ModernBERT Multiclass Disfluency Detection
This model is fine-tuned from answerdotai/ModernBERT-base for multi-class disfluency detection in spoken language.
Training Hyperparameters
The following hyperparameters were used during training:
- Learning rate: 5e-05
- Batch size: 16
- Number of epochs: 15
- Optimizer: OptimizerNames.ADAMW_8BIT
- LR scheduler type: SchedulerType.COSINE
- Warmup ratio: 0.15
- Downloads last month
- 11
Evaluation results
- Accuracy on Disfluency Datasettest set self-reported0.941
- F1 on Disfluency Datasettest set self-reported0.776