New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Letor updates #324
base: master
Are you sure you want to change the base?
Letor updates #324
Conversation
I've rebased onto current master, which was clean apart from |
bfe50b5
to
313eca9
Compare
We are still seeing very similar looking test failures in the linux clang build and on macos (macos uses clang as the default compiler so these may well be the same issue). I've tested with clang (with ubsan and asan enabled) locally and have a clean test run (after a few fixes for merge problems and one bug). I think I'm going to try to peel out logically separate changes and test and merge them separately to try to get to the bottom of this. I've already cherry-picked and merged the changes to drop SVMRanker. I'll likely rebase and force-push this branch to drop out merged changes. If anyone wants to help with this, please speak up and we can come up with a plan. |
This change applies to both ListNetRanker and ListMleRanker. The motivation for Xavier initialization in Neural Networks is to initialize the weights of the network so that the neuron activation functions are not starting out in saturated or dead regions. In other words, we want to initialize the parameters with random values that are not “too small” and not “too large.”
In effect, a bias value allows you to shift the activation function to the left or right, which may be critical for successful learning.
This are due to changes made from fixing ranker implementations, fixing indexing errors, adding Xavier initialisation, adding normalisation of gradient and adding bias combined. This also fixes scorer test which was wrong earlier.
Hardwire it to what default_random_engine gives with current GCC, since that passes the test suite. Maybe this is why some CI builds fail some tests...
I've merged several commits to master, and have reached the point where I can't easily peel off more because there are 3 commits which change the testsuite results, and one commit which updates the testsuite to pass again. I noticed the Xavier initialisation was using That's helped, but there's still one test failing with clang and on macos, but passing elsewhere (and for me locally with clang). There are also two XFAIL-ed testcases passing (which I don't see locally either):
|
This is a replacement for #320, updated for the change from travis to GHA.