I am using TVM frontend to load Tensorflow models, and some placeholders’ shapes are dynamic with -1 . How does TVM solve this problem?
I think the answer is yes based on a recent merged PR: https://github.com/apache/incubator-tvm/pull/4285.
We are working on general dynamic shape supporting. Soon we will be able to parse tf models with symbolic shapes. There is still more work to do to get descent performance.