Merge branch 'fix_reshape_bug' into 'master'

fix bug on reshape op & tar command & merge_duplicate_nodes

See merge request deep-computing/mace!1247

mace/ops/reduce.cc

@@ -1035,6 +1035,8 @@ class ReduceOp<DeviceType::GPU, float> : public ReduceOpBase {
 void RegisterReduce(OpRegistryBase *op_registry) {
   MACE_REGISTER_OP(op_registry, "Reduce", ReduceOp,
                    DeviceType::CPU, float);
+  MACE_REGISTER_OP(op_registry, "Reduce", ReduceOp,
+                   DeviceType::CPU, int);
 #ifdef MACE_ENABLE_QUANTIZE
   MACE_REGISTER_OP(op_registry, "Reduce", ReduceOp,
                    DeviceType::CPU, uint8_t);

mace/ops/reshape.cc

@@ -74,7 +74,7 @@ template <DeviceType D, class T>
 class ReshapeOp : public Operation {
  public:
   explicit ReshapeOp(OpConstructContext *context)
-      : Operation(context),
+      : Operation(context), dim_(Operation::GetRepeatedArgs<int>("dim")),
         has_df_(Operation::GetOptionalArg<int>("has_data_format", 0)) {}
 
   MaceStatus Run(OpContext *context) override {
@@ -85,18 +85,21 @@ class ReshapeOp : public Operation {
     const int32_t *shape_data = shape->data<int32_t>();
     const index_t num_dims = shape->dim_size() == 0 ? 0 : shape->dim(0);
     std::vector<index_t> out_shape;
-    MACE_RETURN_IF_ERROR(
-        GetOutputShape(input, shape_data, num_dims, &out_shape));
 
     // NHWC -> NCHW
-    if (has_df_ && D == DeviceType::CPU && out_shape.size() == 4 &&
-        shape->is_weight()) {
+    std::vector<int32_t> trans_shape_data(shape_data,
+                                          shape_data + shape->size());
+    if (has_df_ && D == DeviceType::CPU && shape->dim_size() == 4 &&
+        out_shape.size() == 4 && dim_.size() == 4) {
       std::vector<int> dst_dims = {0, 3, 1, 2};
-      std::vector<index_t> trans_shape =
-          TransposeShape<index_t, index_t>(out_shape, dst_dims);
-      out_shape = trans_shape;
+      std::vector<int32_t> tmp_shape =
+          TransposeShape<int32_t , int32_t>(trans_shape_data, dst_dims);
+      trans_shape_data = tmp_shape;
     }
 
+    MACE_RETURN_IF_ERROR(
+        GetOutputShape(input, trans_shape_data.data(), num_dims, &out_shape));
+
     Tensor *output = this->Output(OUTPUT);
     output->ReuseTensorBuffer(*input);
     output->Reshape(out_shape);
@@ -105,6 +108,7 @@ class ReshapeOp : public Operation {
   }
 
  private:
+  std::vector<int> dim_;
   bool has_df_;
 
  private:

tools/python/transform/tensorflow_converter.py

@@ -142,7 +142,6 @@ TFTransformGraphOptions = [
     'fold_old_batch_norms',
     'remove_control_dependencies',
     'strip_unused_nodes',
-    'merge_duplicate_nodes',
     'sort_by_execution_order'
 ]
 

tools/python/transform/transformer.py

@@ -1395,7 +1395,8 @@ class Transformer(base_converter.ConverterInterface):
         if op.type == MaceOp.Reshape:
             input_op = self._producer[op.input[0]]
             out_dims_len = len(op.output_shape[0].dims)
-            if len(input_op.output_shape[0].dims) != 4 \
+            if len(input_op.output_shape) != 1 or \
+                len(input_op.output_shape[0].dims) != 4 \
                     or (out_dims_len != 4 and out_dims_len != 2):
                 print("In this model, reshape is not transposable op.")
                 return False

tools/sh_commands.py

@@ -774,15 +774,7 @@ def packaging_lib(libmace_output_dir, project_name):
     six.print_("Start packaging '%s' libs into %s" % (project_name,
                                                       tar_package_path))
     which_sys = platform.system()
-    if which_sys == "Linux":
-        sh.tar(
-            "cvzf",
-            "%s" % tar_package_path,
-            glob.glob("%s/*" % project_dir),
-            "--exclude",
-            "%s/_tmp" % project_dir,
-            _fg=True)
-    elif which_sys == "Darwin":
+    if which_sys == "Linux" or which_sys == "Darwin":
         sh.tar(
             "--exclude",
             "%s/_tmp" % project_dir,