License clarification for diarizen-meeting-base: MIT metadata vs CC BY-NC 4.0 in the GitHub README
Hello, and thank you for releasing this work.
I would like to clarify which license applies to the weights in this specific repository, because two sources disagree.
The model card metadata of BUT-FIT/diarizen-meeting-base declares license: mit. However, the card also asks users to follow the instructions in the BUTSpeechFIT/DiariZen GitHub repository, whose README and MODEL_LICENSE state that the pre-trained model weights are released under CC BY-NC 4.0.
The stated rationale for the non-commercial restriction is the presence of RAMC, MSDWild and DIHARD-3 in the training mixture. This model card lists only AMI, AISHELL-4 and AliMeeting as training data, so that rationale does not appear to apply here.
Could you confirm whether the MIT metadata is correct for this checkpoint, or whether the CC BY-NC 4.0 model license also covers it?
Thank you very much for your time.
@Jyhan003 Apologies for the direct mention — I understand this checkpoint is from an earlier publication and may no longer be actively maintained.
Could you confirm which license applies to the weights in this repository? The model card metadata says MIT, while the GitHub README states CC BY-NC 4.0 for all pre-trained weights. The stated reason for the non-commercial terms is the presence of RAMC, MSDWild and DIHARD-3 in the training mixture, and this card lists only AMI, AISHELL-4 and AliMeeting.
A one-line answer either way would be enough. Thank you.
Hi sorry for the late reply.
And yes, the current model was pretrained using only AMI, AISHELL-4 and AliMeeting. The CC BY-NC 4.0 license only holds for models trained using datasets like RAMC, MSDWild, and DIHARD-3.