GraphSage无监督算法为何需要设置feature_idx和feature_dim两个参数？

Hello，
我们使用Graphsage算法模型训练，由于我们训练数据没有稠密特征，所以，训练时没有设置--feature_idx --feature_dim 这两个参数，运行的过程中有以下报错：

File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/op_def_library.py", line 571, in _apply_op_helper
(input_name, op_type_name, len(values), num_attr.minimum))
ValueError: List argument 'values' to 'ConcatV2' Op with length 0 shorter than minimum length 2.

而将feature_idx和feature_dim设置成-1和0时（--feature_idx -1 --feature_dim 0），同样出现如下错误：
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/op_def_library.py", line 571, in _apply_op_helper
(input_name, op_type_name, len(values), num_attr.minimum))
ValueError: List argument 'values' to 'ConcatV2' Op with length 0 shorter than minimum length 2.

当将feature_idx设置成0，feature_dim设置成0时，出现如下错误：
InvalidArgumentError (see above for traceback): Reshape cannot infer the missing input size for an empty tensor unless all specified input sizes are non-zero
[[node supervisedgraphsage_1/sageencoder_1/shallowencoder_1/Reshape_1 (defined at /opt/euler-0.1.2/tf_euler/python/encoders.py:162) ]]

@yangsiran 所以，想咨询一下，对应训练数据没有稠密特征，如何使用Graphsage算法进行训练？

Start by tracing the GraphSage feature_idx and feature_dim handling in tf_euler/python/encoders.py at line 162, then reproduce the reported empty-feature cases. Compare the resulting ConcatV2 and Reshape failures and determine the supported behavior for training without dense features; done means the valid configuration or required limitation is documented and the errors are addressed or clearly explained.

Written by the indexing model from the issue text.