whisper-tiny-hindi3_test

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

• Loss: 0.4252
• Wer: 52.6981

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: 3.75e-05
• train_batch_size: 16
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant
• lr_scheduler_warmup_steps: 30
• training_steps: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.1072	0.2	20	0.8412	86.0034
0.6316	0.4	40	0.5764	66.9477
0.5548	0.6	60	0.5071	63.4064
0.5014	0.8	80	0.4606	55.3120
0.459	1.0	100	0.4252	52.6981

Framework versions

• Transformers 4.41.1
• Pytorch 2.3.1+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1