zhiics commented on a change in pull request #5855: URL: https://github.com/apache/incubator-tvm/pull/5855#discussion_r444355425
########## File path: tests/python/relay/test_vm_serialization.py ########## @@ -332,6 +284,20 @@ def test_mobilenet(): run_network(mod, params) +def test_vm_shape_of(): Review comment: `vm.shape_of` is implicitly added by memory manifest pass when a dynamic input is detected. It is used to get the shape of a tensor at runtime and it has to be on the CPU as the following instructions (i.e. `shape_func`) would be on CPU as well. The normal `shape_of` instruction doesn't have this constraint so we don't actually convert them. https://github.com/apache/incubator-tvm/blob/ef9bf7d957665901c6980b9ada2aeb8e38d435c6/python/tvm/relay/transform/memory_alloc.py#L141 ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org