arjunshajitech/whisper-small-malayalam-v4

Whisper Small Malayalam - Arjun Shaji

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

• Loss: 0.1677
• Wer: 52.3397

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: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0239	5.1020	1000	0.1136	54.3847
0.002	10.2041	2000	0.1426	52.9827
0.0003	15.3061	3000	0.1584	52.5808
0.0001	20.4082	4000	0.1643	52.3129
0.0001	25.5102	5000	0.1677	52.3397

Framework versions

• Transformers 4.41.0
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1