Skip to content
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.

Sign up for GitHub

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

Jump to bottom

FTRL fails with wrong response - non-deterministically #1891

Closed
arnocandel opened this issue Jun 28, 2019 · 2 comments · Fixed by #1892
Closed

FTRL fails with wrong response - non-deterministically #1891

arnocandel opened this issue Jun 28, 2019 · 2 comments · Fixed by #1892
Assignees
@oleksiyskononenko
Labels
bug Any bugs / errors in datatable; however for severe bugs use [segfault] label
Milestone
Release 0.9.0

Comments

@arnocandel
Copy link
Member

arnocandel commented Jun 28, 2019 •
edited
Loading

0.8.0.dev91

debug.pkl.zip

>>> model, X, y, fit_args = pickle.load(open("debug.pkl", "rb"))
>>> model.fit(X, y, **fit_args)
ValueError: Target column for binomial problem cannot contain more than two labels
>>> model, X, y, fit_args = pickle.load(open("debug.pkl", "rb"))
>>> model.fit(X, y, **fit_args)
FtrlFitOutput(epoch=1.3996499781236327, loss=0.4652366042137146)
>>> model, X, y, fit_args = pickle.load(open("debug.pkl", "rb"))
>>> model.fit(X, y, **fit_args)
FtrlFitOutput(epoch=1.3996499781236327, loss=0.4652366042137146)
>>> model, X, y, fit_args = pickle.load(open("debug.pkl", "rb"))
>>> model.fit(X, y, **fit_args)
ValueError: Target column for binomial problem cannot contain more than two labels
@arnocandel arnocandel added the bug Any bugs / errors in datatable; however for severe bugs use [segfault] label label Jun 28, 2019
@oleksiyskononenko
Copy link
Contributor

Can reproduce it on my side, will take a look.

@oleksiyskononenko oleksiyskononenko mentioned this issue Jun 28, 2019
Merged
@oleksiyskononenko
Copy link
Contributor

This ^^^ PR fixes the problem.

st-pasha pushed a commit that referenced this issue Jun 28, 2019
@oleksiyskononenko @st-pasha
Do binomial check in label_encode functions only when adding new labe…
43ded51 
…ls (#1892)

For binomial problem number of labels should be less or equal to two. This check should only be performed when a new label is encountered in the target column. When running label encoding in parallel, we now do the check making sure the label was not added from another thread.

Closes #1891
@st-pasha st-pasha added this to the Release 0.9.0 milestone Jan 29, 2020
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@oleksiyskononenko oleksiyskononenko

Labels
bug Any bugs / errors in datatable; however for severe bugs use [segfault] label
Projects
None yet
Milestone
Release 0.9.0
Development

Successfully merging a pull request may close this issue.

Do binomial check in label_encode functions only when adding new labels h2oai/datatable
3 participants
@st-pasha @arnocandel @oleksiyskononenko