pydrake: Should ensure that torch>=1.5.0 does not create an issue, relax warning #13707
Labels
component: pydrake
Python API and its supporting Starlark macros
configuration: python
priority: medium
unused team: manipulation
In #12073, the solution (#12108) was to print a warning that points to this issue:
pytorch/pytorch#3059 (comment)
That pytorch issue was closed in the following commit, which is part of
torch>=1.5.0
:pytorch/pytorch@ddff4ef
However, if I now test in a
virtualenv
withtorch==1.5.1
, our warning still triggers:I'm bringing this up b/c I'm looking into #13571 and
isort
, and in Anzu, our import logic gets a little messy if we leave the condition thatpydrake
should be imported beforepytorch
. (Also, false positives here are annoying.)The text was updated successfully, but these errors were encountered: