[RLlib] Issue 21109: Action unsquashing causes inf/NaN actions for unbounded action spaces.

[sven1977]

Issue 21109: Action unsquashing causes inf/NaN actions for unbounded action spaces.

Why are these changes needed?

Closes issue 21109

Related issue number

#21109

Checks

• I've run scripts/format.sh to lint the changes in this PR.
• I've included any doc changes needed for https://docs.ray.io/en/master/.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
• Testing Strategy
  • Unit tests
  • Release tests
  • This PR is not tested :(

[gjoliver]

1 quick suggestion. thanks for the fix.

[sven1977]

Thanks for the review! I can't find your suggestion, though :D

[gjoliver]

Thanks for the review! I can't find your suggestion, though :D

ah, that's really weird!
I was saying maybe we should just keep using (s.dtype == np.float32 or s.dtype == np.float64).
using substring match for type check feels a bit hacky.
personal opinion 😆

[gjoliver]

appreciate the fix!