enable group conv through miopen

Author: Peter Yeh

python/tvm/contrib/miopen.py

@@ -50,7 +50,8 @@ def conv2d_forward(x,
                    dilation_h=1,
                    dilation_w=1,
                    conv_mode=0,
-                   data_type=1):
+                   data_type=1,
+                   group_count=1):
     """Create an extern op that compute 2D convolution with MIOpen
 
     Parameters
@@ -77,13 +78,16 @@ def conv2d_forward(x,
     data_type: int
         0: miopenHalf (fp16)
         1: miopenFloat (fp32)
-
+    group_count: int
+        number of groups 
     Returns
     -------
     y: Tensor
         The result tensor
     """
-    assert (conv_mode == 0 or conv_mode == 1), "0: miopenConvolution / 1: miopenTranspose"
+    assert (0 <= conv_mode <= 2), "0: miopenConvolution / 1: miopenTranspose / 2: miopenGroupConv"
+    if group_count > 1:
+        conv_mode = 2
     oshape = np.zeros((len(x.shape)), dtype=np.int32)
     xshape = x.shape
     wshape = w.shape
@@ -104,6 +108,7 @@ def conv2d_forward(x,
                       wshape[1].value,
                       wshape[2].value,
                       wshape[3].value,
+                      group_count,
                       _get_np_int32_array_handle(oshape))
 
     return _api.extern(

src/contrib/miopen/conv_forward.cc

@@ -50,16 +50,20 @@ TVM_REGISTER_GLOBAL("tvm.contrib.miopen.conv2d.setup")
   const int w_dim1 = args[13];
   const int w_dim2 = args[14];
   const int w_dim3 = args[15];
-  void *out_shape = args[16];
+  const int n_group = args[16];
+  void *out_shape = args[17];
 
   MIOpenThreadEntry* entry_ptr = MIOpenThreadEntry::ThreadLocal();
+  assert(n_group > 0 && "Group Size > 0 is expected");
+  if (n_group > 1)
+    assert(mode > 1 && "Group /Depthwise Conv mode when num of groups > 1");
   // Set Mode
   entry_ptr->conv_entry.mode = static_cast<miopenConvolutionMode_t>(mode);
   // Set Ctx
   entry_ptr->conv_entry.ctx = TVMContext{kDLROCM, 0};
   // Set Data Type
   entry_ptr->conv_entry.data_type = static_cast<miopenDataType_t>(
-      dtype);  // MIOpen supports fp32(miopenFloat), fp16(miopenHalf) at
+      dtype);  // MIOpen supports fp32(miopenFloat), fp16(miopenHalf), int32, int8 at
                // this moment.
   // Set Desc
   MIOPEN_CALL(miopenInitConvolutionDescriptor(entry_ptr->conv_entry.conv_desc,
@@ -70,11 +74,13 @@ TVM_REGISTER_GLOBAL("tvm.contrib.miopen.conv2d.setup")
                                               stride_w,
                                               dilation_h,
                                               dilation_w));
+  if (n_group > 1)
+    MIOPEN_CALL(miopenSetConvolutionGroupCount(entry_ptr->conv_entry.conv_desc, n_group));
   // Set Filter
   MIOPEN_CALL(miopenSet4dTensorDescriptor(entry_ptr->conv_entry.filter_desc,
                                           entry_ptr->conv_entry.data_type,
                                           w_dim0,
-                                          w_dim1,
+                                          w_dim1/n_group,
                                           w_dim2,
                                           w_dim3));
   // Set Input