Add torchvision_vit model to the examples (#33)

Summary:
Pull Request resolved: #33

This diff is adding a selected model for Apple.
Info about the the model: https://pytorch.org/vision/main/models/generated/torchvision.models.vit_b_16.html#torchvision.models.vit_b_16

Differential Revision: D48012005

fbshipit-source-id: 285c7b11704d24b613b6cba1ea64845b60e1b9ee

Author: guangy10

examples/models/TARGETS

@@ -11,5 +11,6 @@ python_library(
         "//executorch/backends:compile_spec_schema",
         "//executorch/examples/models/mobilenet_v2:mv2_export",
         "//executorch/examples/models/mobilenet_v3:mv3_export",
+        "//executorch/examples/models/torchvision_vit:vit_export",
     ],
 )

examples/models/models.py

@@ -12,6 +12,8 @@
 import torch
 from executorch.backends.compile_spec_schema import CompileSpec
 
+from .torchvision_vit import TorchVisionViTModel
+
 
 class MulModule(torch.nn.Module):
     def __init__(self) -> None:
@@ -89,11 +91,18 @@ def gen_mobilenet_v2_model_inputs() -> Tuple[torch.nn.Module, Any]:
     return MV2Model.get_model(), MV2Model.get_example_inputs()
 
 
+def gen_torchvision_vit_model_and_inputs() -> Tuple[torch.nn.Module, Any]:
+    from ..models.torchvision_vit import TorchVisionViTModel
+
+    return TorchVisionViTModel.get_model(), TorchVisionViTModel.get_example_inputs()
+
+
 MODEL_NAME_TO_MODEL = {
     "mul": lambda: (MulModule(), MulModule.get_example_inputs()),
     "linear": lambda: (LinearModule(), LinearModule.get_example_inputs()),
     "add": lambda: (AddModule(), AddModule.get_example_inputs()),
     "add_mul": lambda: (AddMulModule(), AddMulModule.get_example_inputs()),
     "mv2": gen_mobilenet_v2_model_inputs,
     "mv3": gen_mobilenet_v3_model_inputs,
+    "vit": gen_torchvision_vit_model_and_inputs,
 }

examples/models/torchvision_vit/TARGETS

@@ -0,0 +1,14 @@
+load("@fbcode_macros//build_defs:python_library.bzl", "python_library")
+
+python_library(
+    name = "vit_export",
+    srcs = [
+        "__init__.py",
+        "export.py",
+    ],
+    base_module = "executorch.examples.models.torchvision_vit",
+    deps = [
+        "//caffe2:torch",
+        "//pytorch/vision:torchvision",
+    ],
+)

examples/models/torchvision_vit/__init__.py

@@ -0,0 +1,11 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree.
+
+from .export import TorchVisionViTModel
+
+__all__ = [
+    TorchVisionViTModel,
+]

examples/models/torchvision_vit/export.py

@@ -0,0 +1,30 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree.
+
+import logging
+
+import torch
+from torchvision import models
+
+FORMAT = "[%(filename)s:%(lineno)s] %(message)s"
+logging.basicConfig(format=FORMAT)
+
+
+class TorchVisionViTModel:
+    def __init__(self):
+        pass
+
+    @staticmethod
+    def get_model():
+        logging.info("loading torchvision vit_b_16 model")
+        vit_b_16 = models.vit_b_16(weights="IMAGENET1K_V1")
+        logging.info("loaded torchvision vit_b_16 model")
+        return vit_b_16
+
+    @staticmethod
+    def get_example_inputs():
+        input_shape = (1, 3, 224, 224)
+        return (torch.randn(input_shape),)