DeformConv (test), GroupNormalization (Dynamo)

.gitignore

@@ -22,4 +22,6 @@ saved_model/
 node_modules/
 
 .act.secrets
-.act.env
+.act.env
+
+*.sarif

make_test_op/make_DeformConv2d.py

@@ -0,0 +1,37 @@
+#! /usr/bin/env python
+
+import torch
+import torch.nn as nn
+import numpy as np
+import onnx
+from onnxsim import simplify
+import numpy as np
+np.random.seed(0)
+from torchvision.ops import deform_conv2d
+
+class Model(nn.Module):
+    def __init__(self):
+        super(Model, self).__init__()
+
+    def forward(self, x):
+        kh, kw = 3, 3
+        weight = torch.rand(5, 3, kh, kw)
+        offset = torch.rand(4, 2 * kh * kw, 8, 8)
+        mask = torch.rand(4, kh * kw, 8, 8)
+        return deform_conv2d(input, offset, weight, mask=mask)
+
+if __name__ == "__main__":
+    OPSET=19
+    MODEL = f'DeformConv'
+    model = Model()
+    onnx_file = f"{MODEL}_{OPSET}.onnx"
+    x = torch.randn(4, 3, 10, 10)
+    onnx_program = torch.onnx.dynamo_export(model, x)
+    onnx_program.save(onnx_file)
+
+    # model_onnx1 = onnx.load(onnx_file)
+    # model_onnx1 = onnx.shape_inference.infer_shapes(model_onnx1)
+    # onnx.save(model_onnx1, onnx_file)
+    # model_onnx2 = onnx.load(onnx_file)
+    # model_simp, check = simplify(model_onnx2)
+    # onnx.save(model_simp, onnx_file)

make_test_op/make_GroupNormalization.py

@@ -91,4 +91,8 @@ def forward(self, x):
     onnx.save(model_onnx1, onnx_file)
     model_onnx2 = onnx.load(onnx_file)
     model_simp, check = simplify(model_onnx2)
-    onnx.save(model_simp, onnx_file)
+    onnx.save(model_simp, onnx_file)
+
+    onnx_file = f"{MODEL}_{OPSET}_dynamo.onnx"
+    onnx_program = torch.onnx.dynamo_export(model, x)
+    onnx_program.save(onnx_file)