[Relay] Keras frontend upsample and 1 channel conv2d fixes (#3937)

* Fix upsample layout in keras frontend.

* Fixed group conv being used instead of conv when channels=1

* Add new conv2d test to catch bugs when channels=1.

python/tvm/relay/frontend/keras.py

@@ -395,7 +395,7 @@ def _convert_pooling(inexpr, keras_layer, etab):
 def _convert_upsample(inexpr, keras_layer, _):
     _check_data_format(keras_layer)
     upsample_type = type(keras_layer).__name__
-    params = {'layout': 'NHWC'}
+    params = {}
     if upsample_type == 'UpSampling1D':
         h = keras_layer.size
         params['scale'] = h

src/relay/op/nn/convolution.h

@@ -71,7 +71,7 @@ bool Conv2DRel(const Array<Type>& types, int num_inputs, const Attrs& attrs,
     CHECK_EQ(param->dilation.size(), 2);
     Array<IndexExpr> wshape;
 
-    if (tvm::ir::Equal(param->channels, param->groups)) {
+    if (tvm::ir::Equal(param->channels, param->groups) && !tvm::ir::Equal(param->channels, 1)) {
       // infer weight's shape for depthwise convolution
       wshape = {{dshape_nchw[1], param->groups / dshape_nchw[1], param->kernel_size[0],
                  param->kernel_size[1]}};

tests/python/frontend/keras/test_forward.py

@@ -181,6 +181,7 @@ def test_forward_conv():
                                       strides=(2, 2), padding='same'),
                   keras.layers.Conv2D(filters=10, kernel_size=(3, 3),
                                       dilation_rate=(2, 2), padding='same'),
+                  keras.layers.Conv2D(filters=1, kernel_size=(3, 3), padding='same'),
                   keras.layers.DepthwiseConv2D(kernel_size=(3, 3), padding='same'),
                   keras.layers.Conv2DTranspose(filters=10, kernel_size=(3, 3), padding='valid'),
                   keras.layers.SeparableConv2D(filters=10, kernel_size=(3, 3), padding='same')]
@@ -194,7 +195,7 @@ def test_forward_upsample(interpolation='nearest'):
     data = keras.layers.Input(shape=(32, 32, 3))
     x = keras.layers.UpSampling2D(size=(3, 3), interpolation=interpolation)(data)
     keras_model = keras.models.Model(data, x)
-    verify_keras_frontend(keras_model, need_transpose=False)
+    verify_keras_frontend(keras_model)
 
 
 def test_forward_reshape():