-
|
Dear all, You have some results for Albanian language. However, in the notebook example Multilingual_ASR.ipynb the langauage code is missing. |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 7 replies
-
|
Those language codes are the ones used by the Fleurs dataset, which did not include Albanian. Those are slightly different from what's used in our tokenizer.py. In Whisper, you can specify |
Beta Was this translation helpful? Give feedback.
-
|
I have created a fine-tuned version for Albanian. https://huggingface.co/spaces/n-iv/wsq If anyone is interested in collaborating or using the model message me. |
Beta Was this translation helpful? Give feedback.
-
|
@marisbasha please can u conact me https://www.linkedin.com/in/florijan-qosja/ |
Beta Was this translation helpful? Give feedback.
-
|
@marisbasha I am quite interested in using your fined-tuned version for Albanian. Is it possible to access it? Thanks |
Beta Was this translation helpful? Give feedback.
-
Beta Was this translation helpful? Give feedback.
-
|
@marisbasha Thanks! Is the finetuning data available? Was interested in a pure transformer model, so I can use it with CTranslate2 :) |
Beta Was this translation helpful? Give feedback.
Those language codes are the ones used by the Fleurs dataset, which did not include Albanian. Those are slightly different from what's used in our tokenizer.py.
In Whisper, you can specify
--language Albanianor--language sqto tell the model to transcribe Albanian speech, but we unfortunately don't have quantitative metrics on Whisper's performance in Albanian. I'd be curious to know if you can share your experience.