Set nonzero tolerance for RHO_REF_ME with dace_gpu in test_parallel_grid_manager.py#1343
Conversation
|
Mandatory Tests Please make sure you run these tests via comment before you merge!
Optional Tests To run benchmarks you can use:
To run tests and benchmarks with the DaCe backend you can use:
To run test levels ignored by the default test suite (mostly simple datatest for static fields computations) you can use:
For more detailed information please look at CI in the EXCLAIM universe. |
| model_backends.is_cpu_backend(backend) | ||
| or ( | ||
| model_backends.is_gpu_backend(backend) | ||
| and attrs_name == metrics_attributes.DDQZ_Z_FULL_E | ||
| and attrs_name | ||
| in { | ||
| metrics_attributes.DDQZ_Z_FULL_E, | ||
| metrics_attributes.RHO_REF_ME, | ||
| } |
There was a problem hiding this comment.
I am a bit suspicious about the CPU backend being much more permissive, since it does not filter the test cases, although it is unrelated to this PR.
There was a problem hiding this comment.
Yeah, you're right to be suspcicious, but I think it's just a consequence of the special cases being added in separate PRs. For GPU I started adding whitelists for individual fields while CPU was at some point just set to non-zero for all fields without checking which ones need it.
I'm hoping #1303 will improve this a bit. Then dace_cpu should always be expected to have zero diff.
|
cscs-ci run default |
|
cscs-ci run distributed |
|
Restarted https://gitlab.com/cscs-ci/ci-testing/webhook-ci/mirrors/5125340235196978/2255149825504675/-/jobs/15031642841 manually. CI-related failure, not because of icon4py. |
3 participants
I don't think this is a regression. This field is just not tested by default in the usual default pipeline (one of many "extra" fields that are skipped because they take a long time to test for little added value).