Merge pull request #5 from mangguo321/mang/merge_to_yi_branch

Mang/merge to yi branch

ngraph/frontend/paddlepaddle/src/op/scale.cpp

@@ -25,7 +25,15 @@ namespace op {
 NamedOutputs scale (const NodeContext& node) {
     auto data = node.get_ng_input("X");
     auto scale = ngraph::opset6::Constant::create(ngraph::element::f32, {1}, {node.get_attribute<float>("scale")});
-    return node.default_single_output_mapping({std::make_shared<ngraph::opset6::Multiply>(data, scale)}, {"Out"});
+    auto bias = ngraph::opset6::Constant::create(ngraph::element::f32, {1}, {node.get_attribute<float>("bias")});
+    auto bias_after_scale = node.get_attribute<bool>("bias_after_scale");
+    if(!bias_after_scale) {
+        auto node_add = std::make_shared<ngraph::opset6::Add>(data, bias);
+        return node.default_single_output_mapping({std::make_shared<ngraph::opset6::Multiply>(node_add, scale)}, {"Out"});
+    } else {
+        auto node_multiply =  std::make_shared<ngraph::opset6::Multiply>(data, scale);
+        return node.default_single_output_mapping({std::make_shared<ngraph::opset6::Add>(node_multiply, bias)}, {"Out"});
+    }
 }
 
 }}}}

ngraph/test/files/paddlepaddle/gen_scripts/generate_reshape.py

@@ -0,0 +1,43 @@
+#
+# reshape paddle model generator
+#
+import numpy as np
+from save_model import saveModel
+
+data_type = 'float32'
+
+def reshape(name : str, x, out_shape):
+    import paddle as pdpd
+    pdpd.enable_static()
+
+    with pdpd.static.program_guard(pdpd.static.Program(), pdpd.static.Program()):
+        node_x = pdpd.static.data(name='x', shape=x.shape, dtype=data_type)
+        out = pdpd.fluid.layers.reshape(x=node_x, shape=out_shape)
+
+        cpu = pdpd.static.cpu_places(1)
+        exe = pdpd.static.Executor(cpu[0])
+        # startup program will call initializer to initialize the parameters.
+        exe.run(pdpd.static.default_startup_program())
+
+        outs = exe.run(
+            feed={'x': x},
+            fetch_list=[out])             
+
+        saveModel(name, exe, feedkeys=['x'], fetchlist=[out], inputs=[x], outputs=[outs[0]])
+
+    return outs[0]
+
+def main():
+    data = np.array([[[
+        [1, 2, 3, 4],
+        [5, 6, 7, 8],
+        [9, 10, 11, 12],
+        [13, 14, 15, 16],
+    ]]], dtype=np.float32)
+    out_shape = [1, 1, 2, 8]
+    reshape("reshape", data, out_shape)
+
+
+
+if __name__ == "__main__":
+    main()     

ngraph/test/files/paddlepaddle/gen_scripts/generate_scale.py

@@ -0,0 +1,48 @@
+#
+# pool2d paddle model generator
+#
+import numpy as np
+from save_model import saveModel
+
+data_type = 'float32'
+
+def pdpd_scale(name : str, x, scale, bias, attrs : dict):
+    import paddle as pdpd
+    pdpd.enable_static()
+
+    with pdpd.static.program_guard(pdpd.static.Program(), pdpd.static.Program()):
+        node_x = pdpd.static.data(name='x', shape=x.shape, dtype='float32')
+        out = pdpd.scale(x=node_x, scale=scale, bias=bias, 
+                         bias_after_scale=attrs['bias_after_scale'])
+
+        cpu = pdpd.static.cpu_places(1)
+        exe = pdpd.static.Executor(cpu[0])
+        # startup program will call initializer to initialize the parameters.
+        exe.run(pdpd.static.default_startup_program())
+
+        outs = exe.run(
+            feed={'x': x},
+            fetch_list=[out])             
+
+        saveModel(name, exe, feedkeys=['x'], fetchlist=[out], inputs=[x], outputs=[outs[0]])
+
+    return outs[0]
+
+
+def main():
+    scale = 2.0
+    bias = 1.0
+    data = np.random.random([2, 3]).astype("float32")
+    pdpd_attrs = {
+        'bias_after_scale' : True,
+    }  
+    pdpd_scale("scale_test1", data, scale, bias, pdpd_attrs)
+
+    pdpd_attrs = {
+        'bias_after_scale' : False,
+    }
+    pdpd_scale("scale_test2", data, scale, bias, pdpd_attrs)
+
+
+if __name__ == "__main__":
+    main()     

ngraph/test/files/paddlepaddle/models/models.csv

@@ -13,3 +13,6 @@ maxPool_test5,
 #maxPool_test6,
 avgAdaptivePool2D_test1,
 #maxAdaptivePool2D_test1,
+scale_test1,
+scale_test2,
+reshape,