Closed aftersomemath closed 2 days ago
I'm getting
[ RUN ] MuJoCoRolloutTest.test_detect_divergence
[third_party/crosstool/v18/stable/src/libcxx/include/__vector/vector.h:1114](https://cs.corp.google.com/piper///depot/google3/third_party/crosstool/v18/stable/src/libcxx/include/__vector/vector.h?l=1114&ws=tap-presubmit-rerun-server/51936822&snapshot=2): assertion __n < size() failed: vector[] index out of bounds
In line 233, shouldn't you be passing model_ptrs.data()
? But I don't understand how the test are passing on GH...
It looks like the internal check passed, but I did not do anything. I'm worried that there is a bug now (and passing a raw pointer using model_ptrs.data()
would just mask that). Was anything done that made the internal check pass?
They are not passing, I don't think. I'm a bit confused by the situation, but I trust @nimrod-gileadi will sort it out.
I found the bug, I created the std::vector
incorrectly.
All requested changes implemented.
rollout
can now accept a list of models of lengthnroll
, similar to its other arguments.The models must be compatible, meaning
nstate
,ncontrol
,nv
, andnsensordata
must be equal.