[Caffe Frontend] supporting group > 1 cases for Deconv op (apache#8260)

* [Caffe Frontend] supporting group > 1 cases for Deconv op

- Handling group > 1 cases, assuming group == output channels
- Simply decomposed into Relay split, conv2d_transposed, and multi-leveled concatenate ops
- Added some test cases

Signed-off-by: zotanika <[email protected]>

* [Caffe Frontend] amending a test case for Deconv op

Signed-off-by: zotanika <[email protected]>

* explicit importing tvm.testing

* changing split axis to 0, according to PR apache#9336

Author: zotanika

python/tvm/relay/frontend/caffe.py

@@ -515,21 +515,76 @@ def convert_deconv(self, op):
         if weight:
             kh, kw = params["kernel_size"]
             weight_shape = [-1, conv_params.num_output, kh, kw]
-            weight_value = np.asarray(weight.data, np.float32)
+            if not weight.data:
+                if conv_params.weight_filler:
+                    _filler = conv_params.weight_filler.value
+                    weight_value = np.full(weight.shape.dim, _filler, np.float32)
+                else:
+                    raise tvm.error.OpAttributeInvalid("At least weight_filler must be given")
+            else:
+                weight_value = np.asarray(weight.data, np.float32)
             weight_value = np.reshape(weight_value, weight_shape)
 
             # weight shape is in relay's IOHW format rn, we need it to be OIHW
             weight_value = np.transpose(weight_value, [1, 0, 2, 3])
         else:
-            raise Exception("No weight value of layer {} in caffemodel".format(op.name))
+            raise tvm.error.OpAttributeRequired(
+                "No weight value of layer {} in caffemodel".format(op.name)
+            )
 
         weight_expr = self.exp_tab.new_const(weight_value, dtype="float32")
         in_expr = self.exp_tab.get_expr(inputs[0])
-        out = _op.nn.conv2d_transpose(data=in_expr, weight=weight_expr, **params)
+
+        groups = params["groups"]
+        channels = params["channels"]
+
         if bias:
             bias_value = np.asarray(bias.data, np.float32)
             bias_expr = self.exp_tab.new_const(bias_value, dtype="float32")
-            out = _op.nn.bias_add(out, bias_expr)
+
+        if groups > channels:
+            raise tvm.error.OpAttributeInvalid(
+                "Groups cannot be larger than the number of input channels"
+            )
+
+        if groups == channels:
+            inputs_expr = _op.split(in_expr, groups, axis=1)
+            # changing split axis to 0, according to PR #9336
+            weights_expr = _op.split(weight_expr, groups, axis=0)
+            # Preventing to create Concat layer with too many tensors(> 16)
+            q = groups >> 4
+            r = groups % 16
+
+            params["groups"] = 1
+            params["channels"] = 1
+            out = []
+            for lc in range(q):
+                _outputs = []
+                _inputs = [inputs_expr[i] for i in range(lc << 4, (lc << 4) + 16)]
+                _weights = [weights_expr[i] for i in range(lc << 4, (lc << 4) + 16)]
+                for (i, w) in zip(_inputs, _weights):
+                    _out = _op.nn.conv2d_transpose(data=i, weight=w, **params)
+                    if bias:
+                        _out = _op.nn.bias_add(_out, bias_expr)
+                    _outputs.append(_out)
+                out.append(_op.concatenate(_outputs, axis=1))
+            if r != 0:
+                _outputs = []
+                _inputs = [inputs_expr[i] for i in range(groups - r, groups)]
+                _weights = [weights_expr[i] for i in range(groups - r, groups)]
+                for (i, w) in zip(_inputs, _weights):
+                    _out = _op.nn.conv2d_transpose(data=i, weight=w, **params)
+                    if bias:
+                        _out = _op.nn.bias_add(_out, bias_expr)
+                    _outputs.append(_out)
+                out.append(_op.concatenate(_outputs, axis=1))
+            out = _op.concatenate(out, axis=1)
+        elif groups == 1:
+            out = _op.nn.conv2d_transpose(data=in_expr, weight=weight_expr, **params)
+            if bias:
+                out = _op.nn.bias_add(out, bias_expr)
+        else:
+            raise tvm.error.OpAttributeInvalid("Unable to handle.")
         return out
 
     def convert_slice(self, op):

tests/python/frontend/caffe/test_forward.py

@@ -35,6 +35,7 @@
 from caffe.proto import caffe_pb2 as pb
 
 import tvm
+import tvm.testing
 from tvm import relay
 from tvm.contrib import utils, graph_executor
 from tvm.contrib.download import download_testdata
@@ -451,6 +452,35 @@ def test_forward_Deconvolution():
             bias_filler=dict(type="xavier"),
         ),
     )
+    _test_deconvolution(
+        data,
+        convolution_param=dict(
+            num_output=16,
+            bias_term=False,
+            pad=0,
+            kernel_size=2,
+            stride=2,
+            dilation=1,
+            group=16,
+            weight_filler=dict(type="xavier"),
+            bias_filler=dict(type="xavier"),
+        ),
+    )
+    data = np.random.rand(1, 100, 32, 32).astype(np.float32)
+    _test_deconvolution(
+        data,
+        convolution_param=dict(
+            num_output=100,
+            bias_term=False,
+            pad=0,
+            kernel_size=2,
+            stride=2,
+            dilation=1,
+            group=100,
+            weight_filler=dict(type="xavier"),
+            bias_filler=dict(type="xavier"),
+        ),
+    )
 
 
 #######################################################################