!12041 unify parameter output type

From: @zhangbuxue
Reviewed-by: @ginfung,@zh_qh
Signed-off-by: @zh_qh

mindspore/common/parameter.py

@@ -105,7 +105,7 @@ class Parameter(Tensor_):
         >>> print(net(x))
         [[2.]]
         >>> net.weight.set_data(Tensor(np.zeros((1,2))))
-        Parameter (name=w)
+        Parameter (name=w, shape=(1, 2), dtype=Float64, requires_grad=True)
         >>> print(net(x))
         [[0.]]
     """

mindspore/core/abstract/prim_others.cc

@@ -168,6 +168,7 @@ AbstractBasePtr InferImplDepend(const AnalysisEnginePtr &, const PrimitivePtr &p
     MS_LOG(EXCEPTION) << primitive->name() << " input args size should be at lest 1, but got 0";
   }
   auto depends = args_spec_list[0]->Broaden();
+  // For scalar, need to set value to kAnyValue, because broaden scalar will not change the value.
   if (depends->isa<AbstractScalar>()) {
     depends->set_value(kAnyValue);
   }

mindspore/ops/operations/math_ops.py

@@ -218,7 +218,7 @@ class AssignAdd(PrimitiveWithInfer):
         >>> value = Tensor(np.ones([1]).astype(np.int64)*100)
         >>> output = net(value)
         >>> print(output)
-        Parameter (name=global_step)
+        Parameter (name=global_step, shape=(1,), dtype=Int64, requires_grad=True)
     """
     __mindspore_signature__ = (
         sig.make_sig('x', sig.sig_rw.RW_WRITE, dtype=sig.sig_dtype.T),
@@ -273,7 +273,7 @@ class AssignSub(PrimitiveWithInfer):
         >>> value = Tensor(np.ones([1]).astype(np.int32)*100)
         >>> output = net(value)
         >>> print(output)
-        Parameter (name=global_step)
+        Parameter (name=global_step, shape=(1,), dtype=Int32, requires_grad=True)
     """
 
     __mindspore_signature__ = (

mindspore/ops/operations/other_ops.py

@@ -54,7 +54,7 @@ class Assign(PrimitiveWithCheck):
         >>> net = Net()
         >>> output = net(x)
         >>> print(output)
-        Parameter (name=y)
+        Parameter (name=y, shape=(1,), dtype=Float32, requires_grad=True)
     """
     __mindspore_signature__ = (
         sig.make_sig('variable', sig.sig_rw.RW_WRITE, dtype=sig.sig_dtype.T),