No such file or directory error with FMU block

I created FMU file with the below code by Python then loaded the file with FMU block but I got

Error with specified FMU: Cannot load dynamic library 
'/MATLAB Drive/slprj/_fmu/b37a07ca0ba765ad86fa700f844987e2/MachineLearningFMU/binaries/x86_64-linux/ Cannot load MachineLearningFMU.so'
: libpython3.12.so.1.0: cannot open shared object file: No such file or directory

error when I did simulation.

Why do I get this error?

The python code is to load regression_model.pkl file that is a simple machine learning model to predict 2 outputs based on 7 inputs and them save it as .fmu file.

7 inputs (Features):

'accel'

'brake'

'hvac'

'dcdc'

'motor'

'soc'

'vl'

2 outputs (Objective Variables):

'torque'

're_brake'

And I execute this command and create FMU file.

pythonfmu3 build -f mlmodelfmu.py

I wonder if I'm missing something in my python code.

I'd appreciate it you could give me your advice.

Here is my python script (mlmodelfmu.py) to create FMU file.

from pythonfmu3 import Fmi3Causality, Fmi3Slave, Float64
import joblib
import numpy as np


class MachineLearningFMU(Fmi3Slave):
    def __init__(self, **kwargs):
        super().__init__(**kwargs)
        self.author = "AI Assistant"
        self.description = "FMU for regression prediction using pkl model"

        self.model = joblib.load("regression_model.pkl")

        self.inputs = [0.0] * 7

        self.outputs = [0.0] * 2

        input_names = ["accel", "brake", "hvac", "dcdc", "motor", "soc", "vl"]
        for i, name in enumerate(input_names):
            self.register_variable(
                Float64(
                    name,
                    causality=Fmi3Causality.input,
                    getter=lambda i=i: self.inputs[i],
                    setter=lambda v, i=i: self.set_input(i, v),
                )
            )

        output_names = ["torque", "re_brake"]
        for i, name in enumerate(output_names):
            self.register_variable(
                Float64(
                    name,
                    causality=Fmi3Causality.output,
                    getter=lambda i=i: self.outputs[i],
                )
            )

    def set_input(self, index, value):
        self.inputs[index] = value

    def predict_with_model(self):

        input_array = np.array([self.inputs])

        prediction = self.model.predict(input_array)

        for i in range(len(self.outputs)):
            self.outputs[i] = prediction[0][i]

    def do_step(self, current_time, step_size):

        self.predict_with_model()
        return True

Also,this is how I created .pkl file. in the environment of Python 3.10 on local.

n_samples = 1000
X = np.random.rand(n_samples, 7)  
y = np.random.rand(n_samples, 2) 

input_names = ['accel', 'brake', 'hvac', 'dcdc', 
               'motor', 'soc', 'vl']

output_names = ['torque', 're_brake']

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

rf = RandomForestRegressor(n_estimators=100, random_state=42)
model = MultiOutputRegressor(rf)

model.fit(X_train, y_train)

y_pred = model.predict(X_test)

joblib.dump(model, 'regression_model.pkl')

Answer 1

[Venkat Siddarth Reddy]

Hi 翼,

This issue seems to be due to the requirement of Python version "3.12" to execute the FMU. If this version of Python is already installed, I highly recommend reinstalling it and setting up the configuration again.

Additionally,once the Python setup is completed, please generate a new PKL file in "Python 3.12" version.

However if the issue persists even after the above steps, please check the compliance of the FMU with the FMI standard using the "FMU compliance checker". This should reveal more information about the issue.

For more information on "FMU compliance checker", please refer to the following repository:

https://github.com/modelica-tools/FMUComplianceChecker/

Additionally, please refer to the following MATLAB Answers Community post on how to troubleshoot "FMU Import Issues" in Simulink:

https://www.mathworks.com/matlabcentral/answers/1850848-how-do-i-troubleshoot-fmu-import-issues-in-simulink

I hope it helps!

[翼]

Hi @Venkat Siddarth Reddy,

Thank you for telling me.

I generated a new PKL file in "Python 3.12" and executed my python script (mlmodelfmu.py) to create FMU file again. However it didn't work. The same error happend.

("Python 3.12" here means a virtual environment by "pipenv")

I enabled FMU Debug Logging and checked the log file (.txt) but only this is written...

# Debug Logging enabled for block 'fmu_test/FMU'

I wonder if I'm now using MATLAB Online....

Also, I installed FMpy and checked my fmu with the below code.

from fmpy import *
fmu = 'MachineLearningFMU.fmu'
dump(fmu) 

The result says,

Model Info

  FMI Version        3.0
  FMI Type           Co-Simulation
  Model Name         MachineLearningFMU
  Description        FMU for regression prediction using pkl model
  Platforms          linux64, win64
  Continuous States  0
  Event Indicators   0
  Variables          9
  Generation Tool    PythonFMU3 0.1.15
  Generation Date    2024-10-13T08:57:39+00:00


Variables (input, output)

  Name               Causality              Start Value  Unit     Description
  accel              input                            0           
  brake              input                            0           
  hvac               input                            0           
  dcdc               input                            0           
  motor              input                            0           
  soc                input                            0           
  vl                 input                            0           
  torque             output                                       
  re_brake           output  

thenI tried its simulation with this code, but I got the error.

  result = simulate_fmu('MachineLearningFMU.fmu')

Here is the error.

  Failed to validate modelDescription.xml:
  Exactly one independent variable must be defined.

I could not figure out the cause...

Do you have any good ideas?

Best,

[翼]

@Venkat Siddarth Reddy

I solved it! Sorry to bother you..

Thank you.