High CPU utilization with sklearn, function never returns

[jeremiahcompass]

Hello and thanks for Colima.
I found an issue today where using the SciKit Learn (v0.18.2) Linear Regression fit causes high CPU utilization (around 100%-200%) on the container and doesn't complete. So far we've reproduced the issue on 3 developer's systems running Colima 0.2.2 (MacBook Pros Intel CPU) and verified the function returns quickly (<3ms) when running the container with Docker Desktop.
I also reproduced the issue on

colima version HEAD-e1df012
git commit: e1df01257a5aa73a373e04f965d803697df41564

runtime: docker
arch: x86_64
client: v20.10.11
server: v20.10.7`

Here's the python we are running to repro the issue:

import sklearn.linear_model
import time

source_points = [(0.22568627446889877, 0.055740684270858765), (0.12333333119750023, 0.8582248091697693), (0.20058824121952057, 0.9266889095306396)]
target_points = [(0.25509804487228394, 0.9863677620887756), (0.25509804487228394, 0.9863677620887756), (0.25509804487228394, 0.9863677620887756)]

linear_regressor = sklearn.linear_model.LinearRegression()

start = time.monotonic()
linear_regressor.fit(source_points, target_points)
print(f'It took {((time.monotonic() - start) * 1_000_000):.0f}μs')

[abiosoft]

@jeremiahcompass is this running on a mounted volume? How much resources (cpu, ram) is allocated to the VM?

[jeremiahcompass]

@abiosoft Yes, we're using a mounted volume. At first I was using the default CPUs and RAM, then I tried 4 CPUs and 16GB ram with the same result.