Segfault with num_expansions=-1

[HaydenMcT]

Hello,

I seem to be getting segfaults when running maptree with num_expansions=-1, for several datasets. I've attached a small example dataset (10 rows, 2 features) where I seem to be encountering this issue (running from one of my department's linux machines, following the installation guide on the readme, using python 3.10.12): small.txt

Here is an example call which leads to a segfault for me:

from maptree import search as maptree_search
from experiments.globals import load_binary_data

alpha = 0.95
beta = 0.5
rho = (2.5, 2.5)
num_expansions = -1
time_limit = 60

X_train, y_train = load_binary_data('small.txt')

sol = maptree_search(X_train, y_train, alpha, beta, rho, num_expansions, time_limit)

Any guidance you can provide would be much appreciated!

[SullivanC19]

Does it segfault immediately or after a couple minutes?

On Fri, Jun 14, 2024 at 2:51 PM HaydenMcT @.***> wrote:

Hello,

I seem to be getting segfaults when running maptree with num_expansions=-1, for several datasets. I've attached a small example dataset (10 rows, 2 features) where I seem to be encountering this issue (running from one of my department's linux machines, following the installation guide on the readme, using python 3.10.12): small.txt https://github.com/user-attachments/files/15844101/small.txt

Here is an example call which leads to a segfault for me:

from maptree import search as maptree_search from experiments.globals import load_binary_data

alpha = 0.95 beta = 0.5 rho = (2.5, 2.5) num_expansions = -1 time_limit = 60

X_train, y_train = load_binary_data('small.txt')

sol = maptree_search(X_train, y_train, alpha, beta, rho, num_expansions, time_limit)

Any guidance you can provide would be much appreciated!

— Reply to this email directly, view it on GitHub https://github.com/ThrunGroup/maptree/issues/2, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADSFO67HRJP4NOKU2H2FZMDZHNQU3AVCNFSM6AAAAABJLBX5VKVHI2DSMVQWIX3LMV43ASLTON2WKOZSGM2TIMJVHE3DSNY . You are receiving this because you are subscribed to this thread.Message ID: @.***>

[HaydenMcT]

For this dataset, immediately.