Run trunc tests with Float32 thanks to new Enzyme

[kshyatt]

No description provided.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Jutho]

I am surprised this works; I actually moved also the Mooncake tests to Float64 after some issues in #180 .

[kshyatt]

Could it be the custom finite differences method Enzyme is using for fp32?

[kshyatt]

Anyone have objections to me merging this?

[Jutho]

Not really, though as mentioned, I actually moved LQ and QR tests for Mooncake and Enzyme from Float32 to Float64, so this is moving in the opposite direction of consistency.

[Jutho]

Maybe worth testing again if that change of mine was really necessary, or resulting from an underlying problem that I only fixed afterwards?

[kshyatt]

Let me check that also, then

[kshyatt]

I see the failures locally too. Maybe we can test both Float32 and Float64 for these Enzyme + truncation tests, considering they now take much less time, and it was previously erroring out in a horrible way?

[Jutho]

I am fine with any suggestion. I should probably revisit the QR/LQ tests, since I would think that the eig / svd tests are more sensitive to numerical noise than QR/LQ, so if former manage to work for Float32, then why not the latter.