = {})
%add_37 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_28, %unsqueeze_80), kwargs = {})
%squeeze_52 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_27, 3), kwargs = {})
%squeeze_53 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_52, 2), kwargs = {})
%squeeze_54 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_13, 3), kwargs = {})
%squeeze_55 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_54, 2), kwargs = {})
%detach_27 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_53,), kwargs = {})
%detach_28 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_55,), kwargs = {})
%silu_12 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_37,), kwargs = {})
%_param_constant66 : [#users=1] = get_attr[target=_param_constant66]
%_param_constant67 : [#users=1] = get_attr[target=_param_constant67]
%convolution_15 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_12, %_param_constant66, %_param_constant67, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_46 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_15, [1, 32, 16, 16384]), kwargs = {})
%convert_element_type_14 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_46, torch.float32), kwargs = {})
%var_mean_14 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_14, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_28 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_14, 0), kwargs = {})
%_to_copy_28 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_28,), kwargs = {dtype: torch.float16})
%getitem_29 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_14, 1), kwargs = {})
%_to_copy_29 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_29,), kwargs = {dtype: torch.float16})
%add_38 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_28, 1e-06), kwargs = {})
%rsqrt_14 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_38,), kwargs = {})
%sub_14 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_46, %_to_copy_29), kwargs = {})
%mul_29 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_14, %rsqrt_14), kwargs = {})
%view_47 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_29, [1, 512, 128, 128]), kwargs = {})
%_param_constant68 : [#users=1] = get_attr[target=_param_constant68]
%unsqueeze_84 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant68, 0), kwargs = {})
%unsqueeze_85 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_84, 2), kwargs = {})
%unsqueeze_86 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_85, 3), kwargs = {})
%_param_constant69 : [#users=1] = get_attr[target=_param_constant69]
%unsqueeze_87 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant69, 0), kwargs = {})
%unsqueeze_88 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_87, 2), kwargs = {})
%unsqueeze_89 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_88, 3), kwargs = {})
%mul_30 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_47, %unsqueeze_89), kwargs = {})
%add_39 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_30, %unsqueeze_86), kwargs = {})
%squeeze_56 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_29, 3), kwargs = {})
%squeeze_57 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_56, 2), kwargs = {})
%squeeze_58 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_14, 3), kwargs = {})
%squeeze_59 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_58, 2), kwargs = {})
%detach_29 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_57,), kwargs = {})
%detach_30 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_59,), kwargs = {})
%silu_13 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_39,), kwargs = {})
%_param_constant70 : [#users=1] = get_attr[target=_param_constant70]
%_param_constant71 : [#users=1] = get_attr[target=_param_constant71]
%convolution_16 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_13, %_param_constant70, %_param_constant71, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_40 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_6, %convolution_16), kwargs = {})
%div_7 : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_40, 1.0), kwargs = {})
%view_48 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%div_7, [1, 32, 16, 16384]), kwargs = {})
%convert_element_type_15 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_48, torch.float32), kwargs = {})
%var_mean_15 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_15, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_30 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_15, 0), kwargs = {})
%_to_copy_30 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_30,), kwargs = {dtype: torch.float16})
%getitem_31 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_15, 1), kwargs = {})
%_to_copy_31 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_31,), kwargs = {dtype: torch.float16})
%add_41 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_30, 1e-06), kwargs = {})
%rsqrt_15 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_41,), kwargs = {})
%sub_15 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_48, %_to_copy_31), kwargs = {})
%mul_31 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_15, %rsqrt_15), kwargs = {})
%view_49 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_31, [1, 512, 128, 128]), kwargs = {})
%_param_constant72 : [#users=1] = get_attr[target=_param_constant72]
%unsqueeze_90 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant72, 0), kwargs = {})
%unsqueeze_91 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_90, 2), kwargs = {})
%unsqueeze_92 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_91, 3), kwargs = {})
%_param_constant73 : [#users=1] = get_attr[target=_param_constant73]
%unsqueeze_93 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant73, 0), kwargs = {})
%unsqueeze_94 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_93, 2), kwargs = {})
%unsqueeze_95 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_94, 3), kwargs = {})
%mul_32 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_49, %unsqueeze_95), kwargs = {})
%add_42 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_32, %unsqueeze_92), kwargs = {})
%squeeze_60 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_31, 3), kwargs = {})
%squeeze_61 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_60, 2), kwargs = {})
%squeeze_62 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_15, 3), kwargs = {})
%squeeze_63 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_62, 2), kwargs = {})
%detach_31 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_61,), kwargs = {})
%detach_32 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_63,), kwargs = {})
%silu_14 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_42,), kwargs = {})
%_param_constant74 : [#users=1] = get_attr[target=_param_constant74]
%_param_constant75 : [#users=1] = get_attr[target=_param_constant75]
%convolution_17 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_14, %_param_constant74, %_param_constant75, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_50 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_17, [1, 32, 16, 16384]), kwargs = {})
%convert_element_type_16 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_50, torch.float32), kwargs = {})
%var_mean_16 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_16, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_32 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_16, 0), kwargs = {})
%_to_copy_32 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_32,), kwargs = {dtype: torch.float16})
%getitem_33 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_16, 1), kwargs = {})
%_to_copy_33 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_33,), kwargs = {dtype: torch.float16})
%add_43 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_32, 1e-06), kwargs = {})
%rsqrt_16 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_43,), kwargs = {})
%sub_16 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_50, %_to_copy_33), kwargs = {})
%mul_33 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_16, %rsqrt_16), kwargs = {})
%view_51 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_33, [1, 512, 128, 128]), kwargs = {})
%_param_constant76 : [#users=1] = get_attr[target=_param_constant76]
%unsqueeze_96 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant76, 0), kwargs = {})
%unsqueeze_97 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_96, 2), kwargs = {})
%unsqueeze_98 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_97, 3), kwargs = {})
%_param_constant77 : [#users=1] = get_attr[target=_param_constant77]
%unsqueeze_99 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant77, 0), kwargs = {})
%unsqueeze_100 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_99, 2), kwargs = {})
%unsqueeze_101 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_100, 3), kwargs = {})
%mul_34 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_51, %unsqueeze_101), kwargs = {})
%add_44 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_34, %unsqueeze_98), kwargs = {})
%squeeze_64 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_33, 3), kwargs = {})
%squeeze_65 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_64, 2), kwargs = {})
%squeeze_66 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_16, 3), kwargs = {})
%squeeze_67 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_66, 2), kwargs = {})
%detach_33 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_65,), kwargs = {})
%detach_34 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_67,), kwargs = {})
%silu_15 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_44,), kwargs = {})
%_param_constant78 : [#users=1] = get_attr[target=_param_constant78]
%_param_constant79 : [#users=1] = get_attr[target=_param_constant79]
%convolution_18 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_15, %_param_constant78, %_param_constant79, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_45 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_7, %convolution_18), kwargs = {})
%div_8 : [#users=1] = call_function[target=torch.ops.aten.div](args = (%add_45, 1.0), kwargs = {})
%upsample_nearest2d_1 : [#users=1] = call_function[target=torch.ops.aten.upsample_nearest2d](args = (%div_8, [256, 256], 2.0, 2.0), kwargs = {})
%_param_constant80 : [#users=1] = get_attr[target=_param_constant80]
%_param_constant81 : [#users=1] = get_attr[target=_param_constant81]
%convolution_19 : [#users=2] = call_function[target=torch.ops.aten.convolution](args = (%upsample_nearest2d_1, %_param_constant80, %_param_constant81, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_52 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_19, [1, 32, 16, 65536]), kwargs = {})
%convert_element_type_17 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_52, torch.float32), kwargs = {})
%var_mean_17 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_17, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_34 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_17, 0), kwargs = {})
%_to_copy_34 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_34,), kwargs = {dtype: torch.float16})
%getitem_35 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_17, 1), kwargs = {})
%_to_copy_35 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_35,), kwargs = {dtype: torch.float16})
%add_46 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_34, 1e-06), kwargs = {})
%rsqrt_17 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_46,), kwargs = {})
%sub_17 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_52, %_to_copy_35), kwargs = {})
%mul_35 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_17, %rsqrt_17), kwargs = {})
%view_53 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_35, [1, 512, 256, 256]), kwargs = {})
%_param_constant82 : [#users=1] = get_attr[target=_param_constant82]
%unsqueeze_102 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant82, 0), kwargs = {})
%unsqueeze_103 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_102, 2), kwargs = {})
%unsqueeze_104 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_103, 3), kwargs = {})
%_param_constant83 : [#users=1] = get_attr[target=_param_constant83]
%unsqueeze_105 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant83, 0), kwargs = {})
%unsqueeze_106 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_105, 2), kwargs = {})
%unsqueeze_107 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_106, 3), kwargs = {})
%mul_36 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_53, %unsqueeze_107), kwargs = {})
%add_47 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_36, %unsqueeze_104), kwargs = {})
%squeeze_68 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_35, 3), kwargs = {})
%squeeze_69 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_68, 2), kwargs = {})
%squeeze_70 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_17, 3), kwargs = {})
%squeeze_71 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_70, 2), kwargs = {})
%detach_35 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_69,), kwargs = {})
%detach_36 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_71,), kwargs = {})
%silu_16 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_47,), kwargs = {})
%_param_constant84 : [#users=1] = get_attr[target=_param_constant84]
%_param_constant85 : [#users=1] = get_attr[target=_param_constant85]
%convolution_20 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_16, %_param_constant84, %_param_constant85, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_54 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_20, [1, 32, 8, 65536]), kwargs = {})
%convert_element_type_18 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_54, torch.float32), kwargs = {})
%var_mean_18 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_18, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_36 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_18, 0), kwargs = {})
%_to_copy_36 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_36,), kwargs = {dtype: torch.float16})
%getitem_37 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_18, 1), kwargs = {})
%_to_copy_37 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_37,), kwargs = {dtype: torch.float16})
%add_48 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_36, 1e-06), kwargs = {})
%rsqrt_18 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_48,), kwargs = {})
%sub_18 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_54, %_to_copy_37), kwargs = {})
%mul_37 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_18, %rsqrt_18), kwargs = {})
%view_55 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_37, [1, 256, 256, 256]), kwargs = {})
%_param_constant86 : [#users=1] = get_attr[target=_param_constant86]
%unsqueeze_108 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant86, 0), kwargs = {})
%unsqueeze_109 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_108, 2), kwargs = {})
%unsqueeze_110 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_109, 3), kwargs = {})
%_param_constant87 : [#users=1] = get_attr[target=_param_constant87]
%unsqueeze_111 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant87, 0), kwargs = {})
%unsqueeze_112 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_111, 2), kwargs = {})
%unsqueeze_113 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_112, 3), kwargs = {})
%mul_38 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_55, %unsqueeze_113), kwargs = {})
%add_49 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_38, %unsqueeze_110), kwargs = {})
%squeeze_72 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_37, 3), kwargs = {})
%squeeze_73 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_72, 2), kwargs = {})
%squeeze_74 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_18, 3), kwargs = {})
%squeeze_75 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_74, 2), kwargs = {})
%detach_37 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_73,), kwargs = {})
%detach_38 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_75,), kwargs = {})
%silu_17 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_49,), kwargs = {})
%_param_constant88 : [#users=1] = get_attr[target=_param_constant88]
%_param_constant89 : [#users=1] = get_attr[target=_param_constant89]
%convolution_21 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_17, %_param_constant88, %_param_constant89, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%_param_constant90 : [#users=1] = get_attr[target=_param_constant90]
%_param_constant91 : [#users=1] = get_attr[target=_param_constant91]
%convolution_22 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%convolution_19, %_param_constant90, %_param_constant91, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
%add_50 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%convolution_22, %convolution_21), kwargs = {})
%div_9 : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_50, 1.0), kwargs = {})
%view_56 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%div_9, [1, 32, 8, 65536]), kwargs = {})
%convert_element_type_19 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_56, torch.float32), kwargs = {})
%var_mean_19 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_19, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_38 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_19, 0), kwargs = {})
%_to_copy_38 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_38,), kwargs = {dtype: torch.float16})
%getitem_39 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_19, 1), kwargs = {})
%_to_copy_39 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_39,), kwargs = {dtype: torch.float16})
%add_51 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_38, 1e-06), kwargs = {})
%rsqrt_19 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_51,), kwargs = {})
%sub_19 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_56, %_to_copy_39), kwargs = {})
%mul_39 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_19, %rsqrt_19), kwargs = {})
%view_57 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_39, [1, 256, 256, 256]), kwargs = {})
%_param_constant92 : [#users=1] = get_attr[target=_param_constant92]
%unsqueeze_114 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant92, 0), kwargs = {})
%unsqueeze_115 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_114, 2), kwargs = {})
%unsqueeze_116 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_115, 3), kwargs = {})
%_param_constant93 : [#users=1] = get_attr[target=_param_constant93]
%unsqueeze_117 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant93, 0), kwargs = {})
%unsqueeze_118 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_117, 2), kwargs = {})
%unsqueeze_119 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_118, 3), kwargs = {})
%mul_40 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_57, %unsqueeze_119), kwargs = {})
%add_52 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_40, %unsqueeze_116), kwargs = {})
%squeeze_76 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_39, 3), kwargs = {})
%squeeze_77 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_76, 2), kwargs = {})
%squeeze_78 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_19, 3), kwargs = {})
%squeeze_79 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_78, 2), kwargs = {})
%detach_39 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_77,), kwargs = {})
%detach_40 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_79,), kwargs = {})
%silu_18 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_52,), kwargs = {})
%_param_constant94 : [#users=1] = get_attr[target=_param_constant94]
%_param_constant95 : [#users=1] = get_attr[target=_param_constant95]
%convolution_23 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_18, %_param_constant94, %_param_constant95, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_58 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_23, [1, 32, 8, 65536]), kwargs = {})
%convert_element_type_20 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_58, torch.float32), kwargs = {})
%var_mean_20 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_20, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_40 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_20, 0), kwargs = {})
%_to_copy_40 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_40,), kwargs = {dtype: torch.float16})
%getitem_41 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_20, 1), kwargs = {})
%_to_copy_41 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_41,), kwargs = {dtype: torch.float16})
%add_53 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_40, 1e-06), kwargs = {})
%rsqrt_20 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_53,), kwargs = {})
%sub_20 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_58, %_to_copy_41), kwargs = {})
%mul_41 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_20, %rsqrt_20), kwargs = {})
%view_59 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_41, [1, 256, 256, 256]), kwargs = {})
%_param_constant96 : [#users=1] = get_attr[target=_param_constant96]
%unsqueeze_120 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant96, 0), kwargs = {})
%unsqueeze_121 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_120, 2), kwargs = {})
%unsqueeze_122 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_121, 3), kwargs = {})
%_param_constant97 : [#users=1] = get_attr[target=_param_constant97]
%unsqueeze_123 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant97, 0), kwargs = {})
%unsqueeze_124 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_123, 2), kwargs = {})
%unsqueeze_125 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_124, 3), kwargs = {})
%mul_42 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_59, %unsqueeze_125), kwargs = {})
%add_54 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_42, %unsqueeze_122), kwargs = {})
%squeeze_80 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_41, 3), kwargs = {})
%squeeze_81 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_80, 2), kwargs = {})
%squeeze_82 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_20, 3), kwargs = {})
%squeeze_83 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_82, 2), kwargs = {})
%detach_41 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_81,), kwargs = {})
%detach_42 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_83,), kwargs = {})
%silu_19 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_54,), kwargs = {})
%_param_constant98 : [#users=1] = get_attr[target=_param_constant98]
%_param_constant99 : [#users=1] = get_attr[target=_param_constant99]
%convolution_24 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_19, %_param_constant98, %_param_constant99, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_55 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_9, %convolution_24), kwargs = {})
%div_10 : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_55, 1.0), kwargs = {})
%view_60 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%div_10, [1, 32, 8, 65536]), kwargs = {})
%convert_element_type_21 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_60, torch.float32), kwargs = {})
%var_mean_21 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_21, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_42 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_21, 0), kwargs = {})
%_to_copy_42 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_42,), kwargs = {dtype: torch.float16})
%getitem_43 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_21, 1), kwargs = {})
%_to_copy_43 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_43,), kwargs = {dtype: torch.float16})
%add_56 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_42, 1e-06), kwargs = {})
%rsqrt_21 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_56,), kwargs = {})
%sub_21 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_60, %_to_copy_43), kwargs = {})
%mul_43 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_21, %rsqrt_21), kwargs = {})
%view_61 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_43, [1, 256, 256, 256]), kwargs = {})
%_param_constant100 : [#users=1] = get_attr[target=_param_constant100]
%unsqueeze_126 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant100, 0), kwargs = {})
%unsqueeze_127 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_126, 2), kwargs = {})
%unsqueeze_128 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_127, 3), kwargs = {})
%_param_constant101 : [#users=1] = get_attr[target=_param_constant101]
%unsqueeze_129 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant101, 0), kwargs = {})
%unsqueeze_130 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_129, 2), kwargs = {})
%unsqueeze_131 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_130, 3), kwargs = {})
%mul_44 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_61, %unsqueeze_131), kwargs = {})
%add_57 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_44, %unsqueeze_128), kwargs = {})
%squeeze_84 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_43, 3), kwargs = {})
%squeeze_85 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_84, 2), kwargs = {})
%squeeze_86 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_21, 3), kwargs = {})
%squeeze_87 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_86, 2), kwargs = {})
%detach_43 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_85,), kwargs = {})
%detach_44 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_87,), kwargs = {})
%silu_20 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_57,), kwargs = {})
%_param_constant102 : [#users=1] = get_attr[target=_param_constant102]
%_param_constant103 : [#users=1] = get_attr[target=_param_constant103]
%convolution_25 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_20, %_param_constant102, %_param_constant103, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_62 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_25, [1, 32, 8, 65536]), kwargs = {})
%convert_element_type_22 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_62, torch.float32), kwargs = {})
%var_mean_22 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_22, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_44 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_22, 0), kwargs = {})
%_to_copy_44 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_44,), kwargs = {dtype: torch.float16})
%getitem_45 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_22, 1), kwargs = {})
%_to_copy_45 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_45,), kwargs = {dtype: torch.float16})
%add_58 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_44, 1e-06), kwargs = {})
%rsqrt_22 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_58,), kwargs = {})
%sub_22 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_62, %_to_copy_45), kwargs = {})
%mul_45 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_22, %rsqrt_22), kwargs = {})
%view_63 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_45, [1, 256, 256, 256]), kwargs = {})
%_param_constant104 : [#users=1] = get_attr[target=_param_constant104]
%unsqueeze_132 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant104, 0), kwargs = {})
%unsqueeze_133 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_132, 2), kwargs = {})
%unsqueeze_134 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_133, 3), kwargs = {})
%_param_constant105 : [#users=1] = get_attr[target=_param_constant105]
%unsqueeze_135 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant105, 0), kwargs = {})
%unsqueeze_136 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_135, 2), kwargs = {})
%unsqueeze_137 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_136, 3), kwargs = {})
%mul_46 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_63, %unsqueeze_137), kwargs = {})
%add_59 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_46, %unsqueeze_134), kwargs = {})
%squeeze_88 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_45, 3), kwargs = {})
%squeeze_89 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_88, 2), kwargs = {})
%squeeze_90 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_22, 3), kwargs = {})
%squeeze_91 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_90, 2), kwargs = {})
%detach_45 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_89,), kwargs = {})
%detach_46 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_91,), kwargs = {})
%silu_21 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_59,), kwargs = {})
%_param_constant106 : [#users=1] = get_attr[target=_param_constant106]
%_param_constant107 : [#users=1] = get_attr[target=_param_constant107]
%convolution_26 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_21, %_param_constant106, %_param_constant107, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_60 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_10, %convolution_26), kwargs = {})
%div_11 : [#users=1] = call_function[target=torch.ops.aten.div](args = (%add_60, 1.0), kwargs = {})
%upsample_nearest2d_2 : [#users=1] = call_function[target=torch.ops.aten.upsample_nearest2d](args = (%div_11, [512, 512], 2.0, 2.0), kwargs = {})
%_param_constant108 : [#users=1] = get_attr[target=_param_constant108]
%_param_constant109 : [#users=1] = get_attr[target=_param_constant109]
%convolution_27 : [#users=2] = call_function[target=torch.ops.aten.convolution](args = (%upsample_nearest2d_2, %_param_constant108, %_param_constant109, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_64 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_27, [1, 32, 8, 262144]), kwargs = {})
%convert_element_type_23 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_64, torch.float32), kwargs = {})
%var_mean_23 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_23, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_46 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_23, 0), kwargs = {})
%_to_copy_46 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_46,), kwargs = {dtype: torch.float16})
%getitem_47 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_23, 1), kwargs = {})
%_to_copy_47 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_47,), kwargs = {dtype: torch.float16})
%add_61 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_46, 1e-06), kwargs = {})
%rsqrt_23 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_61,), kwargs = {})
%sub_23 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_64, %_to_copy_47), kwargs = {})
%mul_47 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_23, %rsqrt_23), kwargs = {})
%view_65 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_47, [1, 256, 512, 512]), kwargs = {})
%_param_constant110 : [#users=1] = get_attr[target=_param_constant110]
%unsqueeze_138 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant110, 0), kwargs = {})
%unsqueeze_139 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_138, 2), kwargs = {})
%unsqueeze_140 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_139, 3), kwargs = {})
%_param_constant111 : [#users=1] = get_attr[target=_param_constant111]
%unsqueeze_141 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant111, 0), kwargs = {})
%unsqueeze_142 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_141, 2), kwargs = {})
%unsqueeze_143 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_142, 3), kwargs = {})
%mul_48 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_65, %unsqueeze_143), kwargs = {})
%add_62 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_48, %unsqueeze_140), kwargs = {})
%squeeze_92 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_47, 3), kwargs = {})
%squeeze_93 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_92, 2), kwargs = {})
%squeeze_94 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_23, 3), kwargs = {})
%squeeze_95 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_94, 2), kwargs = {})
%detach_47 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_93,), kwargs = {})
%detach_48 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_95,), kwargs = {})
%silu_22 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_62,), kwargs = {})
%_param_constant112 : [#users=1] = get_attr[target=_param_constant112]
%_param_constant113 : [#users=1] = get_attr[target=_param_constant113]
%convolution_28 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_22, %_param_constant112, %_param_constant113, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_66 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_28, [1, 32, 4, 262144]), kwargs = {})
%convert_element_type_24 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_66, torch.float32), kwargs = {})
%var_mean_24 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_24, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_48 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_24, 0), kwargs = {})
%_to_copy_48 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_48,), kwargs = {dtype: torch.float16})
%getitem_49 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_24, 1), kwargs = {})
%_to_copy_49 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_49,), kwargs = {dtype: torch.float16})
%add_63 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_48, 1e-06), kwargs = {})
%rsqrt_24 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_63,), kwargs = {})
%sub_24 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_66, %_to_copy_49), kwargs = {})
%mul_49 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_24, %rsqrt_24), kwargs = {})
%view_67 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_49, [1, 128, 512, 512]), kwargs = {})
%_param_constant114 : [#users=1] = get_attr[target=_param_constant114]
%unsqueeze_144 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant114, 0), kwargs = {})
%unsqueeze_145 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_144, 2), kwargs = {})
%unsqueeze_146 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_145, 3), kwargs = {})
%_param_constant115 : [#users=1] = get_attr[target=_param_constant115]
%unsqueeze_147 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant115, 0), kwargs = {})
%unsqueeze_148 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_147, 2), kwargs = {})
%unsqueeze_149 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_148, 3), kwargs = {})
%mul_50 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_67, %unsqueeze_149), kwargs = {})
%add_64 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_50, %unsqueeze_146), kwargs = {})
%squeeze_96 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_49, 3), kwargs = {})
%squeeze_97 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_96, 2), kwargs = {})
%squeeze_98 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_24, 3), kwargs = {})
%squeeze_99 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_98, 2), kwargs = {})
%detach_49 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_97,), kwargs = {})
%detach_50 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_99,), kwargs = {})
%silu_23 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_64,), kwargs = {})
%_param_constant116 : [#users=1] = get_attr[target=_param_constant116]
%_param_constant117 : [#users=1] = get_attr[target=_param_constant117]
%convolution_29 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_23, %_param_constant116, %_param_constant117, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%_param_constant118 : [#users=1] = get_attr[target=_param_constant118]
%_param_constant119 : [#users=1] = get_attr[target=_param_constant119]
%convolution_30 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%convolution_27, %_param_constant118, %_param_constant119, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
%add_65 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%convolution_30, %convolution_29), kwargs = {})
%div_12 : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_65, 1.0), kwargs = {})
%view_68 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%div_12, [1, 32, 4, 262144]), kwargs = {})
%convert_element_type_25 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_68, torch.float32), kwargs = {})
%var_mean_25 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_25, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_50 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_25, 0), kwargs = {})
%_to_copy_50 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_50,), kwargs = {dtype: torch.float16})
%getitem_51 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_25, 1), kwargs = {})
%_to_copy_51 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_51,), kwargs = {dtype: torch.float16})
%add_66 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_50, 1e-06), kwargs = {})
%rsqrt_25 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_66,), kwargs = {})
%sub_25 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_68, %_to_copy_51), kwargs = {})
%mul_51 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_25, %rsqrt_25), kwargs = {})
%view_69 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_51, [1, 128, 512, 512]), kwargs = {})
%_param_constant120 : [#users=1] = get_attr[target=_param_constant120]
%unsqueeze_150 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant120, 0), kwargs = {})
%unsqueeze_151 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_150, 2), kwargs = {})
%unsqueeze_152 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_151, 3), kwargs = {})
%_param_constant121 : [#users=1] = get_attr[target=_param_constant121]
%unsqueeze_153 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant121, 0), kwargs = {})
%unsqueeze_154 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_153, 2), kwargs = {})
%unsqueeze_155 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_154, 3), kwargs = {})
%mul_52 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_69, %unsqueeze_155), kwargs = {})
%add_67 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_52, %unsqueeze_152), kwargs = {})
%squeeze_100 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_51, 3), kwargs = {})
%squeeze_101 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_100, 2), kwargs = {})
%squeeze_102 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_25, 3), kwargs = {})
%squeeze_103 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_102, 2), kwargs = {})
%detach_51 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_101,), kwargs = {})
%detach_52 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_103,), kwargs = {})
%silu_24 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_67,), kwargs = {})
%_param_constant122 : [#users=1] = get_attr[target=_param_constant122]
%_param_constant123 : [#users=1] = get_attr[target=_param_constant123]
%convolution_31 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_24, %_param_constant122, %_param_constant123, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_70 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_31, [1, 32, 4, 262144]), kwargs = {})
%convert_element_type_26 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_70, torch.float32), kwargs = {})
%var_mean_26 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_26, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_52 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_26, 0), kwargs = {})
%_to_copy_52 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_52,), kwargs = {dtype: torch.float16})
%getitem_53 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_26, 1), kwargs = {})
%_to_copy_53 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_53,), kwargs = {dtype: torch.float16})
%add_68 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_52, 1e-06), kwargs = {})
%rsqrt_26 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_68,), kwargs = {})
%sub_26 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_70, %_to_copy_53), kwargs = {})
%mul_53 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_26, %rsqrt_26), kwargs = {})
%view_71 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_53, [1, 128, 512, 512]), kwargs = {})
%_param_constant124 : [#users=1] = get_attr[target=_param_constant124]
%unsqueeze_156 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant124, 0), kwargs = {})
%unsqueeze_157 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_156, 2), kwargs = {})
%unsqueeze_158 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_157, 3), kwargs = {})
%_param_constant125 : [#users=1] = get_attr[target=_param_constant125]
%unsqueeze_159 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant125, 0), kwargs = {})
%unsqueeze_160 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_159, 2), kwargs = {})
%unsqueeze_161 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_160, 3), kwargs = {})
%mul_54 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_71, %unsqueeze_161), kwargs = {})
%add_69 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_54, %unsqueeze_158), kwargs = {})
%squeeze_104 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_53, 3), kwargs = {})
%squeeze_105 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_104, 2), kwargs = {})
%squeeze_106 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_26, 3), kwargs = {})
%squeeze_107 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_106, 2), kwargs = {})
%detach_53 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_105,), kwargs = {})
%detach_54 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_107,), kwargs = {})
%silu_25 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_69,), kwargs = {})
%_param_constant126 : [#users=1] = get_attr[target=_param_constant126]
%_param_constant127 : [#users=1] = get_attr[target=_param_constant127]
%convolution_32 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_25, %_param_constant126, %_param_constant127, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_70 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_12, %convolution_32), kwargs = {})
%div_13 : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_70, 1.0), kwargs = {})
%view_72 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%div_13, [1, 32, 4, 262144]), kwargs = {})
%convert_element_type_27 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_72, torch.float32), kwargs = {})
%var_mean_27 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_27, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_54 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_27, 0), kwargs = {})
%_to_copy_54 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_54,), kwargs = {dtype: torch.float16})
%getitem_55 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_27, 1), kwargs = {})
%_to_copy_55 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_55,), kwargs = {dtype: torch.float16})
%add_71 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_54, 1e-06), kwargs = {})
%rsqrt_27 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_71,), kwargs = {})
%sub_27 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_72, %_to_copy_55), kwargs = {})
%mul_55 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_27, %rsqrt_27), kwargs = {})
%view_73 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_55, [1, 128, 512, 512]), kwargs = {})
%_param_constant128 : [#users=1] = get_attr[target=_param_constant128]
%unsqueeze_162 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant128, 0), kwargs = {})
%unsqueeze_163 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_162, 2), kwargs = {})
%unsqueeze_164 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_163, 3), kwargs = {})
%_param_constant129 : [#users=1] = get_attr[target=_param_constant129]
%unsqueeze_165 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant129, 0), kwargs = {})
%unsqueeze_166 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_165, 2), kwargs = {})
%unsqueeze_167 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_166, 3), kwargs = {})
%mul_56 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_73, %unsqueeze_167), kwargs = {})
%add_72 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_56, %unsqueeze_164), kwargs = {})
%squeeze_108 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_55, 3), kwargs = {})
%squeeze_109 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_108, 2), kwargs = {})
%squeeze_110 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_27, 3), kwargs = {})
%squeeze_111 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_110, 2), kwargs = {})
%detach_55 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_109,), kwargs = {})
%detach_56 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_111,), kwargs = {})
%silu_26 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_72,), kwargs = {})
%_param_constant130 : [#users=1] = get_attr[target=_param_constant130]
%_param_constant131 : [#users=1] = get_attr[target=_param_constant131]
%convolution_33 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_26, %_param_constant130, %_param_constant131, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_74 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_33, [1, 32, 4, 262144]), kwargs = {})
%convert_element_type_28 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_74, torch.float32), kwargs = {})
%var_mean_28 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_28, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_56 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_28, 0), kwargs = {})
%_to_copy_56 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_56,), kwargs = {dtype: torch.float16})
%getitem_57 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_28, 1), kwargs = {})
%_to_copy_57 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_57,), kwargs = {dtype: torch.float16})
%add_73 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_56, 1e-06), kwargs = {})
%rsqrt_28 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_73,), kwargs = {})
%sub_28 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_74, %_to_copy_57), kwargs = {})
%mul_57 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_28, %rsqrt_28), kwargs = {})
%view_75 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_57, [1, 128, 512, 512]), kwargs = {})
%_param_constant132 : [#users=1] = get_attr[target=_param_constant132]
%unsqueeze_168 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant132, 0), kwargs = {})
%unsqueeze_169 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_168, 2), kwargs = {})
%unsqueeze_170 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_169, 3), kwargs = {})
%_param_constant133 : [#users=1] = get_attr[target=_param_constant133]
%unsqueeze_171 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant133, 0), kwargs = {})
%unsqueeze_172 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_171, 2), kwargs = {})
%unsqueeze_173 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_172, 3), kwargs = {})
%mul_58 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_75, %unsqueeze_173), kwargs = {})
%add_74 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_58, %unsqueeze_170), kwargs = {})
%squeeze_112 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_57, 3), kwargs = {})
%squeeze_113 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_112, 2), kwargs = {})
%squeeze_114 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_28, 3), kwargs = {})
%squeeze_115 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_114, 2), kwargs = {})
%detach_57 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_113,), kwargs = {})
%detach_58 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_115,), kwargs = {})
%silu_27 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_74,), kwargs = {})
%_param_constant134 : [#users=1] = get_attr[target=_param_constant134]
%_param_constant135 : [#users=1] = get_attr[target=_param_constant135]
%convolution_34 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_27, %_param_constant134, %_param_constant135, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_75 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_13, %convolution_34), kwargs = {})
%div_14 : [#users=1] = call_function[target=torch.ops.aten.div](args = (%add_75, 1.0), kwargs = {})
%view_76 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%div_14, [1, 32, 4, 262144]), kwargs = {})
%convert_element_type_29 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_76, torch.float32), kwargs = {})
%var_mean_29 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_29, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_58 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_29, 0), kwargs = {})
%_to_copy_58 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_58,), kwargs = {dtype: torch.float16})
%getitem_59 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_29, 1), kwargs = {})
%_to_copy_59 : [#users=2] = call_function[target=torch.ops.aten._to_copy](args = (%getitem_59,), kwargs = {dtype: torch.float16})
%add_76 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_to_copy_58, 1e-06), kwargs = {})
%rsqrt_29 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_76,), kwargs = {})
%sub_29 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_76, %_to_copy_59), kwargs = {})
%mul_59 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_29, %rsqrt_29), kwargs = {})
%view_77 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_59, [1, 128, 512, 512]), kwargs = {})
%_param_constant136 : [#users=1] = get_attr[target=_param_constant136]
%unsqueeze_174 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant136, 0), kwargs = {})
%unsqueeze_175 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_174, 2), kwargs = {})
%unsqueeze_176 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_175, 3), kwargs = {})
%_param_constant137 : [#users=1] = get_attr[target=_param_constant137]
%unsqueeze_177 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant137, 0), kwargs = {})
%unsqueeze_178 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_177, 2), kwargs = {})
%unsqueeze_179 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_178, 3), kwargs = {})
%mul_60 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_77, %unsqueeze_179), kwargs = {})
%add_77 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_60, %unsqueeze_176), kwargs = {})
%squeeze_116 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%_to_copy_59, 3), kwargs = {})
%squeeze_117 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_116, 2), kwargs = {})
%squeeze_118 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%rsqrt_29, 3), kwargs = {})
%squeeze_119 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_118, 2), kwargs = {})
%detach_59 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_117,), kwargs = {})
%detach_60 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_119,), kwargs = {})
%silu_28 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%add_77,), kwargs = {})
%_param_constant138 : [#users=1] = get_attr[target=_param_constant138]
%_param_constant139 : [#users=1] = get_attr[target=_param_constant139]
%convolution_35 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_28, %_param_constant138, %_param_constant139, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%div_15 : [#users=1] = call_function[target=torch.ops.aten.div](args = (%convolution_35, 2), kwargs = {})
%add_78 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_15, 0.5), kwargs = {})
%clamp : [#users=1] = call_function[target=torch.ops.aten.clamp](args = (%add_78, 0, 1), kwargs = {})
%mul_61 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%clamp, 255.0), kwargs = {})
%round_1 : [#users=1] = call_function[target=torch.ops.aten.round](args = (%mul_61,), kwargs = {})
return round_1
graph():
%arg0_1 : [#users=1] = placeholder[target=arg0_1]
%mul : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%arg0_1, 5.489980785067252), kwargs = {})
%_param_constant0 : [#users=1] = get_attr[target=_param_constant0]
%_param_constant1 : [#users=1] = get_attr[target=_param_constant1]
%convolution : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%mul, %_param_constant0, %_param_constant1, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
%_param_constant2 : [#users=1] = get_attr[target=_param_constant2]
%_param_constant3 : [#users=1] = get_attr[target=_param_constant3]
%convolution_1 : [#users=2] = call_function[target=torch.ops.aten.convolution](args = (%convolution, %_param_constant2, %_param_constant3, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_1, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view, torch.float32), kwargs = {})
%var_mean : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem : [#users=1] = call_function[target=operator.getitem](args = (%var_mean, 0), kwargs = {})
%getitem_1 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean, 1), kwargs = {})
%add : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem, 1e-06), kwargs = {})
%rsqrt : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add,), kwargs = {})
%sub : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view, %getitem_1), kwargs = {})
%mul_1 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub, %rsqrt), kwargs = {})
%view_1 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_1, [1, 512, 64, 64]), kwargs = {})
%_param_constant4 : [#users=1] = get_attr[target=_param_constant4]
%unsqueeze : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant4, 0), kwargs = {})
%unsqueeze_1 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze, 2), kwargs = {})
%unsqueeze_2 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_1, 3), kwargs = {})
%_param_constant5 : [#users=1] = get_attr[target=_param_constant5]
%unsqueeze_3 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant5, 0), kwargs = {})
%unsqueeze_4 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_3, 2), kwargs = {})
%unsqueeze_5 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_4, 3), kwargs = {})
%mul_2 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_1, %unsqueeze_5), kwargs = {})
%add_1 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_2, %unsqueeze_2), kwargs = {})
%convert_element_type_1 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_1, torch.float16), kwargs = {})
%convert_element_type_2 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_1, torch.float16), kwargs = {})
%convert_element_type_3 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt, torch.float16), kwargs = {})
%squeeze : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_2, 3), kwargs = {})
%squeeze_1 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze, 2), kwargs = {})
%squeeze_2 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_3, 3), kwargs = {})
%squeeze_3 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_2, 2), kwargs = {})
%detach : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_1,), kwargs = {})
%detach_1 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_3,), kwargs = {})
%silu : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_1,), kwargs = {})
%_param_constant6 : [#users=1] = get_attr[target=_param_constant6]
%_param_constant7 : [#users=1] = get_attr[target=_param_constant7]
%convolution_2 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu, %_param_constant6, %_param_constant7, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_2 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_2, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type_4 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_2, torch.float32), kwargs = {})
%var_mean_1 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_4, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_2 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_1, 0), kwargs = {})
%getitem_3 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_1, 1), kwargs = {})
%add_2 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_2, 1e-06), kwargs = {})
%rsqrt_1 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_2,), kwargs = {})
%sub_1 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_2, %getitem_3), kwargs = {})
%mul_3 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_1, %rsqrt_1), kwargs = {})
%view_3 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_3, [1, 512, 64, 64]), kwargs = {})
%_param_constant8 : [#users=1] = get_attr[target=_param_constant8]
%unsqueeze_6 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant8, 0), kwargs = {})
%unsqueeze_7 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_6, 2), kwargs = {})
%unsqueeze_8 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_7, 3), kwargs = {})
%_param_constant9 : [#users=1] = get_attr[target=_param_constant9]
%unsqueeze_9 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant9, 0), kwargs = {})
%unsqueeze_10 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_9, 2), kwargs = {})
%unsqueeze_11 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_10, 3), kwargs = {})
%mul_4 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_3, %unsqueeze_11), kwargs = {})
%add_3 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_4, %unsqueeze_8), kwargs = {})
%convert_element_type_5 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_3, torch.float16), kwargs = {})
%convert_element_type_6 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_3, torch.float16), kwargs = {})
%convert_element_type_7 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_1, torch.float16), kwargs = {})
%squeeze_4 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_6, 3), kwargs = {})
%squeeze_5 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_4, 2), kwargs = {})
%squeeze_6 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_7, 3), kwargs = {})
%squeeze_7 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_6, 2), kwargs = {})
%detach_2 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_5,), kwargs = {})
%detach_3 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_7,), kwargs = {})
%silu_1 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_5,), kwargs = {})
%_param_constant10 : [#users=1] = get_attr[target=_param_constant10]
%_param_constant11 : [#users=1] = get_attr[target=_param_constant11]
%convolution_3 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_1, %_param_constant10, %_param_constant11, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_4 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%convolution_1, %convolution_3), kwargs = {})
%div : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_4, 1), kwargs = {})
%view_4 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%div, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type_8 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_4, torch.float32), kwargs = {})
%var_mean_2 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_8, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_4 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_2, 0), kwargs = {})
%getitem_5 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_2, 1), kwargs = {})
%add_5 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_4, 1e-06), kwargs = {})
%rsqrt_2 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_5,), kwargs = {})
%sub_2 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_4, %getitem_5), kwargs = {})
%mul_5 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_2, %rsqrt_2), kwargs = {})
%view_5 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_5, [1, 512, 64, 64]), kwargs = {})
%_param_constant12 : [#users=1] = get_attr[target=_param_constant12]
%unsqueeze_12 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant12, 0), kwargs = {})
%unsqueeze_13 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_12, 2), kwargs = {})
%unsqueeze_14 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_13, 3), kwargs = {})
%_param_constant13 : [#users=1] = get_attr[target=_param_constant13]
%unsqueeze_15 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant13, 0), kwargs = {})
%unsqueeze_16 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_15, 2), kwargs = {})
%unsqueeze_17 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_16, 3), kwargs = {})
%mul_6 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_5, %unsqueeze_17), kwargs = {})
%add_6 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_6, %unsqueeze_14), kwargs = {})
%convert_element_type_9 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_6, torch.float16), kwargs = {})
%convert_element_type_10 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_5, torch.float16), kwargs = {})
%convert_element_type_11 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_2, torch.float16), kwargs = {})
%squeeze_8 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_10, 3), kwargs = {})
%squeeze_9 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_8, 2), kwargs = {})
%squeeze_10 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_11, 3), kwargs = {})
%squeeze_11 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_10, 2), kwargs = {})
%detach_4 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_9,), kwargs = {})
%detach_5 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_11,), kwargs = {})
%view_6 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%convert_element_type_9, [1, 512, 4096]), kwargs = {})
%transpose : [#users=3] = call_function[target=torch.ops.aten.transpose](args = (%view_6, 1, 2), kwargs = {})
%_param_constant14 : [#users=1] = get_attr[target=_param_constant14]
%t : [#users=1] = call_function[target=torch.ops.aten.t](args = (%_param_constant14,), kwargs = {})
%expand : [#users=1] = call_function[target=torch.ops.aten.expand](args = (%transpose, [1, 4096, 512]), kwargs = {})
%view_7 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%expand, [1, 4096, 512]), kwargs = {})
%expand_1 : [#users=1] = call_function[target=torch.ops.aten.expand](args = (%t, [1, 512, 512]), kwargs = {})
%view_8 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%expand_1, [1, 512, 512]), kwargs = {})
%bmm : [#users=1] = call_function[target=torch.ops.aten.bmm](args = (%view_7, %view_8), kwargs = {})
%_unsafe_view : [#users=1] = call_function[target=torch.ops.aten._unsafe_view](args = (%bmm, [1, 4096, 512]), kwargs = {})
%_param_constant15 : [#users=1] = get_attr[target=_param_constant15]
%add_7 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_unsafe_view, %_param_constant15), kwargs = {})
%_param_constant16 : [#users=1] = get_attr[target=_param_constant16]
%t_1 : [#users=1] = call_function[target=torch.ops.aten.t](args = (%_param_constant16,), kwargs = {})
%expand_2 : [#users=1] = call_function[target=torch.ops.aten.expand](args = (%transpose, [1, 4096, 512]), kwargs = {})
%view_9 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%expand_2, [1, 4096, 512]), kwargs = {})
%expand_3 : [#users=1] = call_function[target=torch.ops.aten.expand](args = (%t_1, [1, 512, 512]), kwargs = {})
%view_10 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%expand_3, [1, 512, 512]), kwargs = {})
%bmm_1 : [#users=1] = call_function[target=torch.ops.aten.bmm](args = (%view_9, %view_10), kwargs = {})
%_unsafe_view_1 : [#users=1] = call_function[target=torch.ops.aten._unsafe_view](args = (%bmm_1, [1, 4096, 512]), kwargs = {})
%_param_constant17 : [#users=1] = get_attr[target=_param_constant17]
%add_8 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_unsafe_view_1, %_param_constant17), kwargs = {})
%_param_constant18 : [#users=1] = get_attr[target=_param_constant18]
%t_2 : [#users=1] = call_function[target=torch.ops.aten.t](args = (%_param_constant18,), kwargs = {})
%expand_4 : [#users=1] = call_function[target=torch.ops.aten.expand](args = (%transpose, [1, 4096, 512]), kwargs = {})
%view_11 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%expand_4, [1, 4096, 512]), kwargs = {})
%expand_5 : [#users=1] = call_function[target=torch.ops.aten.expand](args = (%t_2, [1, 512, 512]), kwargs = {})
%view_12 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%expand_5, [1, 512, 512]), kwargs = {})
%bmm_2 : [#users=1] = call_function[target=torch.ops.aten.bmm](args = (%view_11, %view_12), kwargs = {})
%_unsafe_view_2 : [#users=1] = call_function[target=torch.ops.aten._unsafe_view](args = (%bmm_2, [1, 4096, 512]), kwargs = {})
%_param_constant19 : [#users=1] = get_attr[target=_param_constant19]
%add_9 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%_unsafe_view_2, %_param_constant19), kwargs = {})
%view_13 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%add_7, [1, 4096, 1, 512]), kwargs = {})
%permute : [#users=1] = call_function[target=torch.ops.aten.permute](args = (%view_13, [0, 2, 1, 3]), kwargs = {})
%view_14 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%permute, [1, 4096, 512]), kwargs = {})
%view_15 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%add_8, [1, 4096, 1, 512]), kwargs = {})
%permute_1 : [#users=1] = call_function[target=torch.ops.aten.permute](args = (%view_15, [0, 2, 1, 3]), kwargs = {})
%view_16 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%permute_1, [1, 4096, 512]), kwargs = {})
%view_17 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%add_9, [1, 4096, 1, 512]), kwargs = {})
%permute_2 : [#users=1] = call_function[target=torch.ops.aten.permute](args = (%view_17, [0, 2, 1, 3]), kwargs = {})
%view_18 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%permute_2, [1, 4096, 512]), kwargs = {})
%empty : [#users=1] = call_function[target=torch.ops.aten.empty](args = ([1, 4096, 4096],), kwargs = {dtype: torch.float16, device: cuda:0, pin_memory: False})
%transpose_1 : [#users=1] = call_function[target=torch.ops.aten.transpose](args = (%view_16, -1, -2), kwargs = {})
%baddbmm : [#users=1] = call_function[target=torch.ops.aten.baddbmm](args = (%empty, %view_14, %transpose_1), kwargs = {beta: 0, alpha: 0.044194173824159216})
%_to_copy : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%baddbmm,), kwargs = {dtype: torch.float32})
%_softmax : [#users=2] = call_function[target=torch.ops.aten._softmax](args = (%_to_copy, -1, False), kwargs = {})
%detach_6 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%_softmax,), kwargs = {})
%_to_copy_1 : [#users=1] = call_function[target=torch.ops.aten._to_copy](args = (%_softmax,), kwargs = {dtype: torch.float16, device: cuda:0})
%bmm_3 : [#users=1] = call_function[target=torch.ops.aten.bmm](args = (%_to_copy_1, %view_18), kwargs = {})
%view_19 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%bmm_3, [1, 1, 4096, 512]), kwargs = {})
%permute_3 : [#users=1] = call_function[target=torch.ops.aten.permute](args = (%view_19, [0, 2, 1, 3]), kwargs = {})
%view_20 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%permute_3, [1, 4096, 512]), kwargs = {})
%_param_constant20 : [#users=1] = get_attr[target=_param_constant20]
%t_3 : [#users=1] = call_function[target=torch.ops.aten.t](args = (%_param_constant20,), kwargs = {})
%view_21 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%view_20, [4096, 512]), kwargs = {})
%_param_constant21 : [#users=1] = get_attr[target=_param_constant21]
%addmm : [#users=1] = call_function[target=torch.ops.aten.addmm](args = (%_param_constant21, %view_21, %t_3), kwargs = {})
%view_22 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%addmm, [1, 4096, 512]), kwargs = {})
%transpose_2 : [#users=1] = call_function[target=torch.ops.aten.transpose](args = (%view_22, -1, -2), kwargs = {})
%view_23 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%transpose_2, [1, 512, 64, 64]), kwargs = {})
%add_10 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%view_23, %div), kwargs = {})
%div_1 : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_10, 1), kwargs = {})
%clone : [#users=1] = call_function[target=torch.ops.aten.clone](args = (%div_1,), kwargs = {memory_format: torch.contiguous_format})
%view_24 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%clone, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type_12 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_24, torch.float32), kwargs = {})
%var_mean_3 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_12, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_6 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_3, 0), kwargs = {})
%getitem_7 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_3, 1), kwargs = {})
%add_11 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_6, 1e-06), kwargs = {})
%rsqrt_3 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_11,), kwargs = {})
%sub_3 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_24, %getitem_7), kwargs = {})
%mul_7 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_3, %rsqrt_3), kwargs = {})
%view_25 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_7, [1, 512, 64, 64]), kwargs = {})
%_param_constant22 : [#users=1] = get_attr[target=_param_constant22]
%unsqueeze_18 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant22, 0), kwargs = {})
%unsqueeze_19 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_18, 2), kwargs = {})
%unsqueeze_20 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_19, 3), kwargs = {})
%_param_constant23 : [#users=1] = get_attr[target=_param_constant23]
%unsqueeze_21 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant23, 0), kwargs = {})
%unsqueeze_22 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_21, 2), kwargs = {})
%unsqueeze_23 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_22, 3), kwargs = {})
%mul_8 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_25, %unsqueeze_23), kwargs = {})
%add_12 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_8, %unsqueeze_20), kwargs = {})
%convert_element_type_13 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_12, torch.float16), kwargs = {})
%convert_element_type_14 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_7, torch.float16), kwargs = {})
%convert_element_type_15 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_3, torch.float16), kwargs = {})
%squeeze_12 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_14, 3), kwargs = {})
%squeeze_13 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_12, 2), kwargs = {})
%squeeze_14 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_15, 3), kwargs = {})
%squeeze_15 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_14, 2), kwargs = {})
%detach_7 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_13,), kwargs = {})
%detach_8 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_15,), kwargs = {})
%silu_2 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_13,), kwargs = {})
%_param_constant24 : [#users=1] = get_attr[target=_param_constant24]
%_param_constant25 : [#users=1] = get_attr[target=_param_constant25]
%convolution_4 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_2, %_param_constant24, %_param_constant25, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_26 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_4, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type_16 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_26, torch.float32), kwargs = {})
%var_mean_4 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_16, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_8 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_4, 0), kwargs = {})
%getitem_9 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_4, 1), kwargs = {})
%add_13 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_8, 1e-06), kwargs = {})
%rsqrt_4 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_13,), kwargs = {})
%sub_4 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_26, %getitem_9), kwargs = {})
%mul_9 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_4, %rsqrt_4), kwargs = {})
%view_27 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_9, [1, 512, 64, 64]), kwargs = {})
%_param_constant26 : [#users=1] = get_attr[target=_param_constant26]
%unsqueeze_24 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant26, 0), kwargs = {})
%unsqueeze_25 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_24, 2), kwargs = {})
%unsqueeze_26 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_25, 3), kwargs = {})
%_param_constant27 : [#users=1] = get_attr[target=_param_constant27]
%unsqueeze_27 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant27, 0), kwargs = {})
%unsqueeze_28 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_27, 2), kwargs = {})
%unsqueeze_29 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_28, 3), kwargs = {})
%mul_10 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_27, %unsqueeze_29), kwargs = {})
%add_14 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_10, %unsqueeze_26), kwargs = {})
%convert_element_type_17 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_14, torch.float16), kwargs = {})
%convert_element_type_18 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_9, torch.float16), kwargs = {})
%convert_element_type_19 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_4, torch.float16), kwargs = {})
%squeeze_16 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_18, 3), kwargs = {})
%squeeze_17 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_16, 2), kwargs = {})
%squeeze_18 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_19, 3), kwargs = {})
%squeeze_19 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_18, 2), kwargs = {})
%detach_9 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_17,), kwargs = {})
%detach_10 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_19,), kwargs = {})
%silu_3 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_17,), kwargs = {})
%_param_constant28 : [#users=1] = get_attr[target=_param_constant28]
%_param_constant29 : [#users=1] = get_attr[target=_param_constant29]
%convolution_5 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_3, %_param_constant28, %_param_constant29, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_15 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_1, %convolution_5), kwargs = {})
%div_2 : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_15, 1), kwargs = {})
%clone_1 : [#users=1] = call_function[target=torch.ops.aten.clone](args = (%div_2,), kwargs = {memory_format: torch.contiguous_format})
%view_28 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%clone_1, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type_20 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_28, torch.float32), kwargs = {})
%var_mean_5 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_20, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_10 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_5, 0), kwargs = {})
%getitem_11 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_5, 1), kwargs = {})
%add_16 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_10, 1e-06), kwargs = {})
%rsqrt_5 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_16,), kwargs = {})
%sub_5 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_28, %getitem_11), kwargs = {})
%mul_11 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_5, %rsqrt_5), kwargs = {})
%view_29 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_11, [1, 512, 64, 64]), kwargs = {})
%_param_constant30 : [#users=1] = get_attr[target=_param_constant30]
%unsqueeze_30 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant30, 0), kwargs = {})
%unsqueeze_31 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_30, 2), kwargs = {})
%unsqueeze_32 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_31, 3), kwargs = {})
%_param_constant31 : [#users=1] = get_attr[target=_param_constant31]
%unsqueeze_33 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant31, 0), kwargs = {})
%unsqueeze_34 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_33, 2), kwargs = {})
%unsqueeze_35 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_34, 3), kwargs = {})
%mul_12 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_29, %unsqueeze_35), kwargs = {})
%add_17 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_12, %unsqueeze_32), kwargs = {})
%convert_element_type_21 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_17, torch.float16), kwargs = {})
%convert_element_type_22 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_11, torch.float16), kwargs = {})
%convert_element_type_23 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_5, torch.float16), kwargs = {})
%squeeze_20 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_22, 3), kwargs = {})
%squeeze_21 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_20, 2), kwargs = {})
%squeeze_22 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_23, 3), kwargs = {})
%squeeze_23 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_22, 2), kwargs = {})
%detach_11 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_21,), kwargs = {})
%detach_12 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_23,), kwargs = {})
%silu_4 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_21,), kwargs = {})
%_param_constant32 : [#users=1] = get_attr[target=_param_constant32]
%_param_constant33 : [#users=1] = get_attr[target=_param_constant33]
%convolution_6 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_4, %_param_constant32, %_param_constant33, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_30 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_6, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type_24 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_30, torch.float32), kwargs = {})
%var_mean_6 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_24, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_12 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_6, 0), kwargs = {})
%getitem_13 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_6, 1), kwargs = {})
%add_18 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_12, 1e-06), kwargs = {})
%rsqrt_6 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_18,), kwargs = {})
%sub_6 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_30, %getitem_13), kwargs = {})
%mul_13 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_6, %rsqrt_6), kwargs = {})
%view_31 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_13, [1, 512, 64, 64]), kwargs = {})
%_param_constant34 : [#users=1] = get_attr[target=_param_constant34]
%unsqueeze_36 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant34, 0), kwargs = {})
%unsqueeze_37 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_36, 2), kwargs = {})
%unsqueeze_38 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_37, 3), kwargs = {})
%_param_constant35 : [#users=1] = get_attr[target=_param_constant35]
%unsqueeze_39 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant35, 0), kwargs = {})
%unsqueeze_40 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_39, 2), kwargs = {})
%unsqueeze_41 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_40, 3), kwargs = {})
%mul_14 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_31, %unsqueeze_41), kwargs = {})
%add_19 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_14, %unsqueeze_38), kwargs = {})
%convert_element_type_25 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_19, torch.float16), kwargs = {})
%convert_element_type_26 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_13, torch.float16), kwargs = {})
%convert_element_type_27 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_6, torch.float16), kwargs = {})
%squeeze_24 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_26, 3), kwargs = {})
%squeeze_25 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_24, 2), kwargs = {})
%squeeze_26 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_27, 3), kwargs = {})
%squeeze_27 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_26, 2), kwargs = {})
%detach_13 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_25,), kwargs = {})
%detach_14 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_27,), kwargs = {})
%silu_5 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_25,), kwargs = {})
%_param_constant36 : [#users=1] = get_attr[target=_param_constant36]
%_param_constant37 : [#users=1] = get_attr[target=_param_constant37]
%convolution_7 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_5, %_param_constant36, %_param_constant37, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_20 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_2, %convolution_7), kwargs = {})
%div_3 : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_20, 1.0), kwargs = {})
%clone_2 : [#users=1] = call_function[target=torch.ops.aten.clone](args = (%div_3,), kwargs = {memory_format: torch.contiguous_format})
%view_32 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%clone_2, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type_28 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_32, torch.float32), kwargs = {})
%var_mean_7 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_28, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_14 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_7, 0), kwargs = {})
%getitem_15 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_7, 1), kwargs = {})
%add_21 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_14, 1e-06), kwargs = {})
%rsqrt_7 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_21,), kwargs = {})
%sub_7 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_32, %getitem_15), kwargs = {})
%mul_15 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_7, %rsqrt_7), kwargs = {})
%view_33 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_15, [1, 512, 64, 64]), kwargs = {})
%_param_constant38 : [#users=1] = get_attr[target=_param_constant38]
%unsqueeze_42 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant38, 0), kwargs = {})
%unsqueeze_43 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_42, 2), kwargs = {})
%unsqueeze_44 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_43, 3), kwargs = {})
%_param_constant39 : [#users=1] = get_attr[target=_param_constant39]
%unsqueeze_45 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant39, 0), kwargs = {})
%unsqueeze_46 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_45, 2), kwargs = {})
%unsqueeze_47 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_46, 3), kwargs = {})
%mul_16 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_33, %unsqueeze_47), kwargs = {})
%add_22 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_16, %unsqueeze_44), kwargs = {})
%convert_element_type_29 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_22, torch.float16), kwargs = {})
%convert_element_type_30 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_15, torch.float16), kwargs = {})
%convert_element_type_31 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_7, torch.float16), kwargs = {})
%squeeze_28 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_30, 3), kwargs = {})
%squeeze_29 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_28, 2), kwargs = {})
%squeeze_30 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_31, 3), kwargs = {})
%squeeze_31 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_30, 2), kwargs = {})
%detach_15 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_29,), kwargs = {})
%detach_16 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_31,), kwargs = {})
%silu_6 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_29,), kwargs = {})
%_param_constant40 : [#users=1] = get_attr[target=_param_constant40]
%_param_constant41 : [#users=1] = get_attr[target=_param_constant41]
%convolution_8 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_6, %_param_constant40, %_param_constant41, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_34 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_8, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type_32 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_34, torch.float32), kwargs = {})
%var_mean_8 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_32, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_16 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_8, 0), kwargs = {})
%getitem_17 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_8, 1), kwargs = {})
%add_23 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_16, 1e-06), kwargs = {})
%rsqrt_8 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_23,), kwargs = {})
%sub_8 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_34, %getitem_17), kwargs = {})
%mul_17 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_8, %rsqrt_8), kwargs = {})
%view_35 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_17, [1, 512, 64, 64]), kwargs = {})
%_param_constant42 : [#users=1] = get_attr[target=_param_constant42]
%unsqueeze_48 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant42, 0), kwargs = {})
%unsqueeze_49 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_48, 2), kwargs = {})
%unsqueeze_50 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_49, 3), kwargs = {})
%_param_constant43 : [#users=1] = get_attr[target=_param_constant43]
%unsqueeze_51 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant43, 0), kwargs = {})
%unsqueeze_52 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_51, 2), kwargs = {})
%unsqueeze_53 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_52, 3), kwargs = {})
%mul_18 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_35, %unsqueeze_53), kwargs = {})
%add_24 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_18, %unsqueeze_50), kwargs = {})
%convert_element_type_33 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_24, torch.float16), kwargs = {})
%convert_element_type_34 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_17, torch.float16), kwargs = {})
%convert_element_type_35 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_8, torch.float16), kwargs = {})
%squeeze_32 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_34, 3), kwargs = {})
%squeeze_33 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_32, 2), kwargs = {})
%squeeze_34 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_35, 3), kwargs = {})
%squeeze_35 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_34, 2), kwargs = {})
%detach_17 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_33,), kwargs = {})
%detach_18 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_35,), kwargs = {})
%silu_7 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_33,), kwargs = {})
%_param_constant44 : [#users=1] = get_attr[target=_param_constant44]
%_param_constant45 : [#users=1] = get_attr[target=_param_constant45]
%convolution_9 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_7, %_param_constant44, %_param_constant45, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_25 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_3, %convolution_9), kwargs = {})
%div_4 : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_25, 1.0), kwargs = {})
%clone_3 : [#users=1] = call_function[target=torch.ops.aten.clone](args = (%div_4,), kwargs = {memory_format: torch.contiguous_format})
%view_36 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%clone_3, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type_36 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_36, torch.float32), kwargs = {})
%var_mean_9 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_36, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_18 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_9, 0), kwargs = {})
%getitem_19 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_9, 1), kwargs = {})
%add_26 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_18, 1e-06), kwargs = {})
%rsqrt_9 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_26,), kwargs = {})
%sub_9 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_36, %getitem_19), kwargs = {})
%mul_19 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_9, %rsqrt_9), kwargs = {})
%view_37 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_19, [1, 512, 64, 64]), kwargs = {})
%_param_constant46 : [#users=1] = get_attr[target=_param_constant46]
%unsqueeze_54 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant46, 0), kwargs = {})
%unsqueeze_55 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_54, 2), kwargs = {})
%unsqueeze_56 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_55, 3), kwargs = {})
%_param_constant47 : [#users=1] = get_attr[target=_param_constant47]
%unsqueeze_57 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant47, 0), kwargs = {})
%unsqueeze_58 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_57, 2), kwargs = {})
%unsqueeze_59 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_58, 3), kwargs = {})
%mul_20 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_37, %unsqueeze_59), kwargs = {})
%add_27 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_20, %unsqueeze_56), kwargs = {})
%convert_element_type_37 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_27, torch.float16), kwargs = {})
%convert_element_type_38 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_19, torch.float16), kwargs = {})
%convert_element_type_39 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_9, torch.float16), kwargs = {})
%squeeze_36 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_38, 3), kwargs = {})
%squeeze_37 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_36, 2), kwargs = {})
%squeeze_38 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_39, 3), kwargs = {})
%squeeze_39 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_38, 2), kwargs = {})
%detach_19 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_37,), kwargs = {})
%detach_20 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_39,), kwargs = {})
%silu_8 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_37,), kwargs = {})
%_param_constant48 : [#users=1] = get_attr[target=_param_constant48]
%_param_constant49 : [#users=1] = get_attr[target=_param_constant49]
%convolution_10 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_8, %_param_constant48, %_param_constant49, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_38 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_10, [1, 32, 16, 4096]), kwargs = {})
%convert_element_type_40 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_38, torch.float32), kwargs = {})
%var_mean_10 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_40, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_20 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_10, 0), kwargs = {})
%getitem_21 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_10, 1), kwargs = {})
%add_28 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_20, 1e-06), kwargs = {})
%rsqrt_10 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_28,), kwargs = {})
%sub_10 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_38, %getitem_21), kwargs = {})
%mul_21 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_10, %rsqrt_10), kwargs = {})
%view_39 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_21, [1, 512, 64, 64]), kwargs = {})
%_param_constant50 : [#users=1] = get_attr[target=_param_constant50]
%unsqueeze_60 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant50, 0), kwargs = {})
%unsqueeze_61 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_60, 2), kwargs = {})
%unsqueeze_62 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_61, 3), kwargs = {})
%_param_constant51 : [#users=1] = get_attr[target=_param_constant51]
%unsqueeze_63 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant51, 0), kwargs = {})
%unsqueeze_64 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_63, 2), kwargs = {})
%unsqueeze_65 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_64, 3), kwargs = {})
%mul_22 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_39, %unsqueeze_65), kwargs = {})
%add_29 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_22, %unsqueeze_62), kwargs = {})
%convert_element_type_41 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_29, torch.float16), kwargs = {})
%convert_element_type_42 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_21, torch.float16), kwargs = {})
%convert_element_type_43 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_10, torch.float16), kwargs = {})
%squeeze_40 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_42, 3), kwargs = {})
%squeeze_41 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_40, 2), kwargs = {})
%squeeze_42 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_43, 3), kwargs = {})
%squeeze_43 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_42, 2), kwargs = {})
%detach_21 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_41,), kwargs = {})
%detach_22 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_43,), kwargs = {})
%silu_9 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_41,), kwargs = {})
%_param_constant52 : [#users=1] = get_attr[target=_param_constant52]
%_param_constant53 : [#users=1] = get_attr[target=_param_constant53]
%convolution_11 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_9, %_param_constant52, %_param_constant53, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_30 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_4, %convolution_11), kwargs = {})
%div_5 : [#users=1] = call_function[target=torch.ops.aten.div](args = (%add_30, 1.0), kwargs = {})
%upsample_nearest2d : [#users=1] = call_function[target=torch.ops.aten.upsample_nearest2d](args = (%div_5, [128, 128], 2.0, 2.0), kwargs = {})
%_param_constant54 : [#users=1] = get_attr[target=_param_constant54]
%_param_constant55 : [#users=1] = get_attr[target=_param_constant55]
%convolution_12 : [#users=2] = call_function[target=torch.ops.aten.convolution](args = (%upsample_nearest2d, %_param_constant54, %_param_constant55, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_40 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_12, [1, 32, 16, 16384]), kwargs = {})
%convert_element_type_44 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_40, torch.float32), kwargs = {})
%var_mean_11 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_44, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_22 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_11, 0), kwargs = {})
%getitem_23 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_11, 1), kwargs = {})
%add_31 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_22, 1e-06), kwargs = {})
%rsqrt_11 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_31,), kwargs = {})
%sub_11 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_40, %getitem_23), kwargs = {})
%mul_23 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_11, %rsqrt_11), kwargs = {})
%view_41 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_23, [1, 512, 128, 128]), kwargs = {})
%_param_constant56 : [#users=1] = get_attr[target=_param_constant56]
%unsqueeze_66 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant56, 0), kwargs = {})
%unsqueeze_67 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_66, 2), kwargs = {})
%unsqueeze_68 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_67, 3), kwargs = {})
%_param_constant57 : [#users=1] = get_attr[target=_param_constant57]
%unsqueeze_69 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant57, 0), kwargs = {})
%unsqueeze_70 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_69, 2), kwargs = {})
%unsqueeze_71 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_70, 3), kwargs = {})
%mul_24 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_41, %unsqueeze_71), kwargs = {})
%add_32 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_24, %unsqueeze_68), kwargs = {})
%convert_element_type_45 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_32, torch.float16), kwargs = {})
%convert_element_type_46 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_23, torch.float16), kwargs = {})
%convert_element_type_47 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_11, torch.float16), kwargs = {})
%squeeze_44 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_46, 3), kwargs = {})
%squeeze_45 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_44, 2), kwargs = {})
%squeeze_46 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_47, 3), kwargs = {})
%squeeze_47 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_46, 2), kwargs = {})
%detach_23 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_45,), kwargs = {})
%detach_24 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_47,), kwargs = {})
%silu_10 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_45,), kwargs = {})
%_param_constant58 : [#users=1] = get_attr[target=_param_constant58]
%_param_constant59 : [#users=1] = get_attr[target=_param_constant59]
%convolution_13 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_10, %_param_constant58, %_param_constant59, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_42 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_13, [1, 32, 16, 16384]), kwargs = {})
%convert_element_type_48 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_42, torch.float32), kwargs = {})
%var_mean_12 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_48, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_24 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_12, 0), kwargs = {})
%getitem_25 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_12, 1), kwargs = {})
%add_33 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_24, 1e-06), kwargs = {})
%rsqrt_12 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_33,), kwargs = {})
%sub_12 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_42, %getitem_25), kwargs = {})
%mul_25 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_12, %rsqrt_12), kwargs = {})
%view_43 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_25, [1, 512, 128, 128]), kwargs = {})
%_param_constant60 : [#users=1] = get_attr[target=_param_constant60]
%unsqueeze_72 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant60, 0), kwargs = {})
%unsqueeze_73 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_72, 2), kwargs = {})
%unsqueeze_74 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_73, 3), kwargs = {})
%_param_constant61 : [#users=1] = get_attr[target=_param_constant61]
%unsqueeze_75 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant61, 0), kwargs = {})
%unsqueeze_76 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_75, 2), kwargs = {})
%unsqueeze_77 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_76, 3), kwargs = {})
%mul_26 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_43, %unsqueeze_77), kwargs = {})
%add_34 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_26, %unsqueeze_74), kwargs = {})
%convert_element_type_49 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_34, torch.float16), kwargs = {})
%convert_element_type_50 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_25, torch.float16), kwargs = {})
%convert_element_type_51 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_12, torch.float16), kwargs = {})
%squeeze_48 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_50, 3), kwargs = {})
%squeeze_49 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_48, 2), kwargs = {})
%squeeze_50 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_51, 3), kwargs = {})
%squeeze_51 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_50, 2), kwargs = {})
%detach_25 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_49,), kwargs = {})
%detach_26 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_51,), kwargs = {})
%silu_11 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_49,), kwargs = {})
%_param_constant62 : [#users=1] = get_attr[target=_param_constant62]
%_param_constant63 : [#users=1] = get_attr[target=_param_constant63]
%convolution_14 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_11, %_param_constant62, %_param_constant63, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_35 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%convolution_12, %convolution_14), kwargs = {})
%div_6 : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_35, 1.0), kwargs = {})
%view_44 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%div_6, [1, 32, 16, 16384]), kwargs = {})
%convert_element_type_52 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_44, torch.float32), kwargs = {})
%var_mean_13 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_52, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_26 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_13, 0), kwargs = {})
%getitem_27 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_13, 1), kwargs = {})
%add_36 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_26, 1e-06), kwargs = {})
%rsqrt_13 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_36,), kwargs = {})
%sub_13 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_44, %getitem_27), kwargs = {})
%mul_27 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_13, %rsqrt_13), kwargs = {})
%view_45 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_27, [1, 512, 128, 128]), kwargs = {})
%_param_constant64 : [#users=1] = get_attr[target=_param_constant64]
%unsqueeze_78 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant64, 0), kwargs = {})
%unsqueeze_79 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_78, 2), kwargs = {})
%unsqueeze_80 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_79, 3), kwargs = {})
%_param_constant65 : [#users=1] = get_attr[target=_param_constant65]
%unsqueeze_81 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant65, 0), kwargs = {})
%unsqueeze_82 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_81, 2), kwargs = {})
%unsqueeze_83 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_82, 3), kwargs = {})
%mul_28 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_45, %unsqueeze_83), kwargs = {})
%add_37 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_28, %unsqueeze_80), kwargs = {})
%convert_element_type_53 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_37, torch.float16), kwargs = {})
%convert_element_type_54 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_27, torch.float16), kwargs = {})
%convert_element_type_55 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_13, torch.float16), kwargs = {})
%squeeze_52 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_54, 3), kwargs = {})
%squeeze_53 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_52, 2), kwargs = {})
%squeeze_54 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_55, 3), kwargs = {})
%squeeze_55 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_54, 2), kwargs = {})
%detach_27 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_53,), kwargs = {})
%detach_28 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_55,), kwargs = {})
%silu_12 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_53,), kwargs = {})
%_param_constant66 : [#users=1] = get_attr[target=_param_constant66]
%_param_constant67 : [#users=1] = get_attr[target=_param_constant67]
%convolution_15 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_12, %_param_constant66, %_param_constant67, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_46 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_15, [1, 32, 16, 16384]), kwargs = {})
%convert_element_type_56 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_46, torch.float32), kwargs = {})
%var_mean_14 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_56, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_28 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_14, 0), kwargs = {})
%getitem_29 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_14, 1), kwargs = {})
%add_38 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_28, 1e-06), kwargs = {})
%rsqrt_14 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_38,), kwargs = {})
%sub_14 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_46, %getitem_29), kwargs = {})
%mul_29 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_14, %rsqrt_14), kwargs = {})
%view_47 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_29, [1, 512, 128, 128]), kwargs = {})
%_param_constant68 : [#users=1] = get_attr[target=_param_constant68]
%unsqueeze_84 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant68, 0), kwargs = {})
%unsqueeze_85 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_84, 2), kwargs = {})
%unsqueeze_86 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_85, 3), kwargs = {})
%_param_constant69 : [#users=1] = get_attr[target=_param_constant69]
%unsqueeze_87 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant69, 0), kwargs = {})
%unsqueeze_88 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_87, 2), kwargs = {})
%unsqueeze_89 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_88, 3), kwargs = {})
%mul_30 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_47, %unsqueeze_89), kwargs = {})
%add_39 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_30, %unsqueeze_86), kwargs = {})
%convert_element_type_57 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_39, torch.float16), kwargs = {})
%convert_element_type_58 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_29, torch.float16), kwargs = {})
%convert_element_type_59 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_14, torch.float16), kwargs = {})
%squeeze_56 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_58, 3), kwargs = {})
%squeeze_57 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_56, 2), kwargs = {})
%squeeze_58 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_59, 3), kwargs = {})
%squeeze_59 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_58, 2), kwargs = {})
%detach_29 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_57,), kwargs = {})
%detach_30 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_59,), kwargs = {})
%silu_13 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_57,), kwargs = {})
%_param_constant70 : [#users=1] = get_attr[target=_param_constant70]
%_param_constant71 : [#users=1] = get_attr[target=_param_constant71]
%convolution_16 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_13, %_param_constant70, %_param_constant71, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_40 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_6, %convolution_16), kwargs = {})
%div_7 : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_40, 1.0), kwargs = {})
%view_48 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%div_7, [1, 32, 16, 16384]), kwargs = {})
%convert_element_type_60 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_48, torch.float32), kwargs = {})
%var_mean_15 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_60, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_30 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_15, 0), kwargs = {})
%getitem_31 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_15, 1), kwargs = {})
%add_41 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_30, 1e-06), kwargs = {})
%rsqrt_15 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_41,), kwargs = {})
%sub_15 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_48, %getitem_31), kwargs = {})
%mul_31 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_15, %rsqrt_15), kwargs = {})
%view_49 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_31, [1, 512, 128, 128]), kwargs = {})
%_param_constant72 : [#users=1] = get_attr[target=_param_constant72]
%unsqueeze_90 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant72, 0), kwargs = {})
%unsqueeze_91 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_90, 2), kwargs = {})
%unsqueeze_92 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_91, 3), kwargs = {})
%_param_constant73 : [#users=1] = get_attr[target=_param_constant73]
%unsqueeze_93 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant73, 0), kwargs = {})
%unsqueeze_94 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_93, 2), kwargs = {})
%unsqueeze_95 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_94, 3), kwargs = {})
%mul_32 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_49, %unsqueeze_95), kwargs = {})
%add_42 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_32, %unsqueeze_92), kwargs = {})
%convert_element_type_61 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_42, torch.float16), kwargs = {})
%convert_element_type_62 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_31, torch.float16), kwargs = {})
%convert_element_type_63 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_15, torch.float16), kwargs = {})
%squeeze_60 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_62, 3), kwargs = {})
%squeeze_61 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_60, 2), kwargs = {})
%squeeze_62 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_63, 3), kwargs = {})
%squeeze_63 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_62, 2), kwargs = {})
%detach_31 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_61,), kwargs = {})
%detach_32 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_63,), kwargs = {})
%silu_14 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_61,), kwargs = {})
%_param_constant74 : [#users=1] = get_attr[target=_param_constant74]
%_param_constant75 : [#users=1] = get_attr[target=_param_constant75]
%convolution_17 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_14, %_param_constant74, %_param_constant75, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_50 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_17, [1, 32, 16, 16384]), kwargs = {})
%convert_element_type_64 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_50, torch.float32), kwargs = {})
%var_mean_16 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_64, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_32 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_16, 0), kwargs = {})
%getitem_33 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_16, 1), kwargs = {})
%add_43 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_32, 1e-06), kwargs = {})
%rsqrt_16 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_43,), kwargs = {})
%sub_16 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_50, %getitem_33), kwargs = {})
%mul_33 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_16, %rsqrt_16), kwargs = {})
%view_51 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_33, [1, 512, 128, 128]), kwargs = {})
%_param_constant76 : [#users=1] = get_attr[target=_param_constant76]
%unsqueeze_96 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant76, 0), kwargs = {})
%unsqueeze_97 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_96, 2), kwargs = {})
%unsqueeze_98 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_97, 3), kwargs = {})
%_param_constant77 : [#users=1] = get_attr[target=_param_constant77]
%unsqueeze_99 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant77, 0), kwargs = {})
%unsqueeze_100 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_99, 2), kwargs = {})
%unsqueeze_101 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_100, 3), kwargs = {})
%mul_34 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_51, %unsqueeze_101), kwargs = {})
%add_44 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_34, %unsqueeze_98), kwargs = {})
%convert_element_type_65 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_44, torch.float16), kwargs = {})
%convert_element_type_66 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_33, torch.float16), kwargs = {})
%convert_element_type_67 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_16, torch.float16), kwargs = {})
%squeeze_64 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_66, 3), kwargs = {})
%squeeze_65 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_64, 2), kwargs = {})
%squeeze_66 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_67, 3), kwargs = {})
%squeeze_67 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_66, 2), kwargs = {})
%detach_33 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_65,), kwargs = {})
%detach_34 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_67,), kwargs = {})
%silu_15 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_65,), kwargs = {})
%_param_constant78 : [#users=1] = get_attr[target=_param_constant78]
%_param_constant79 : [#users=1] = get_attr[target=_param_constant79]
%convolution_18 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_15, %_param_constant78, %_param_constant79, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_45 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_7, %convolution_18), kwargs = {})
%div_8 : [#users=1] = call_function[target=torch.ops.aten.div](args = (%add_45, 1.0), kwargs = {})
%upsample_nearest2d_1 : [#users=1] = call_function[target=torch.ops.aten.upsample_nearest2d](args = (%div_8, [256, 256], 2.0, 2.0), kwargs = {})
%_param_constant80 : [#users=1] = get_attr[target=_param_constant80]
%_param_constant81 : [#users=1] = get_attr[target=_param_constant81]
%convolution_19 : [#users=2] = call_function[target=torch.ops.aten.convolution](args = (%upsample_nearest2d_1, %_param_constant80, %_param_constant81, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_52 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_19, [1, 32, 16, 65536]), kwargs = {})
%convert_element_type_68 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_52, torch.float32), kwargs = {})
%var_mean_17 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_68, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_34 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_17, 0), kwargs = {})
%getitem_35 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_17, 1), kwargs = {})
%add_46 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_34, 1e-06), kwargs = {})
%rsqrt_17 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_46,), kwargs = {})
%sub_17 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_52, %getitem_35), kwargs = {})
%mul_35 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_17, %rsqrt_17), kwargs = {})
%view_53 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_35, [1, 512, 256, 256]), kwargs = {})
%_param_constant82 : [#users=1] = get_attr[target=_param_constant82]
%unsqueeze_102 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant82, 0), kwargs = {})
%unsqueeze_103 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_102, 2), kwargs = {})
%unsqueeze_104 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_103, 3), kwargs = {})
%_param_constant83 : [#users=1] = get_attr[target=_param_constant83]
%unsqueeze_105 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant83, 0), kwargs = {})
%unsqueeze_106 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_105, 2), kwargs = {})
%unsqueeze_107 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_106, 3), kwargs = {})
%mul_36 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_53, %unsqueeze_107), kwargs = {})
%add_47 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_36, %unsqueeze_104), kwargs = {})
%convert_element_type_69 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_47, torch.float16), kwargs = {})
%convert_element_type_70 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_35, torch.float16), kwargs = {})
%convert_element_type_71 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_17, torch.float16), kwargs = {})
%squeeze_68 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_70, 3), kwargs = {})
%squeeze_69 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_68, 2), kwargs = {})
%squeeze_70 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_71, 3), kwargs = {})
%squeeze_71 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_70, 2), kwargs = {})
%detach_35 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_69,), kwargs = {})
%detach_36 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_71,), kwargs = {})
%silu_16 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_69,), kwargs = {})
%_param_constant84 : [#users=1] = get_attr[target=_param_constant84]
%_param_constant85 : [#users=1] = get_attr[target=_param_constant85]
%convolution_20 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_16, %_param_constant84, %_param_constant85, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_54 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_20, [1, 32, 8, 65536]), kwargs = {})
%convert_element_type_72 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_54, torch.float32), kwargs = {})
%var_mean_18 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_72, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_36 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_18, 0), kwargs = {})
%getitem_37 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_18, 1), kwargs = {})
%add_48 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_36, 1e-06), kwargs = {})
%rsqrt_18 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_48,), kwargs = {})
%sub_18 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_54, %getitem_37), kwargs = {})
%mul_37 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_18, %rsqrt_18), kwargs = {})
%view_55 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_37, [1, 256, 256, 256]), kwargs = {})
%_param_constant86 : [#users=1] = get_attr[target=_param_constant86]
%unsqueeze_108 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant86, 0), kwargs = {})
%unsqueeze_109 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_108, 2), kwargs = {})
%unsqueeze_110 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_109, 3), kwargs = {})
%_param_constant87 : [#users=1] = get_attr[target=_param_constant87]
%unsqueeze_111 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant87, 0), kwargs = {})
%unsqueeze_112 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_111, 2), kwargs = {})
%unsqueeze_113 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_112, 3), kwargs = {})
%mul_38 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_55, %unsqueeze_113), kwargs = {})
%add_49 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_38, %unsqueeze_110), kwargs = {})
%convert_element_type_73 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_49, torch.float16), kwargs = {})
%convert_element_type_74 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_37, torch.float16), kwargs = {})
%convert_element_type_75 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_18, torch.float16), kwargs = {})
%squeeze_72 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_74, 3), kwargs = {})
%squeeze_73 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_72, 2), kwargs = {})
%squeeze_74 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_75, 3), kwargs = {})
%squeeze_75 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_74, 2), kwargs = {})
%detach_37 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_73,), kwargs = {})
%detach_38 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_75,), kwargs = {})
%silu_17 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_73,), kwargs = {})
%_param_constant88 : [#users=1] = get_attr[target=_param_constant88]
%_param_constant89 : [#users=1] = get_attr[target=_param_constant89]
%convolution_21 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_17, %_param_constant88, %_param_constant89, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%_param_constant90 : [#users=1] = get_attr[target=_param_constant90]
%_param_constant91 : [#users=1] = get_attr[target=_param_constant91]
%convolution_22 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%convolution_19, %_param_constant90, %_param_constant91, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
%add_50 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%convolution_22, %convolution_21), kwargs = {})
%div_9 : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_50, 1.0), kwargs = {})
%view_56 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%div_9, [1, 32, 8, 65536]), kwargs = {})
%convert_element_type_76 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_56, torch.float32), kwargs = {})
%var_mean_19 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_76, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_38 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_19, 0), kwargs = {})
%getitem_39 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_19, 1), kwargs = {})
%add_51 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_38, 1e-06), kwargs = {})
%rsqrt_19 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_51,), kwargs = {})
%sub_19 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_56, %getitem_39), kwargs = {})
%mul_39 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_19, %rsqrt_19), kwargs = {})
%view_57 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_39, [1, 256, 256, 256]), kwargs = {})
%_param_constant92 : [#users=1] = get_attr[target=_param_constant92]
%unsqueeze_114 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant92, 0), kwargs = {})
%unsqueeze_115 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_114, 2), kwargs = {})
%unsqueeze_116 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_115, 3), kwargs = {})
%_param_constant93 : [#users=1] = get_attr[target=_param_constant93]
%unsqueeze_117 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant93, 0), kwargs = {})
%unsqueeze_118 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_117, 2), kwargs = {})
%unsqueeze_119 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_118, 3), kwargs = {})
%mul_40 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_57, %unsqueeze_119), kwargs = {})
%add_52 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_40, %unsqueeze_116), kwargs = {})
%convert_element_type_77 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_52, torch.float16), kwargs = {})
%convert_element_type_78 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_39, torch.float16), kwargs = {})
%convert_element_type_79 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_19, torch.float16), kwargs = {})
%squeeze_76 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_78, 3), kwargs = {})
%squeeze_77 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_76, 2), kwargs = {})
%squeeze_78 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_79, 3), kwargs = {})
%squeeze_79 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_78, 2), kwargs = {})
%detach_39 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_77,), kwargs = {})
%detach_40 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_79,), kwargs = {})
%silu_18 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_77,), kwargs = {})
%_param_constant94 : [#users=1] = get_attr[target=_param_constant94]
%_param_constant95 : [#users=1] = get_attr[target=_param_constant95]
%convolution_23 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_18, %_param_constant94, %_param_constant95, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_58 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_23, [1, 32, 8, 65536]), kwargs = {})
%convert_element_type_80 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_58, torch.float32), kwargs = {})
%var_mean_20 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_80, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_40 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_20, 0), kwargs = {})
%getitem_41 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_20, 1), kwargs = {})
%add_53 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_40, 1e-06), kwargs = {})
%rsqrt_20 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_53,), kwargs = {})
%sub_20 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_58, %getitem_41), kwargs = {})
%mul_41 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_20, %rsqrt_20), kwargs = {})
%view_59 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_41, [1, 256, 256, 256]), kwargs = {})
%_param_constant96 : [#users=1] = get_attr[target=_param_constant96]
%unsqueeze_120 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant96, 0), kwargs = {})
%unsqueeze_121 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_120, 2), kwargs = {})
%unsqueeze_122 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_121, 3), kwargs = {})
%_param_constant97 : [#users=1] = get_attr[target=_param_constant97]
%unsqueeze_123 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant97, 0), kwargs = {})
%unsqueeze_124 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_123, 2), kwargs = {})
%unsqueeze_125 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_124, 3), kwargs = {})
%mul_42 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_59, %unsqueeze_125), kwargs = {})
%add_54 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_42, %unsqueeze_122), kwargs = {})
%convert_element_type_81 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_54, torch.float16), kwargs = {})
%convert_element_type_82 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_41, torch.float16), kwargs = {})
%convert_element_type_83 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_20, torch.float16), kwargs = {})
%squeeze_80 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_82, 3), kwargs = {})
%squeeze_81 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_80, 2), kwargs = {})
%squeeze_82 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_83, 3), kwargs = {})
%squeeze_83 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_82, 2), kwargs = {})
%detach_41 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_81,), kwargs = {})
%detach_42 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_83,), kwargs = {})
%silu_19 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_81,), kwargs = {})
%_param_constant98 : [#users=1] = get_attr[target=_param_constant98]
%_param_constant99 : [#users=1] = get_attr[target=_param_constant99]
%convolution_24 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_19, %_param_constant98, %_param_constant99, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_55 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_9, %convolution_24), kwargs = {})
%div_10 : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_55, 1.0), kwargs = {})
%view_60 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%div_10, [1, 32, 8, 65536]), kwargs = {})
%convert_element_type_84 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_60, torch.float32), kwargs = {})
%var_mean_21 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_84, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_42 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_21, 0), kwargs = {})
%getitem_43 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_21, 1), kwargs = {})
%add_56 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_42, 1e-06), kwargs = {})
%rsqrt_21 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_56,), kwargs = {})
%sub_21 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_60, %getitem_43), kwargs = {})
%mul_43 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_21, %rsqrt_21), kwargs = {})
%view_61 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_43, [1, 256, 256, 256]), kwargs = {})
%_param_constant100 : [#users=1] = get_attr[target=_param_constant100]
%unsqueeze_126 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant100, 0), kwargs = {})
%unsqueeze_127 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_126, 2), kwargs = {})
%unsqueeze_128 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_127, 3), kwargs = {})
%_param_constant101 : [#users=1] = get_attr[target=_param_constant101]
%unsqueeze_129 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant101, 0), kwargs = {})
%unsqueeze_130 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_129, 2), kwargs = {})
%unsqueeze_131 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_130, 3), kwargs = {})
%mul_44 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_61, %unsqueeze_131), kwargs = {})
%add_57 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_44, %unsqueeze_128), kwargs = {})
%convert_element_type_85 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_57, torch.float16), kwargs = {})
%convert_element_type_86 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_43, torch.float16), kwargs = {})
%convert_element_type_87 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_21, torch.float16), kwargs = {})
%squeeze_84 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_86, 3), kwargs = {})
%squeeze_85 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_84, 2), kwargs = {})
%squeeze_86 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_87, 3), kwargs = {})
%squeeze_87 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_86, 2), kwargs = {})
%detach_43 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_85,), kwargs = {})
%detach_44 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_87,), kwargs = {})
%silu_20 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_85,), kwargs = {})
%_param_constant102 : [#users=1] = get_attr[target=_param_constant102]
%_param_constant103 : [#users=1] = get_attr[target=_param_constant103]
%convolution_25 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_20, %_param_constant102, %_param_constant103, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_62 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_25, [1, 32, 8, 65536]), kwargs = {})
%convert_element_type_88 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_62, torch.float32), kwargs = {})
%var_mean_22 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_88, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_44 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_22, 0), kwargs = {})
%getitem_45 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_22, 1), kwargs = {})
%add_58 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_44, 1e-06), kwargs = {})
%rsqrt_22 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_58,), kwargs = {})
%sub_22 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_62, %getitem_45), kwargs = {})
%mul_45 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_22, %rsqrt_22), kwargs = {})
%view_63 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_45, [1, 256, 256, 256]), kwargs = {})
%_param_constant104 : [#users=1] = get_attr[target=_param_constant104]
%unsqueeze_132 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant104, 0), kwargs = {})
%unsqueeze_133 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_132, 2), kwargs = {})
%unsqueeze_134 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_133, 3), kwargs = {})
%_param_constant105 : [#users=1] = get_attr[target=_param_constant105]
%unsqueeze_135 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant105, 0), kwargs = {})
%unsqueeze_136 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_135, 2), kwargs = {})
%unsqueeze_137 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_136, 3), kwargs = {})
%mul_46 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_63, %unsqueeze_137), kwargs = {})
%add_59 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_46, %unsqueeze_134), kwargs = {})
%convert_element_type_89 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_59, torch.float16), kwargs = {})
%convert_element_type_90 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_45, torch.float16), kwargs = {})
%convert_element_type_91 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_22, torch.float16), kwargs = {})
%squeeze_88 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_90, 3), kwargs = {})
%squeeze_89 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_88, 2), kwargs = {})
%squeeze_90 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_91, 3), kwargs = {})
%squeeze_91 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_90, 2), kwargs = {})
%detach_45 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_89,), kwargs = {})
%detach_46 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_91,), kwargs = {})
%silu_21 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_89,), kwargs = {})
%_param_constant106 : [#users=1] = get_attr[target=_param_constant106]
%_param_constant107 : [#users=1] = get_attr[target=_param_constant107]
%convolution_26 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_21, %_param_constant106, %_param_constant107, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_60 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_10, %convolution_26), kwargs = {})
%div_11 : [#users=1] = call_function[target=torch.ops.aten.div](args = (%add_60, 1.0), kwargs = {})
%upsample_nearest2d_2 : [#users=1] = call_function[target=torch.ops.aten.upsample_nearest2d](args = (%div_11, [512, 512], 2.0, 2.0), kwargs = {})
%_param_constant108 : [#users=1] = get_attr[target=_param_constant108]
%_param_constant109 : [#users=1] = get_attr[target=_param_constant109]
%convolution_27 : [#users=2] = call_function[target=torch.ops.aten.convolution](args = (%upsample_nearest2d_2, %_param_constant108, %_param_constant109, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_64 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_27, [1, 32, 8, 262144]), kwargs = {})
%convert_element_type_92 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_64, torch.float32), kwargs = {})
%var_mean_23 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_92, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_46 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_23, 0), kwargs = {})
%getitem_47 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_23, 1), kwargs = {})
%add_61 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_46, 1e-06), kwargs = {})
%rsqrt_23 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_61,), kwargs = {})
%sub_23 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_64, %getitem_47), kwargs = {})
%mul_47 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_23, %rsqrt_23), kwargs = {})
%view_65 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_47, [1, 256, 512, 512]), kwargs = {})
%_param_constant110 : [#users=1] = get_attr[target=_param_constant110]
%unsqueeze_138 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant110, 0), kwargs = {})
%unsqueeze_139 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_138, 2), kwargs = {})
%unsqueeze_140 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_139, 3), kwargs = {})
%_param_constant111 : [#users=1] = get_attr[target=_param_constant111]
%unsqueeze_141 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant111, 0), kwargs = {})
%unsqueeze_142 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_141, 2), kwargs = {})
%unsqueeze_143 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_142, 3), kwargs = {})
%mul_48 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_65, %unsqueeze_143), kwargs = {})
%add_62 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_48, %unsqueeze_140), kwargs = {})
%convert_element_type_93 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_62, torch.float16), kwargs = {})
%convert_element_type_94 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_47, torch.float16), kwargs = {})
%convert_element_type_95 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_23, torch.float16), kwargs = {})
%squeeze_92 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_94, 3), kwargs = {})
%squeeze_93 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_92, 2), kwargs = {})
%squeeze_94 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_95, 3), kwargs = {})
%squeeze_95 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_94, 2), kwargs = {})
%detach_47 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_93,), kwargs = {})
%detach_48 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_95,), kwargs = {})
%silu_22 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_93,), kwargs = {})
%_param_constant112 : [#users=1] = get_attr[target=_param_constant112]
%_param_constant113 : [#users=1] = get_attr[target=_param_constant113]
%convolution_28 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_22, %_param_constant112, %_param_constant113, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_66 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_28, [1, 32, 4, 262144]), kwargs = {})
%convert_element_type_96 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_66, torch.float32), kwargs = {})
%var_mean_24 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_96, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_48 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_24, 0), kwargs = {})
%getitem_49 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_24, 1), kwargs = {})
%add_63 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_48, 1e-06), kwargs = {})
%rsqrt_24 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_63,), kwargs = {})
%sub_24 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_66, %getitem_49), kwargs = {})
%mul_49 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_24, %rsqrt_24), kwargs = {})
%view_67 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_49, [1, 128, 512, 512]), kwargs = {})
%_param_constant114 : [#users=1] = get_attr[target=_param_constant114]
%unsqueeze_144 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant114, 0), kwargs = {})
%unsqueeze_145 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_144, 2), kwargs = {})
%unsqueeze_146 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_145, 3), kwargs = {})
%_param_constant115 : [#users=1] = get_attr[target=_param_constant115]
%unsqueeze_147 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant115, 0), kwargs = {})
%unsqueeze_148 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_147, 2), kwargs = {})
%unsqueeze_149 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_148, 3), kwargs = {})
%mul_50 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_67, %unsqueeze_149), kwargs = {})
%add_64 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_50, %unsqueeze_146), kwargs = {})
%convert_element_type_97 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_64, torch.float16), kwargs = {})
%convert_element_type_98 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_49, torch.float16), kwargs = {})
%convert_element_type_99 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_24, torch.float16), kwargs = {})
%squeeze_96 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_98, 3), kwargs = {})
%squeeze_97 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_96, 2), kwargs = {})
%squeeze_98 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_99, 3), kwargs = {})
%squeeze_99 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_98, 2), kwargs = {})
%detach_49 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_97,), kwargs = {})
%detach_50 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_99,), kwargs = {})
%silu_23 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_97,), kwargs = {})
%_param_constant116 : [#users=1] = get_attr[target=_param_constant116]
%_param_constant117 : [#users=1] = get_attr[target=_param_constant117]
%convolution_29 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_23, %_param_constant116, %_param_constant117, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%_param_constant118 : [#users=1] = get_attr[target=_param_constant118]
%_param_constant119 : [#users=1] = get_attr[target=_param_constant119]
%convolution_30 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%convolution_27, %_param_constant118, %_param_constant119, [1, 1], [0, 0], [1, 1], False, [0, 0], 1), kwargs = {})
%add_65 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%convolution_30, %convolution_29), kwargs = {})
%div_12 : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_65, 1.0), kwargs = {})
%view_68 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%div_12, [1, 32, 4, 262144]), kwargs = {})
%convert_element_type_100 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_68, torch.float32), kwargs = {})
%var_mean_25 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_100, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_50 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_25, 0), kwargs = {})
%getitem_51 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_25, 1), kwargs = {})
%add_66 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_50, 1e-06), kwargs = {})
%rsqrt_25 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_66,), kwargs = {})
%sub_25 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_68, %getitem_51), kwargs = {})
%mul_51 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_25, %rsqrt_25), kwargs = {})
%view_69 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_51, [1, 128, 512, 512]), kwargs = {})
%_param_constant120 : [#users=1] = get_attr[target=_param_constant120]
%unsqueeze_150 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant120, 0), kwargs = {})
%unsqueeze_151 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_150, 2), kwargs = {})
%unsqueeze_152 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_151, 3), kwargs = {})
%_param_constant121 : [#users=1] = get_attr[target=_param_constant121]
%unsqueeze_153 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant121, 0), kwargs = {})
%unsqueeze_154 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_153, 2), kwargs = {})
%unsqueeze_155 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_154, 3), kwargs = {})
%mul_52 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_69, %unsqueeze_155), kwargs = {})
%add_67 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_52, %unsqueeze_152), kwargs = {})
%convert_element_type_101 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_67, torch.float16), kwargs = {})
%convert_element_type_102 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_51, torch.float16), kwargs = {})
%convert_element_type_103 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_25, torch.float16), kwargs = {})
%squeeze_100 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_102, 3), kwargs = {})
%squeeze_101 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_100, 2), kwargs = {})
%squeeze_102 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_103, 3), kwargs = {})
%squeeze_103 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_102, 2), kwargs = {})
%detach_51 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_101,), kwargs = {})
%detach_52 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_103,), kwargs = {})
%silu_24 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_101,), kwargs = {})
%_param_constant122 : [#users=1] = get_attr[target=_param_constant122]
%_param_constant123 : [#users=1] = get_attr[target=_param_constant123]
%convolution_31 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_24, %_param_constant122, %_param_constant123, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_70 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_31, [1, 32, 4, 262144]), kwargs = {})
%convert_element_type_104 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_70, torch.float32), kwargs = {})
%var_mean_26 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_104, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_52 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_26, 0), kwargs = {})
%getitem_53 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_26, 1), kwargs = {})
%add_68 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_52, 1e-06), kwargs = {})
%rsqrt_26 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_68,), kwargs = {})
%sub_26 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_70, %getitem_53), kwargs = {})
%mul_53 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_26, %rsqrt_26), kwargs = {})
%view_71 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_53, [1, 128, 512, 512]), kwargs = {})
%_param_constant124 : [#users=1] = get_attr[target=_param_constant124]
%unsqueeze_156 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant124, 0), kwargs = {})
%unsqueeze_157 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_156, 2), kwargs = {})
%unsqueeze_158 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_157, 3), kwargs = {})
%_param_constant125 : [#users=1] = get_attr[target=_param_constant125]
%unsqueeze_159 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant125, 0), kwargs = {})
%unsqueeze_160 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_159, 2), kwargs = {})
%unsqueeze_161 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_160, 3), kwargs = {})
%mul_54 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_71, %unsqueeze_161), kwargs = {})
%add_69 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_54, %unsqueeze_158), kwargs = {})
%convert_element_type_105 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_69, torch.float16), kwargs = {})
%convert_element_type_106 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_53, torch.float16), kwargs = {})
%convert_element_type_107 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_26, torch.float16), kwargs = {})
%squeeze_104 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_106, 3), kwargs = {})
%squeeze_105 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_104, 2), kwargs = {})
%squeeze_106 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_107, 3), kwargs = {})
%squeeze_107 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_106, 2), kwargs = {})
%detach_53 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_105,), kwargs = {})
%detach_54 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_107,), kwargs = {})
%silu_25 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_105,), kwargs = {})
%_param_constant126 : [#users=1] = get_attr[target=_param_constant126]
%_param_constant127 : [#users=1] = get_attr[target=_param_constant127]
%convolution_32 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_25, %_param_constant126, %_param_constant127, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_70 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_12, %convolution_32), kwargs = {})
%div_13 : [#users=2] = call_function[target=torch.ops.aten.div](args = (%add_70, 1.0), kwargs = {})
%view_72 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%div_13, [1, 32, 4, 262144]), kwargs = {})
%convert_element_type_108 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_72, torch.float32), kwargs = {})
%var_mean_27 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_108, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_54 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_27, 0), kwargs = {})
%getitem_55 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_27, 1), kwargs = {})
%add_71 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_54, 1e-06), kwargs = {})
%rsqrt_27 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_71,), kwargs = {})
%sub_27 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_72, %getitem_55), kwargs = {})
%mul_55 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_27, %rsqrt_27), kwargs = {})
%view_73 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_55, [1, 128, 512, 512]), kwargs = {})
%_param_constant128 : [#users=1] = get_attr[target=_param_constant128]
%unsqueeze_162 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant128, 0), kwargs = {})
%unsqueeze_163 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_162, 2), kwargs = {})
%unsqueeze_164 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_163, 3), kwargs = {})
%_param_constant129 : [#users=1] = get_attr[target=_param_constant129]
%unsqueeze_165 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant129, 0), kwargs = {})
%unsqueeze_166 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_165, 2), kwargs = {})
%unsqueeze_167 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_166, 3), kwargs = {})
%mul_56 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_73, %unsqueeze_167), kwargs = {})
%add_72 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_56, %unsqueeze_164), kwargs = {})
%convert_element_type_109 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_72, torch.float16), kwargs = {})
%convert_element_type_110 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_55, torch.float16), kwargs = {})
%convert_element_type_111 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_27, torch.float16), kwargs = {})
%squeeze_108 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_110, 3), kwargs = {})
%squeeze_109 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_108, 2), kwargs = {})
%squeeze_110 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_111, 3), kwargs = {})
%squeeze_111 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_110, 2), kwargs = {})
%detach_55 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_109,), kwargs = {})
%detach_56 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_111,), kwargs = {})
%silu_26 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_109,), kwargs = {})
%_param_constant130 : [#users=1] = get_attr[target=_param_constant130]
%_param_constant131 : [#users=1] = get_attr[target=_param_constant131]
%convolution_33 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_26, %_param_constant130, %_param_constant131, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%view_74 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%convolution_33, [1, 32, 4, 262144]), kwargs = {})
%convert_element_type_112 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_74, torch.float32), kwargs = {})
%var_mean_28 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_112, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_56 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_28, 0), kwargs = {})
%getitem_57 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_28, 1), kwargs = {})
%add_73 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_56, 1e-06), kwargs = {})
%rsqrt_28 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_73,), kwargs = {})
%sub_28 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_74, %getitem_57), kwargs = {})
%mul_57 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_28, %rsqrt_28), kwargs = {})
%view_75 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_57, [1, 128, 512, 512]), kwargs = {})
%_param_constant132 : [#users=1] = get_attr[target=_param_constant132]
%unsqueeze_168 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant132, 0), kwargs = {})
%unsqueeze_169 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_168, 2), kwargs = {})
%unsqueeze_170 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_169, 3), kwargs = {})
%_param_constant133 : [#users=1] = get_attr[target=_param_constant133]
%unsqueeze_171 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant133, 0), kwargs = {})
%unsqueeze_172 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_171, 2), kwargs = {})
%unsqueeze_173 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_172, 3), kwargs = {})
%mul_58 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_75, %unsqueeze_173), kwargs = {})
%add_74 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_58, %unsqueeze_170), kwargs = {})
%convert_element_type_113 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_74, torch.float16), kwargs = {})
%convert_element_type_114 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_57, torch.float16), kwargs = {})
%convert_element_type_115 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_28, torch.float16), kwargs = {})
%squeeze_112 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_114, 3), kwargs = {})
%squeeze_113 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_112, 2), kwargs = {})
%squeeze_114 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_115, 3), kwargs = {})
%squeeze_115 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_114, 2), kwargs = {})
%detach_57 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_113,), kwargs = {})
%detach_58 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_115,), kwargs = {})
%silu_27 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_113,), kwargs = {})
%_param_constant134 : [#users=1] = get_attr[target=_param_constant134]
%_param_constant135 : [#users=1] = get_attr[target=_param_constant135]
%convolution_34 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_27, %_param_constant134, %_param_constant135, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%add_75 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_13, %convolution_34), kwargs = {})
%div_14 : [#users=1] = call_function[target=torch.ops.aten.div](args = (%add_75, 1.0), kwargs = {})
%view_76 : [#users=2] = call_function[target=torch.ops.aten.view](args = (%div_14, [1, 32, 4, 262144]), kwargs = {})
%convert_element_type_116 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%view_76, torch.float32), kwargs = {})
%var_mean_29 : [#users=2] = call_function[target=torch.ops.aten.var_mean](args = (%convert_element_type_116, [2, 3]), kwargs = {correction: 0, keepdim: True})
%getitem_58 : [#users=1] = call_function[target=operator.getitem](args = (%var_mean_29, 0), kwargs = {})
%getitem_59 : [#users=2] = call_function[target=operator.getitem](args = (%var_mean_29, 1), kwargs = {})
%add_76 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%getitem_58, 1e-06), kwargs = {})
%rsqrt_29 : [#users=2] = call_function[target=torch.ops.aten.rsqrt](args = (%add_76,), kwargs = {})
%sub_29 : [#users=1] = call_function[target=torch.ops.aten.sub](args = (%view_76, %getitem_59), kwargs = {})
%mul_59 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%sub_29, %rsqrt_29), kwargs = {})
%view_77 : [#users=1] = call_function[target=torch.ops.aten.view](args = (%mul_59, [1, 128, 512, 512]), kwargs = {})
%_param_constant136 : [#users=1] = get_attr[target=_param_constant136]
%unsqueeze_174 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant136, 0), kwargs = {})
%unsqueeze_175 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_174, 2), kwargs = {})
%unsqueeze_176 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_175, 3), kwargs = {})
%_param_constant137 : [#users=1] = get_attr[target=_param_constant137]
%unsqueeze_177 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%_param_constant137, 0), kwargs = {})
%unsqueeze_178 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_177, 2), kwargs = {})
%unsqueeze_179 : [#users=1] = call_function[target=torch.ops.aten.unsqueeze](args = (%unsqueeze_178, 3), kwargs = {})
%mul_60 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%view_77, %unsqueeze_179), kwargs = {})
%add_77 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%mul_60, %unsqueeze_176), kwargs = {})
%convert_element_type_117 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%add_77, torch.float16), kwargs = {})
%convert_element_type_118 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%getitem_59, torch.float16), kwargs = {})
%convert_element_type_119 : [#users=1] = call_function[target=torch.ops.prims.convert_element_type](args = (%rsqrt_29, torch.float16), kwargs = {})
%squeeze_116 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_118, 3), kwargs = {})
%squeeze_117 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_116, 2), kwargs = {})
%squeeze_118 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%convert_element_type_119, 3), kwargs = {})
%squeeze_119 : [#users=1] = call_function[target=torch.ops.aten.squeeze](args = (%squeeze_118, 2), kwargs = {})
%detach_59 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_117,), kwargs = {})
%detach_60 : [#users=0] = call_function[target=torch.ops.aten.detach](args = (%squeeze_119,), kwargs = {})
%silu_28 : [#users=1] = call_function[target=torch.ops.aten.silu](args = (%convert_element_type_117,), kwargs = {})
%_param_constant138 : [#users=1] = get_attr[target=_param_constant138]
%_param_constant139 : [#users=1] = get_attr[target=_param_constant139]
%convolution_35 : [#users=1] = call_function[target=torch.ops.aten.convolution](args = (%silu_28, %_param_constant138, %_param_constant139, [1, 1], [1, 1], [1, 1], False, [0, 0], 1), kwargs = {})
%div_15 : [#users=1] = call_function[target=torch.ops.aten.div](args = (%convolution_35, 2), kwargs = {})
%add_78 : [#users=1] = call_function[target=torch.ops.aten.add](args = (%div_15, 0.5), kwargs = {})
%clamp : [#users=1] = call_function[target=torch.ops.aten.clamp](args = (%add_78, 0, 1), kwargs = {})
%mul_61 : [#users=1] = call_function[target=torch.ops.aten.mul](args = (%clamp, 255.0), kwargs = {})
%round_1 : [#users=1] = call_function[target=torch.ops.aten.round](args = (%mul_61,), kwargs = {})
return round_1
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment