Fix: MPI reproducibility #38
Draft
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
alecksphillips
force-pushed
the
fix/MPI-reproducibility
branch
2 times, most recently
from
58231e1 to
c9190bb
Compare
separate calculation of cdf in get_num_copies to test separately add test (currently fails due to numerical inaccuracy)
Add logcumsumexp to Executor and Executor_MPI with accompanying tests Executor_MPI.logsumexp(x) currently implemented naively as log(exec.cumsum(log(x))) and thus fails tests due to numerical inaccuracy
Use exec.logcumsumexp in calculate cdf for numerical stability (currently still unstable with MPI)
Mark test_logcumsumexp in test_executor_MPI
For MPI implementation of logcumsumexp, use inclusive_prefix_logsumexp
from redistribution submodule.
Tests in tests/test_executor_MPI.py all pass including for large array
of numbers for logcumsumexp, however running working example (e.g.
examples/mpi/decision_tree_example.py) with MPI results in error when
performing reduction operation:
```
Fatal Python error: exception in user-defined reduction operation
Traceback (most recent call last):
File "mpi4py\MPI\opimpl.pxi", line 99, in mpi4py.MPI.op_user_mpi
File "mpi4py\MPI\opimpl.pxi", line 90, in mpi4py.MPI.op_user_py
File
"...\discretesampling\base\executor\MPI\distributed_fixed_size_redistribution\prefix_sum.py",
line 7, in LSE~
x = np.frombuffer(xmem, dtype=np.float64)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ValueError: buffer size must be a multiple of element size
```
alecksphillips
force-pushed
the
fix/MPI-reproducibility
branch
from
c9190bb to
c45e8e3
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There is currently a lack of reproducibility with MPI, due to numerical instability when calculating the cdf for resampling.
This pull request will (eventually) do a few things: