fix h class model on OrangePi (#2220)

Author: lvyufeng

mindtorch/_apis/npu.py

@@ -791,10 +791,10 @@ def select(condition, input, other):
         return pyboost.select_op(condition, input, other)
     return legacy.select(condition, input, other)
 
-def mean(input, axis, keepdims, dtype):
+def mean(input, dim, keepdim, dtype):
     if use_pyboost():
-        return pyboost.mean_ext_op(input, axis, keepdims, dtype)
-    return legacy.reduce_mean(input, axis, keepdims)
+        return pyboost.mean_ext_op(input, dim, keepdim, dtype)
+    return legacy.reduce_mean(input, dim, keepdim)
 
 def index(input, index):
     if use_pyboost():
@@ -1552,9 +1552,29 @@ def one_hot(tensor, num_classes):
     return legacy.one_hot(tensor, num_classes, on_value, off_value, -1)
 
 def var(input, dim=None, correction=1, keepdim=False):
-    if use_pyboost():
+    if use_pyboost() and not ON_ORANGE_PI:
         return pyboost.var_op(input, dim, correction, keepdim)
-    return legacy.var(input, dim, correction, keepdim)
+    if dim is None:
+        input_mean = mean(input, (), False, None)
+    else:
+        input_mean = mean(input, dim=dim, keepdim=True, dtype=None)
+    
+    # 计算与均值的平方差
+    squared_diff = pow(sub(input, input_mean, 1), 2) 
+    # 计算方差
+    if dim is None:
+        variance = mean(squared_diff, (), False, None)
+        n = input.numel()  # 总元素个数
+    else:
+        variance = mean(squared_diff, dim=dim, keepdim=keepdim, dtype=None)
+        n = input.size(dim)  # 指定维度的元素个数
+    
+    # 无偏估计校正
+    if correction and n > 1:
+        variance = mul(variance, (n / (n - 1)))
+    
+    return variance
+
 
 def linspace(start, end, steps, dtype=None):
     if use_pyboost() and not ON_ORANGE_PI: