anuragshas/whisper-small-hi

Whisper Small Hindi

This model is a fine-tuned version of openai/whisper-small on the mozilla-foundation/common_voice_11_0 hi dataset. It achieves the following results on the evaluation set:

• Loss: 0.6260
• Wer: 22.4292

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: 7e-06
• train_batch_size: 64
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• training_steps: 3000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0176	7.01	500	0.4165	22.5066
0.0015	14.01	1000	0.5186	22.2573
0.0004	21.02	1500	0.5741	22.2401
0.0002	28.02	2000	0.6025	22.3834
0.0002	36.01	2500	0.6197	22.3977
0.0002	43.01	3000	0.6260	22.4292

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.0+cu117
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2