whisper-medium-konnakol

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

• Loss: 0.2686
• Wer: 48.1013

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: 1e-05
• train_batch_size: 2
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• training_steps: 250
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0661	40.0	50	0.2011	49.7890
0.0015	80.0	100	0.2589	48.1013
0.0003	120.0	150	0.2683	48.5232
0.0001	160.0	200	0.2667	48.1013
0.0	200.0	250	0.2686	48.1013

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1