Pyro: Test failure of enum-parallel gradients after PyTorch #5776

Created on 21 Mar 2018  路  5Comments  路  Source: pyro-ppl/pyro

With the latest pytorch master, many parallel enum tests in test_enum are failing due to mismatch in the gradient computation.

To replicate, checkout the commit 5fa3aac610ee234338dbc11eb5b6d4a133cb483d in PyTorch master (https://github.com/pytorch/pytorch/pull/5776), build PyTorch and run these tests

pytest -v --tb=short tests/infer/test_enum.py

Example of a failing test - test_elbo_iarange_iarange 2-2-None-None-parallel-None.

@fritzo, @eb8680 - I thought that there could be some unexpected interactions between the dice elbo change and upstream PyTorch. Turns out that is not exactly the case as 11 of our tests fail even before the dice elbo change, but there are more failures (79) with dice elbo. Could you guys take a look?

This could either be a Pyro bug or something in PyTorch upstream.

bug high priority

Most helpful comment

I found the bug, I'll send a patch soon.

All 5 comments

Hey @neerajprad, thank you for your post, I'm looking into this now.

I found the bug, I'll send a patch soon.

Thanks, @cpuhrsch! Curious to see where the bug was.

Fixed upstream by pytorch/pytorch#5926 and in Pyro by #917.

Was this page helpful?
0 / 5 - 0 ratings

Related issues

neerajprad picture neerajprad  路  4Comments

lundlab-kaltinel picture lundlab-kaltinel  路  3Comments

martinjankowiak picture martinjankowiak  路  3Comments

null-a picture null-a  路  4Comments

jpchen picture jpchen  路  5Comments