Certificate Generator Error Despite Completing All 4 Audio Course Assignments
Hello @MariaK ,
My Hugging Face username is Fasih779, and I have successfully completed all four Audio Course hands-on assignments:
Unit 4: Audio Classification ✅
Unit 5: Automatic Speech Recognition ✅
Unit 6: Text-to-Speech ✅
Unit 7: Audio Applications ✅ — my public speech-to-speech translation demo passed the community-fixed Unit 7 assessment.
The community-fixed progress checker verifies Units 4, 5, and 6. I have also attached a screenshot confirming that my Unit 7 demo passed the assessment.
However, when I enter my information in MariaK/Audio-Course-Certification, all three outputs—message, PDF and image—display Error.
It appears that the certificate generator’s private progress-checking service may be unavailable or affected by the deleted Unit 7 dataset. Could you please verify my completed assignments and repair/restart the certificate service, or advise how I can receive my certificate?
Details:
Hugging Face username: Fasih779
Name: Muhammad Fasih Malhi
Unit 7 demo: Fasih779/speech-to-speech-translation
Unit 6 model: Fasih779/speecht5-librispeech-dummy-tts
I have attached:
Screenshot showing Units 4, 5, and 6 as passed.
Screenshot showing my Unit 7 demo passed.
Screenshot showing the certificate generator error.
@MariaK , the official certificate form still returns Error in all three outputs for username Rohith9059 on 25 October 2026. Refreshing the form and resubmitting did not resolve it, and no certificate was issued.
Three required training projects are published with their actual evaluation results and Trainer records:
- Unit 4: https://huggingface.co/Rohith9059/ast-finetuned-gtzan — GTZAN accuracy 0.87 on the fixed seed-42 held-out split.
- Unit 5: https://huggingface.co/Rohith9059/whisper-tiny-minds14-en-us — MINDS-14 en-US normalized WER 0.31582054309327035, using the prescribed 450-example training split.
- Unit 6: https://huggingface.co/Rohith9059/speecht5-finetuned-voxpopuli-nl — SpeechT5 fine-tuned for 1,000 optimizer updates on three hours of Dutch VoxPopuli audio. Generated speech quality remains limited, as documented in its model card.
Requested certificate name: Rohith kumar Dhamagatla.
Could you restore the official checker or review these submissions and advise how to receive the official certificate? Thank you.
Same situation on certificate page.
Username: markkeung
Best regards
Hi @MariaK , I'm seeing the same problem. When I click "Check if I pass and get the certificate" with username MohammadDaleen, all three outputs show Error.
I've passed 3 of 4 hands-on assignments. All three models are public:
- Unit 4: MohammadDaleen/distilhubert-finetuned-gtzan has Accuracy 0.87 on
marsyas/gtzan. - Unit 5: MohammadDaleen/whisper-tiny-finetuned-minds14-en has WER 0.2904 on
PolyAI/minds14. - Unit 6: MohammadDaleen/speecht5_finetuned_english_dialects is tagged
text-to-speech.
A likely cause, in case it helps: app.py was last changed in 2023, when it was written for Gradio 3, but the Space now runs Gradio 5.6.0 (sdk_version was bumped in November 2024). certification() returns output_row.update(visible=visible), and Gradio 4 removed the .update() method, so every request fails at that line. The fix would be to return gr.update(visible=visible) or gr.Row(visible=visible) instead. The gradio_client call to the private checker may also need updating, since newer clients return data directly rather than a file path to pass to open(result).
Thanks!