RikeshSilwal/whisper-small-hi-transfer-ne

Whisper small nepali - Rikesh Silwal

This model is a fine-tuned version of openai/whisper-small on the slr43 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3583
• Wer: 33.7199

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: 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: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0061	9.62	1000	0.3096	36.4853
0.0001	19.23	2000	0.3306	34.2551
0.0	28.85	3000	0.3525	33.5712
0.0	38.46	4000	0.3583	33.7199

Framework versions

• Transformers 4.38.2
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2