Added Reshape / View support.

Author: gmalivenko

README.md

@@ -74,7 +74,8 @@ Layers:
 
 Reshape:
 
-* View (only with 0.2)
+* View
+* Reshape (only with 0.4)
 * Transpose (only with 0.4)
 
 Activations:

pytorch2keras/layers.py

@@ -530,7 +530,7 @@ def convert_tanh(params, w_name, scope_name, inputs, layers, weights):
 
 def convert_transpose(params, w_name, scope_name, inputs, layers, weights):
     """
-    Convert tanh layer.
+    Convert transpose layer.
 
    Args:
         params: dictionary with layer parameters
@@ -551,6 +551,25 @@ def convert_transpose(params, w_name, scope_name, inputs, layers, weights):
         layers[scope_name] = permute(layers[inputs[0]])
 
 
+def convert_reshape(params, w_name, scope_name, inputs, layers, weights):
+    """
+    Convert reshape layer.
+
+   Args:
+        params: dictionary with layer parameters
+        w_name: name prefix in state_dict
+        scope_name: pytorch scope name
+        inputs: pytorch node inputs
+        layers: dictionary with keras tensors
+        weights: pytorch state_dict
+    """
+    print('Converting reshape ...')
+
+    tf_name = w_name + str(random.random())
+    reshape = keras.layers.Reshape(params['shape'], name=tf_name)
+    layers[scope_name] = reshape(layers[inputs[0]])
+
+
 def convert_matmul(params, w_name, scope_name, inputs, layers, weights):
     """
     Convert tanh layer.
@@ -604,5 +623,6 @@ def convert_matmul(params, w_name, scope_name, inputs, layers, weights):
     'Softmax': convert_softmax,
     'Tanh': convert_tanh,
     'Transpose': convert_transpose,
+    'Reshape': convert_reshape,
     'MatMul': convert_matmul,
 }

tests/view.py

@@ -0,0 +1,51 @@
+import keras  # work around segfault
+import sys
+import numpy as np
+
+import torch
+import torch.nn as nn
+from torch.autograd import Variable
+
+sys.path.append('../pytorch2keras')
+from converter import pytorch_to_keras
+
+
+class TestView(nn.Module):
+    """Module for View conversion testing
+    """
+
+    def __init__(self, inp=10, out=16, kernel_size=3, bias=True):
+        super(TestView, self).__init__()
+        self.conv2d = nn.Conv2d(inp, out, kernel_size=kernel_size, bias=bias)
+
+    def forward(self, x):
+        x = self.conv2d(x)
+        x = x.view([x.size(0), -1, 2, 1, 1, 1, 1, 1])
+        x = torch.nn.Tanh()(x)
+        return x
+
+
+if __name__ == '__main__':
+    max_error = 0
+    for i in range(100):
+        kernel_size = np.random.randint(1, 7)
+        inp = 2 * np.random.randint(kernel_size + 1, 10)
+        out = 2 * np.random.randint(1, 10)
+
+        model = TestView(inp, out, kernel_size, inp % 2)
+
+        input_np = np.random.uniform(0, 1, (1, inp, inp, inp))
+        input_var = Variable(torch.FloatTensor(input_np))
+        output = model(input_var)
+
+        k_model = pytorch_to_keras(model, input_var, (inp, inp, inp,), verbose=True)
+
+        pytorch_output = output.data.numpy()
+        keras_output = k_model.predict(input_np)
+
+        error = np.max(pytorch_output - keras_output)
+        print(error)
+        if max_error < error:
+            max_error = error
+
+    print('Max error: {0}'.format(max_error))