feat(components): PyTorch - Convert to ONNX from PyTorch ScriptModule (…

…#5207)

components/PyTorch/Convert_to_OnnxModel_from_PyTorchScriptModule/component.py

@@ -0,0 +1,36 @@
+from kfp.components import create_component_from_func, InputPath, OutputPath
+
+
+def convert_to_onnx_from_pytorch_script_module(
+    model_path: InputPath('PyTorchScriptModule'),
+    converted_model_path: OutputPath('OnnxModel'),
+    list_of_input_shapes: list,
+):
+    '''Creates fully-connected network in PyTorch ScriptModule format'''
+    import torch
+    model = torch.jit.load(model_path)
+    example_inputs = [
+        torch.ones(*input_shape)
+        for input_shape in list_of_input_shapes
+    ]
+    example_outputs = model.forward(*example_inputs)
+    torch.onnx.export(
+        model=model,
+        args=example_inputs,
+        f=converted_model_path,
+        verbose=True,
+        training=torch.onnx.TrainingMode.EVAL,
+        example_outputs=example_outputs,
+    )
+
+
+if __name__ == '__main__':
+    convert_to_onnx_from_pytorch_script_module_op = create_component_from_func(
+        convert_to_onnx_from_pytorch_script_module,
+        output_component_file='component.yaml',
+        base_image='pytorch/pytorch:1.7.1-cuda11.0-cudnn8-runtime',
+        packages_to_install=[],
+        annotations={
+            "author": "Alexey Volkov <alexey.volkov@ark-kun.com>",
+        },
+    )

components/PyTorch/Convert_to_OnnxModel_from_PyTorchScriptModule/component.yaml

@@ -0,0 +1,63 @@
+name: Convert to onnx from pytorch script module
+description: Creates fully-connected network in PyTorch ScriptModule format
+metadata:
+  annotations: {author: Alexey Volkov <alexey.volkov@ark-kun.com>}
+inputs:
+- {name: model, type: PyTorchScriptModule}
+- {name: list_of_input_shapes, type: JsonArray}
+outputs:
+- {name: converted_model, type: OnnxModel}
+implementation:
+  container:
+    image: pytorch/pytorch:1.7.1-cuda11.0-cudnn8-runtime
+    command:
+    - sh
+    - -ec
+    - |
+      program_path=$(mktemp)
+      printf "%s" "$0" > "$program_path"
+      python3 -u "$program_path" "$@"
+    - |
+      def _make_parent_dirs_and_return_path(file_path: str):
+          import os
+          os.makedirs(os.path.dirname(file_path), exist_ok=True)
+          return file_path
+
+      def convert_to_onnx_from_pytorch_script_module(
+          model_path,
+          converted_model_path,
+          list_of_input_shapes,
+      ):
+          '''Creates fully-connected network in PyTorch ScriptModule format'''
+          import torch
+          model = torch.jit.load(model_path)
+          example_inputs = [
+              torch.ones(*input_shape)
+              for input_shape in list_of_input_shapes
+          ]
+          example_outputs = model.forward(*example_inputs)
+          torch.onnx.export(
+              model=model,
+              args=example_inputs,
+              f=converted_model_path,
+              verbose=True,
+              training=torch.onnx.TrainingMode.EVAL,
+              example_outputs=example_outputs,
+          )
+
+      import json
+      import argparse
+      _parser = argparse.ArgumentParser(prog='Convert to onnx from pytorch script module', description='Creates fully-connected network in PyTorch ScriptModule format')
+      _parser.add_argument("--model", dest="model_path", type=str, required=True, default=argparse.SUPPRESS)
+      _parser.add_argument("--list-of-input-shapes", dest="list_of_input_shapes", type=json.loads, required=True, default=argparse.SUPPRESS)
+      _parser.add_argument("--converted-model", dest="converted_model_path", type=_make_parent_dirs_and_return_path, required=True, default=argparse.SUPPRESS)
+      _parsed_args = vars(_parser.parse_args())
+
+      _outputs = convert_to_onnx_from_pytorch_script_module(**_parsed_args)
+    args:
+    - --model
+    - {inputPath: model}
+    - --list-of-input-shapes
+    - {inputValue: list_of_input_shapes}
+    - --converted-model
+    - {outputPath: converted_model}