Headless mode flags not recognized (“Unable to read input file -model / -experiment / -output”)

[soltanisgeo]

Hello Community,
Maybe a silly question but I am trying to run GAMA in headless mode on a Linux HPC cluster I downloaded the package: GAMA_1.9.3_Linux_with_JDK.zip
However, when I launch GAMA using the usual headless arguments:

#!/bin/bash
#$ -l h_rt=4:00:00
#$ -l mem=8G
#$ -cwd
#$ -j y
#$ -N gama_tcem

export JAVA_HOME=/…./software/gama_headless/jdk
export PATH=$JAVA_HOME/bin:$PATH

GAMA=/…./software/gama_headless
LAUNCHER=$GAMA/plugins/org.eclipse.equinox.launcher_1.6.500.v20230717-2134.jar
WORKSPACE=/…/GAMA_workspace
MODEL=/…../GAMA_workspace/CityEscape/models/TCEM.gaml
EXPERIMENT=RunHeadless
OUTPUT=/…../GAMA_output/outputCityEscape_output.csv

$JAVA_HOME/bin/java
-Xms512m -Xmx6G
-Dgama.headless=true
-DWORKSPACE=$WORKSPACE
-classpath $LAUNCHER
org.eclipse.equinox.launcher.Main
-application msi.gama.headless.product
-model $MODEL
-experiment $EXPERIMENT
-output_file $OUTPUT

GAMA produces errors such as:

Unable to read input file -model. File not exist!
Unable to read input file -experiment. File not exist!
Unable to read input file -output. File not exist!

It seems the launcher is treating headless flags as filenames and seems the distribution may not include the correct headless argument parser. This occurs even with a minimal test model.

Could you please confirm which download provides the proper Linux headless build?
Is the “Linux_with_JDK” package expected to support headless mode, or is a separate gama-headless package required for 1.9.3/2025?

Thank you!

[lesquoyb]

The problem is not in the release but in the way you call gama headless, those parameters just do not exist.
I recommend that you read the documentation on headless batch here to get the proper syntax.
You can also get more recent releases of gama as I think some bugs in headless batch have been solved since 1.9.3

[soltanisgeo]

Hi again Baptiste, and thank you for your answer.
I checked again the headless batch doc and reran my tests with GAMA 1.9.3 headless on Linux. This time it starts but when I started, the execution consistently reaches to this part:

GAMA : Loading constants done
GAML : Plugins with language additions loaded
GAMA configuring and loading...
GAML : Artifacts built
GAMA couldn't compile your input file.

At that point, the following happens:
A .workspaceX directory is created.
.workspaceX/.metadata is created.
.workspaceX/.metadata/.plugins exists.
but, .workspaceX/.metadata/.log is never created.

The launcher then prints the error:
cat: .workspaceX/.metadata/.log: No such file or directory

You can check this on "headless_run.log" attached
headless_run.log
I have also confirmed directly from the filesystem that there is no .log file anywhere under .metadata (no hidden file, no *.log).
I can also confirm that the paths are correct, since the same model, with the same folder structure and file references, runs correctly in the Windows GUI.

What makes this confusing is that the model runs in the GUI, headless clearly starts correctly (plugins load, etc.), but the failure happens before any .metadata/.log is written. Without a log file, it is impossible to see why the headless compiler is rejecting the model.

So my questions are:
Is there a way to force GAMA headless to print GAML compilation errors when -batch fails?
If .metadata/.log is not created, where should headless users look for diagnostic output when?
Are there cases you know where a model that runs in the GUI will fail to compile in headless, and if so, is there a recommended workflow to validate a model specifically for headless execution?

Best
Saeed

p.s: Headless bash ..
gama-headless.txt

[RoiArthurB]

Hi @soltanisgeo , the missing log file is not very important and is just that the headless script in 1.9.3 didn't check if the file exists before trying to print it.

The important error you're having is GAMA couldn't compile your input file. which might indicate that you're not feeding the parameters in the good order and we can't give you any support if you're not giving us the exact command you're trying...