could you please give more information about this model?
#1
by
Toen
- opened
Could you please provide more information about this model? I have no clue about it. For instance:
How is it trained?
What data does the training rely on?
How to use it?
I'll update the model cards soonish,
however in the mean time:
It is based on the Whisper-small model, it was trained 15 Epochs on (warmup, linear decay)
Its was trained on 70% of an modified/filtered version of https://huggingface.co/datasets/laion/synthetic_vocal_bursts
Batch Size: 50 * 8 (GPUs)
LR: 0.00001
Evaluation Table is also coming soon.