devkyle/Akan-tiny-2000ms-1k

whisper-tiny-akan

This model is a fine-tuned version of openai/whisper-tiny on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2315
• Wer: 13.2075

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: 0.0001
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.7229	5.0	250	0.8173	56.7933
0.2324	10.0	500	0.8472	50.4318
0.0673	15.0	750	0.9745	49.5250
0.0297	20.0	1000	1.0408	49.4675
0.0245	25.0	1250	0.2170	13.1495
0.0088	30.0	1500	0.2295	13.7591
0.0025	35.0	1750	0.2324	13.9478
0.0019	40.0	2000	0.2315	13.2075

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.0+cu121
• Datasets 3.0.0
• Tokenizers 0.19.1