You are viewing a plain text version of this content. The canonical link for it is here.

Posted to dev@tvm.apache.org by heliqi <no...@github.com> on 2020/04/01 04:38:11 UTC

Re: [apache/incubator-tvm] [RFC] Enhance TensorFlow Frontend Control Flow Support (#4969)

Have you try the nlp model? I use latest code, dynamic sahpe don't working.
recursively find the input to the control flow nodes , have some problems with fixed dynamic input.

For example, I set the 'Placeholder' op(the original shape is dynamic (-1,-1)) shape as (1,30) in the from_tensorflow interface.First, for each control flow node, we backtrack all its ancestor nodes until input nodes.But all the nodes we first backtracked do not necessarily contain 'Placeholder' node, then the shape of these nodes is dynamic.until some control node search for and get the 'Placeholder' node.

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/apache/incubator-tvm/issues/4969#issuecomment-607026858