retinanet_resnet50_fpn issue

[swimdi]

Currently retinanet_resnet50_fpn has these issue

1. because it's just 🚧, so these new appeared not yet have lowering

aten.empty.memory_format
aten.fill.Scalar
aten.split_with_sizes.default

2. aten.stack.default: this input varation shape too large

# Statically allocated circular buffers on core range [(x=0,y=0) - (x=0,y=0)] grow to 3580704 B which is beyond max L1 size of 1499136 B
aten_aten_stack_default = [
    ["List[Tensor] tensors = [<[13600]>, <[13600]>, <[13600]>, <[13600]>]", "int dim = 1"],
]

Reproduce method is

• Remove this line
  

  pytorch2.0_ttnn/torch_ttnn/passes/lowering/to_tt_guard.py

  Line 83 in 3be2fd5

  ["List[Tensor] tensors = [<[13600]>, <[13600]>, <[13600]>, <[13600]>]", "int dim = 1"],

• Run pytest tests/models/torchvision/test_torchvision_object_detection.py -k retinanet_resnet50_fpn

3. aten.clone.default: torch.int32 cause this error

E       RuntimeError: TT_FATAL @ /tmp/build-via-sdist-5pcqh003/ttnn-0.54.0rc11+wormhole.b0/ttnn/cpp/ttnn/operations/data_movement/concat/device/concat_device_operation.cpp:42: in_ref.get_dtype() == first_input.get_dtype()
E       info:
E       All Tensors should have same dtypes.

Reproduce method is

• Add torch.int32: TtnnInt32(), in here

  pytorch2.0_ttnn/torch_ttnn/passes/lowering/to_tt_pass.py

  Lines 360 to 363 in 3be2fd5

  	dtype_map = {
  	torch.float32: TtnnBfloat16(),
  	torch.bfloat16: TtnnBfloat16(),
  	}

• Run pytest tests/models/torchvision/test_torchvision_object_detection.py -k retinanet_resnet50_fpn