[AutoWEKA] Error: Could not find or load main class weka.classifiers.meta.AutoWEKAClassifier

[learsi1911]

Hi, I am trying to run some experiments in AutoWEKA but I got the above error. I think the problem is that weka.jar is no longer included in the autoweka archive. From my setup.sh I haven't changed anything, I'm using a "stable" version and WEKA is downloaded from: https://prdownloads.sourceforge.net/weka/weka-3-8-5-azul-zulu-linux.zip

[PGijsbers]

Can you provide the exact command and the branch you used (or commit hash, if not using the head)?

[learsi1911]

Yes, this is the command I used: python runbenchmark.py AutoWEKA openml/t/2073 pygmo_c -s force pygmo_c is a constrain with 5 folds and 1 hour time budget The branch is master

[cynthiamaia]

Hi,

I have the same problem, when running autoweka, it gives this error. CalledProcessError: Command 'java -cp /home/cynthia/automlmaster/project-automl/automlbenchmark/frameworks/AutoWEKA/lib/autoweka/autoweka.jar weka.classifiers .meta.AutoWEKAClassifier -t "/home/cynthi…, did anyone manage to solve?

[PGijsbers]

Can you describe to me the steps to reproduce this problem? And produce the full installation log when trying to install AutoWEKA? I installed from scratch and could not reproduce the error, specifically I performed the following steps and it worked correctly:

git clone https://github.com/openml/automlbenchmark.git
cd automlbenchmark
python -m venv venv
source venv/bin/activate
python -m pip install -r requirements.txt
python runbenchmark.py autoweka -t iris -f 0 -m docker -s force

It successfully installed AutoWEKA and its dependency (a pr last summer was supposed to fix this, is it possible you are using old code?).

[alanwilter]

Indeed, autoweka work with iris, example, but when trying with my own dataset, all setup is done fine but it eventually fails during execution with:

...
**** AutoWEKA [vlatest]****

Using 37406MB memory per run on 4 parallel runs.
Running cmd `java -cp /bench/frameworks/AutoWEKA/lib/autoweka/autoweka.jar:/bench/frameworks/AutoWEKA/lib/weka/weka.jar weka.classifiers.meta.AutoWEKAClassifier -t "/input/test_data/differentiate_cancer_train.csv" -T "/input/test_data/differentiate_cancer_test.csv" -memLimit 37406 -classifications "weka.classifiers.evaluation.output.prediction.CSV -distribution -file \"/output/predictions/teddata/0/predictions.weka_pred.csv\"" -timeLimit 6 -parallelRuns 4 -metric areaUnderROC -seed 24977`
2022-09-21 09:08:03.676 [Thread-4] WARN AutoWEKAClassifier - 2022-09-21 09:08:03.676 [main] WARN Experiment - [WARN ] [PROCESS-ERR]  2022-09-21 09:08:03.643 [main] ERROR Wrapper - java.lang.NoClassDefFoundError: no/uib/cipr/matrix/Vector
2022-09-21 09:08:03.676 [Thread-4] WARN AutoWEKAClassifier - 2022-09-21 09:08:03.676 [main] WARN Experiment - [WARN ] [PROCESS-ERR]  java.lang.RuntimeException: java.lang.NoClassDefFoundError: no/uib/cipr/matrix/Vector
2022-09-21 09:08:03.677 [Thread-4] WARN AutoWEKAClassifier - 2022-09-21 09:08:03.676 [main] WARN Experiment - [WARN ] [PROCESS-ERR]     at autoweka.WorkerThread.run(WorkerThread.java:34)
2022-09-21 09:08:03.677 [Thread-4] WARN AutoWEKAClassifier - 2022-09-21 09:08:03.677 [main] WARN Experiment - [WARN ] [PROCESS-ERR]  Caused by: java.lang.NoClassDefFoundError: no/uib/cipr/matrix/Vector
2022-09-21 09:08:03.677 [Thread-4] WARN AutoWEKAClassifier - 2022-09-21 09:08:03.677 [main] WARN Experiment - [WARN ] [PROCESS-ERR]     at java.lang.Class.forName0(Native Method)
...

However, given my recent experience with other frameworks I suspect it suffers from the same issue I reported in #498, i.e., my input data is in CSV when it should be ARFF.

[alanwilter]

Good news, with help of https://github.com/anaavila/convert-csv-to-arff and tweaking my target to @attribute 'target' {0,1} in my new ARFF file and now AutoWeka worked fine for my dataset.

I would close this ticket. But as a suggestion, please add some info about ARFF and the input data in the README, FAQ etc. There's not a single world about it. I'm gonna test now the other frameworks that I'm sure need ARFF just to confirm my suspicions.