muneebharoon/whisper-kashmiri-medium

Whisper Medium ks - Muneeb Haroon

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

• Loss: 1.2108
• Wer: 42.3256

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: 32
• 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: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0024	71.43	1000	1.0586	44.0930
0.0002	142.86	2000	1.2108	42.3256
0.0001	214.29	3000	1.3416	42.5581
0.0	285.71	4000	1.4482	42.9302
0.0	357.14	5000	1.5533	42.9302
0.0	428.57	6000	1.6342	43.4884
0.0	500.0	7000	1.7142	43.1628
0.0	571.43	8000	1.7657	43.4419
0.0	642.86	9000	1.8178	43.5349
0.0	714.29	10000	1.8312	43.6744

Framework versions

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