URGENT: Software registration fails with HF_HUB_CACHE error

PAN
1

Dear TIRA Team / Organizers,

I am currently participating in the PAN CLEF 2026 Generative AI Authorship Verification task. With the May 7 deadline approaching extremely fast, I am running into a critical error when trying to submit my software via tira-run.

To avoid uploading an oversized Docker image, I am using the --mount-hf-model feature. During the execution, the process proceeds to the final registration phase. However, at the very last Upload TIRA_SOFTWARE step, the process crashes with the following error:

Upload of software failed with error {'status': 1, 'message': "Model Qwen/Qwen2.5-1.5B does not in HF_HUB_CACHE = '/mnt/ceph/tira/data/publicly-shared-datasets/huggingface/hub'. Expected a directory '/mnt/ceph/tira/data/publicly-shared-datasets/huggingface/hub/models--Qwen--Qwen2.5-1.5B' to exist. Please ensure that the model is available via your preferred way to download it."}

Could you please look into this error? If this indicates that the model needs to be available in the TIRA shared Ceph storage for mounting, I would be extremely grateful if you could assist in making it available. If there are any other steps I should take on my end to resolve this, please let me know.

Given the upcoming deadline, your immediate support would be a lifesaver. Thank you for your time and assistance!

Best regards,

Junlong Yang

2

Dear Junlong,

Thanks for reaching out and for participating in PAN, we are really happy about this!

We extend the deadline for submissions by two weeks, we anounced this in the Discord chat now and will update the web page soon.

The `Qwen/Qwen2.5-1.5B` model is not available at the moment in the TIRA backend, I will download it and update you once it is there, then you can mount it.

Best regards,

Maik

3

Dear Junlong,

Sorry that it took a bit, the `Qwen/Qwen2.5-1.5B` is now available in TIRA, so you should now be able to submit your approach.

Could you try it?

Best regards,

Maik

4

Dear Maik,

Thank you very much for your prompt help and for making the model available.

I have just retried the submission, and it was successful! Everything is working perfectly now.

Thanks again for your support.

Best regards,

Junlong Yang

5

Dear Junlong,

Cool, thanks for the confirmation!

Best regards,

Maik