提交 0cc1d7ba authored 作者: Razvan Pascanu's avatar Razvan Pascanu

Bug in infer shape

When collecting shapes from the outputs of the inner function, we should keep track that mit_mot arguments have multiple outputs for one argument.
上级 692ab088
......@@ -1086,8 +1086,9 @@ class Scan(PureOp):
offset = 1 + self.n_seqs
scan_outs = [x for x in input_shapes[offset:offset + n_outs]]
offset += n_outs
outs_shape_n = self.n_mit_mot_outs + self.n_mit_sot + self.n_sit_sot
for x in xrange(self.n_nit_sot):
out_shape_x = outs_shape[n_outs + x]
out_shape_x = outs_shape[outs_shape_n + x]
if out_shape_x is None:
# This output is not a tensor, and has no shape
scan_outs.append(None)
......
Markdown 格式
0%
您添加了 0 到此讨论。请谨慎行事。
请先完成此评论的编辑!
注册 或者 后发表评论