Evaluation is stuck at running status

[dongfang91]

Hello,

I was testing my codalab shared task, and the evaluation was stuck at the running status.

Here is my compettion link:

https://codalab.lisn.upsaclay.fr/competitions/22552?secret_key=c81b0267-2116-4211-b948-c282889b194a

[Didayolo]

Hello,

Could you please share more details?

• Are you using the default queue, or your own queue of compute workers?
• Did it work in the past?
• Is it results or code submission?
• Are the submissions supposed to be quick to compute? Do you have a sound execution time limit?

[dongfang91]

Are you using the default queue, or your own queue of compute workers?

• I am using the default queue.

Did it work in the past?

• I just started this task. I can see it saved the score output, and the logging file, but did not see any score at the codalab results page.

Is it results or code submission?

• It's the result submission.

Are the submissions supposed to be quick to compute? Do you have a sound execution time limit?

• It is very quick to run, took less than 1 second in the local environment.

[Didayolo]

Hi @dongfang91,

To investigate the problem, I dumped your competition and uploaded it on Codabench.

For the upload to work, I just had to rename labels of phases and leaderboard that were too long (e.g. Track2 Sentence Segmention, results into Track2 Results).

Then when I try to upload the submission, I get the following error:

2025-04-07 11:22:40,681 root         INFO     -------------------------------------------
2025-04-07 11:22:40,681 root         INFO     Input directory: /app/input
2025-04-07 11:22:40,682 root         INFO     Output directory: /app/output
2025-04-07 11:22:40,682 root         ERROR    More than one valid files found in archive.                   
Make sure only one valid file is available.

This log may give you more information about your problem. Also, it is strongly suggested to work on new projects directly on Codabench.

[dongfang91]

Thank you. I changed to the codabench and the evaluation scripts. Now it works.