Bump version to 1.8.1

[fritzo]

In preparation for 1.8.1 release:

• update version checks in notebooks to 1.8.1
• update funsor pinned version to 0.4.3
• xfailed a bunch of pyro.contrib.funsor enumeration tests Some pyro.contrib.funsor tests are failing in torch==1.11.0 #3046

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[fritzo]

Looks like we're seeing #3046 again with funsor==0.4.3. @ordabayevy to try to get tests to pass I've reverted #3048 (which had simply removed xfails)

[fritzo]

Gosh looking at the funsor changelog, it looks like the only relevant change was @eb8680's and my #584 which modified partial_sum_product() 🤔

[fritzo]

Indeed git bisect shows the offending commit is @eb8680's and my pyro-ppl/funsor#584

[eb8680]

It looks like re-running the tests made them pass? There were at least a couple tests off by a factor of 2 rather than numerical error, which is odd. I'll look into the failures this evening. I don't think it's worth blocking a Pyro release on contrib.funsor bugs though - I'd be OK with either releasing as-is with some tests marked xfail or reverting the change in Funsor. I'll look into the failures tonight.

[fritzo]

@eb8680 the tests passed because I xfailed them 😊 LMK if you want to pair debug on our funsor PR. I'm ok either reverting that PR or releasing pyro with some contrib.funsor bugs, but I'd prefer to do a quick bugfix release of funsor 0.4.4 if we can find & fix the error. It took so long to get that PR merged...

[fritzo]

@fehiepsi-- @eb8680 are ok merging this with the xfailed tests in contrib.funsor and we can look into fixing those tests later. I think the PyTorch update is higher priority.

[fritzo]

Thanks @martinjankowiak !