latest tf_sparse only support tf2.*?

[caowencai]

tf_sparse/tf_sparse/sparse_matrix.py

Line 12 in d3066fb

from tensorflow.python.ops.linalg.sparse.sparse_csr_matrix_ops import sparse_tensor_to_csr_sparse_matrix, \

As showed above, tensorflow.python.ops.linalg.sparse only exists in tf==2.* version，and there exists import error at tf_sparse here when using tf_geometric with tf=1.5.0

[hujunxianligong]

Yes. Currently tf_sparse only supports tf 2.x. We will consider to make it compatible with tf 1.x in the future.

…

Message ID: ***@***.***>

[caowencai]

Yes. Currently tf_sparse only supports tf 2.x. We will consider to make it compatible with tf 1.x in the future.
…
Message ID: @.***>

so also, tf_geometric only supports tf 2.x? So upset.

[hujunxianligong]

I think if you're using tf 1.x with eager execution, you should upgrade to tf 2.x.
It's more efficient in many cases.

[caowencai]

I think if you're using tf 1.x with eager execution, you should upgrade to tf 2.x. It's more efficient in many cases.

You're right. But our seving system is using tf1.x, which cannot be destroyed by Artillery of Italy ，哈哈