Fix the value error typo of AdamW's betas' valid values checking

[dourgey]

What does this PR do?

Fixes the value error typo of AdamW's betas' valid values checking(raise ValueError())
"should be in [0.0, 1.0[" -> "should be in [0.0, 1.0]"

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a Github issue or the forum? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

[sgugger]

The test asserts that betas[0] < 1.0 so it seems like the current error is correct.

[dourgey]

The test asserts that betas[0] < 1.0 so it seems like the current error is correct.

Thanks for your reply.
I mean, that there is a incorrect format of the interval indicates, as shown below:

if I initialize a AdamW optimizer with error betas' values like [0.1, 2], there would raise a ValueError like this:

It seems like a incorrect format of the interval indicates. The correct format may should be Invalid beta parameter: 2 - should be in [0.0, 1.0], but the current version is Invalid beta parameter: 2 - should be in [0.0, 1.0[. Is my understanding correct?

And after fixed, the error is shown as follows:

[sgugger]

No, both beta1 and beta2 are tests to be < 1.0, not <= 1.0, so the upper bound needs to be excluded. If you put 1.0 as a value, you will get the error.

[dourgey]

No, both beta1 and beta2 are tests to be < 1.0, not <= 1.0, so the upper bound needs to be excluded. If you put 1.0 as a value, you will get the error.

Yeah you are right. I made a mistake on this. Besides, my question is the error message should be in [0.0, 1.0[ is a correct expression form of [0.0, 1.0) or a spelling mistake? I see that the expression form in original paper of Adam is:

If it's a spelling mistake, I submit a new commit, please have a look, thanks a lot.

[sgugger]

It's the European (or at least French, not sure about all the neighbors!) way of writing an interval without the upper bound, yours is the American way ;-)

It's more inline with the paper so let's take yours.