metadata

license: mit
base_model: distil-whisper/distil-medium.en
tags:
  - generated_from_trainer
datasets:
  - dysarthria
metrics:
  - accuracy
model-index:
  - name: distil-medium.en-dysarthia-non-dysathira-detection-fm
    results:
      - task:
          name: Audio Classification
          type: audio-classification
        dataset:
          name: Dysarthria
          type: dysarthria
          config: default
          split: train
          args: default
        metrics:
          - name: Accuracy
            type: accuracy
            value: 0.975

distil-medium.en-dysarthia-non-dysathira-detection-fm

This model is a fine-tuned version of distil-whisper/distil-medium.en on the Dysarthria dataset. It achieves the following results on the evaluation set:

Loss: 0.1593
Accuracy: 0.975

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 5e-05
train_batch_size: 2
eval_batch_size: 2
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.1
num_epochs: 10
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.875	1.0	160	0.4243	0.8875
0.0264	2.0	320	1.2160	0.8125
0.0115	3.0	480	0.2111	0.95
0.3632	4.0	640	0.1856	0.975
0.0933	5.0	800	0.7655	0.9
0.0003	6.0	960	0.6221	0.9
0.0001	7.0	1120	0.1163	0.9875
0.0001	8.0	1280	0.3188	0.95
0.0001	9.0	1440	0.1662	0.975
0.0001	10.0	1600	0.1593	0.975

Framework versions

Transformers 4.41.0.dev0
Pytorch 2.2.1+cu121
Datasets 2.19.1
Tokenizers 0.19.1