Can't run .fit() inside docker container. stan_model/prophet_model.bin -> Segmentation Fault

[MattiaPeiretti]

I am running prophet 1.1.6 in docker, and i get the following error:

RuntimeError: Error during optimization! Command '/usr/local/lib/python3.12/site-packages/prophet/stan_model/prophet_model.bin random seed=4955 data file=/tmp/tmp8nj_ajwk/d6fyv2_s.json init=/tmp/tmp8nj_ajwk/sh420jq2.json output file=/tmp/tmp8nj_ajwk/prophet_modelsihs9kil/prophet_model-20250109155857.csv method=optimize algorithm=newton iter=10000' failed: 

When I exec into the container, and try to run

/usr/local/lib/python3.12/site-packages/prophet/stan_model/prophet_model.bin random seed=4955 data file=/tmp/tmp8nj_ajwk/d6fyv2_s.json init=/tmp/tmp8nj_ajwk/sh420jq2.json output file=/tmp/tmp8nj_ajwk/prophet_modelsihs9kil/prophet_model-20250109155857.csv method=optimize algorithm=newton iter=10000

or

/usr/local/lib/python3.12/site-packages/prophet/stan_model/prophet_model.bin info

I get

Segmentation fault

This is my dockefile:

# Use the official Python base image
FROM python:3.12-slim

# Set the working directory inside the container
WORKDIR /app

# Copy the requirements file to the working directory
COPY requirements.txt .

RUN apt-get update
RUN apt-get --assume-yes install libpq-dev gcc g++ build-essential python3-dev python3-pip python3-venv

# Install the Python dependencies
RUN pip install --upgrade pip
RUN pip install -r requirements.txt

# Copy the application code to the working directory
COPY src/ ./src/
COPY alembic/ ./alembic/
COPY alembic.ini/ .
COPY build.sh .

# Expose the port on which the application will run
EXPOSE 8080

RUN chmod +x build.sh
# Run the FastAPI application using uvicorn server
CMD ["./build.sh"]

I am running this docker container on a M2 Apple silicon mac book air

UPDATE:

I ran very similar code on a Google Cloud Run Function and I get the same issue

[MattiaPeiretti]

Update 2:

I created a docker image where cmdstan is build from source, and by running prophet like the following, it's possible to highjack the actual binaries of cmdstan used:

        model = Prophet(
            growth='flat',
            yearly_seasonality=True,
            weekly_seasonality=False,
            daily_seasonality=False,
            changepoint_range=0.95,
            changepoint_prior_scale=0.025,
            interval_width=0.8,
        )

        cmdstanpy.set_cmdstan_path('/opt/cmdstan-2.34.1/')

This way I could get a more verbose error message:

20:03:11 - cmdstanpy - INFO - Chain [1] start processing
20:03:11 - cmdstanpy - INFO - Chain [1] done processing
20:03:11 - cmdstanpy - ERROR - Chain [1] error: terminated by signal 11 Unknown error -11
Optimization terminated abnormally. Falling back to Newton.
20:03:11 - cmdstanpy - INFO - Chain [1] start processing
20:03:11 - cmdstanpy - INFO - Chain [1] done processing
20:03:11 - cmdstanpy - ERROR - Chain [1] error: terminated by signal 11 Unknown error -11

[MattiaPeiretti]

UPDATE 3:

If I run Prophet on my local machine instead of Docker. It works like a charm.

Maybe an issue with dockerisation?