Remove check_pickle argument in delayed.

[jjerphan]

Context: check_pickle should have been removed in 0.13.0 as indicated in the DeprecationWarning.

Changes: This PR removes this argument, behavior, and the associated test.

Associated Issue: #900

[pierreglaser]

Hi @jjerphan, why was this closed? We should probably merge this before releasing joblib 0.15 / 1.0

[jjerphan]

Hi @pierreglaser ; I just made some housekeeping on my stalled PRs, I reopen it now.

[pierreglaser]

@jjerphan If you find some time, feel free to finish this up :) we should release joblib early next week.

[codecov]

Codecov Report

Merging #903 (264633b) into master (53a8173) will increase coverage by 0.62%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master     #903      +/-   ##
==========================================
+ Coverage   93.93%   94.56%   +0.62%     
==========================================
  Files          47       47              
  Lines        6956     6933      -23     
==========================================
+ Hits         6534     6556      +22     
+ Misses        422      377      -45     

Impacted Files	Coverage Δ
joblib/test/test_parallel.py	97.09% <ø> (+0.17%)	⬆️
joblib/parallel.py	96.41% <100.00%> (-0.04%)	⬇️
joblib/test/test_memory.py	98.56% <0.00%> (+0.14%)	⬆️
joblib/pool.py	87.80% <0.00%> (+0.81%)	⬆️
joblib/func_inspect.py	92.26% <0.00%> (+1.19%)	⬆️
joblib/logger.py	86.84% <0.00%> (+1.31%)	⬆️
joblib/_parallel_backends.py	96.48% <0.00%> (+1.56%)	⬆️
joblib/disk.py	92.06% <0.00%> (+1.58%)	⬆️
joblib/test/test_memmapping.py	99.23% <0.00%> (+1.90%)	⬆️
... and 3 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 53a8173...264633b. Read the comment docs.

[jjerphan]

Hi @pierreglaser. I haven't managed to find a reason for the "failure" of the codecov pipeline.

[pierreglaser]

Don't worry about the codecov complaints. This issue is affecting a lot of PRs, and I think joblib should use a coverage threshold like scikit-learn does.

[pierreglaser]

@ogrisel if you want to take a look :) that would be nice to have this merged for 0.16.

[jjerphan]

OK, thank you @pierreglaser for this piece of feedback.

[jjerphan]

Small up.

[tomMoral]

This seems a bit overdue! thx @jjerphan.

Could you simply add an entry in the change log saying that you removed this deprecated argument?

[jjerphan]

@tomMoral: thanks for the follow up! 🙂

CI checks on Azure are still failing but it is unrelated to those changes.

[ogrisel]

CI checks on Azure are still failing but it is unrelated to those changes.

It looks all green to me.

[ogrisel]

Thanks!

[tomMoral]

Thanks!

[dmalzl]

Hi there,

I am currently trying to run pomgranate to fit a hidden markov model to some data. This library uses joblib to parallelize the model training. However, when I try to train it I get the following error:

Empty                                     Traceback (most recent call last)
~/.conda/envs/mypython/lib/python3.6/site-packages/joblib/parallel.py in dispatch_one_batch(self, iterator)
    819             try:
--> 820                 tasks = self._ready_batches.get(block=False)
    821             except queue.Empty:

~/.conda/envs/mypython/lib/python3.6/queue.py in get(self, block, timeout)
    160                 if not self._qsize():
--> 161                     raise Empty
    162             elif timeout is None:

Empty:

During handling of the above exception, another exception occurred:

TypeError                                 Traceback (most recent call last)
<ipython-input-7-dcc6125ae320> in <module>
----> 1 df = hmm(df, 3)

~/.conda/envs/mypython/lib/python3.6/site-packages/hmm_bigwigs/bigwig_hmm.py in hmm(df, num_states)
     40     vals = df["value"].values
     41     model = HiddenMarkovModel.from_samples(
---> 42         NormalDistribution, X=[vals], n_components=num_states
     43     )
     44     states = model.predict(vals)

~/.conda/envs/mypython/lib/python3.6/site-packages/pomegranate/hmm.pyx in pomegranate.hmm.HiddenMarkovModel.from_samples()

~/.conda/envs/mypython/lib/python3.6/site-packages/pomegranate/kmeans.pyx in pomegranate.kmeans.Kmeans.fit()

~/.conda/envs/mypython/lib/python3.6/site-packages/pomegranate/kmeans.pyx in genexpr()

~/.conda/envs/mypython/lib/python3.6/site-packages/joblib/parallel.py in __call__(self, iterable)
   1039             # remaining jobs.
   1040             self._iterating = False
-> 1041             if self.dispatch_one_batch(iterator):
   1042                 self._iterating = self._original_iterator is not None
   1043

~/.conda/envs/mypython/lib/python3.6/site-packages/joblib/parallel.py in dispatch_one_batch(self, iterator)
    829                 big_batch_size = batch_size * n_jobs
    830
--> 831                 islice = list(itertools.islice(iterator, big_batch_size))
    832                 if len(islice) == 0:
    833                     return False

~/.conda/envs/mypython/lib/python3.6/site-packages/pomegranate/kmeans.pyx in genexpr()

TypeError: delayed() got an unexpected keyword argument 'check_pickle'

I checked the input file and every variable I feed the model and all is fine there. I tried to find out where the 'check_pickle' kwarg came from but I was not able to find the culprit. Do you know what's going on?

Thanks

[jjerphan]

@dmalzl: the check_pickle has been removed from delayed in joblib. This has been fixed accordingly in pomegranate by jmschrei/pomegranate@42d14be. Updating pomgranate to v0.14.0 should fix the problem:

pip install - U pomegranate

[dmalzl]

Ah okay. Since I just installed it today, I thought I already had the newest version. Seems the conda recipe is outdated then. That's what you get from blind trust.

Anyway, works now after updating.
Thanks for the help

[jjerphan]

You're welcome.