update for matmul_v2 and reshape2 orig2prim

python/paddle/autograd/primrules.py

@@ -87,6 +87,11 @@ def trans(shape):
 
     assert len(x.shape) < 4 and len(
         y.shape) < 4, 'Do not support multi batchsize dimensions currently.'
+
+    if len(x.shape == 1):
+        x = broadcast(x, shape=[1, x.shape[0]])
+    if len(y.shape == 1):
+        y = broadcast(y, shape=[y.shape[0], 1])
     if op.attr('trans_x'):
         x = transpose(x, shape=trans(x.shape))
     if op.attr('trans_y'):
@@ -117,13 +122,15 @@ def tanh_orig2prim(op, x):
     return tanh(x)
 
 
-## NOTE(lml): The second output of reshape2 Xshape, can't be described by prim ops, use paddle.shape() interface instead.
+## NOTE(lml): The second output of reshape2 Xshape, which is only used in reshape2_grad, is meanlingless in new autograd mechanism, thus we use a zero tensor instead.
 @REGISTER_ORIG2PRIM('reshape2')
 def reshape2_orig2prim(op, shape_t, shape_tl, x):
     assert shape_t is None, 'Can not lower reshape2 into prim ops with shapetensor.'
     assert shape_tl is None, 'Can not lower reshape2 into prim ops with shapetensorlist.'
-    y, _ = get_output_vars(op)
-    return reshape(x, shape=y.shape), paddle.shape(x)
+    y, xshape = get_output_vars(op)
+    return reshape(
+        x, shape=y.shape), fill_const(
+            shape=xshape.shape, dtype=xshape.dtype, value=0.0)
 
 
 @REGISTER_ORIG2PRIM('concat')

python/paddle/autograd/primx.py

@@ -293,6 +293,7 @@ def transpose(self, ys_dot, xs_dot, ys_bar=None, retain_fwd=False):
 
         return ys_bar, xs_bar
 
+
 def _gradients(ys, xs, ys_bar=None):
     """ A drop-in replacement of paddle.gradients for computing
     the gradients of `xs` against `ys` using primitive ops based
@@ -383,8 +384,8 @@ def _lower(block, reverse, update_var_list):
     for op_idx in reversed(ops_to_remove):
         block._remove_op(op_idx)
     for var_name in vars_to_remove:
-        # block._remove_var(var_name)
-        del block.vars[var_name]
+        block._remove_var(var_name)
+        # del block.vars[var_name]
 
     if update_var_list is not None:
         for i in range(len(update_var_list)):