[pt2] bug fix: invert condition in checkFloatingOrComplex

[nkaretnikov]

Stack from ghstack:

• [pt2] add meta and SymInt support for linalg_matrix_exp #102945
• -> [pt2] bug fix: invert condition in checkFloatingOrComplex #102944

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/102944

• 📄 Preview Python docs built from this PR
• 📄 Preview C++ docs built from this PR
• ❓ Need help or want to give feedback on the CI? Visit the bot commands wiki or our office hours

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit d266966:
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[nkaretnikov]

CPP code:

static inline void checkFloatingOrComplex(const Tensor& t, const char* const f_name, const bool allow_low_precision_dtypes=true) {
  auto dtype = t.scalar_type();
  TORCH_CHECK((at::isFloatingType(dtype) || at::isComplexType(dtype)),
              f_name, ": Expected a floating point or complex tensor as input. Got ", dtype);
  if (!allow_low_precision_dtypes) {  // <--------- here
    TORCH_CHECK(dtype == kFloat || dtype == kDouble || dtype == kComplexFloat || dtype == kComplexDouble,
                f_name, ": Low precision dtypes not supported. Got ", dtype);
  }
}

[lezcano]

Approving to move forward, but can you add an error sample to one of the linalg ops to make sure we're testing this?

[nkaretnikov]

Note: waiting until #103240 lands (again) before I rebase this and merge.

UPD: Decided to land this first if possible. Looks like the PR I'm referencing here is not reviewed and has some failures.