shripadbhat/whisper-tiny-hi-1000steps

Whisper tiny Hindi

This model is a fine-tuned version of openai/whisper-tiny on the Common Voice 11.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.5538
• Wer: 41.5453

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
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• training_steps: 1000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.7718	0.73	100	0.8130	55.6890
0.5169	1.47	200	0.6515	48.2517
0.3986	2.21	300	0.6001	44.9931
0.3824	2.94	400	0.5720	43.5171
0.3328	3.67	500	0.5632	42.5112
0.2919	4.41	600	0.5594	42.7863
0.2654	5.15	700	0.5552	41.6428
0.2618	5.88	800	0.5530	41.8893
0.2442	6.62	900	0.5539	41.5740
0.238	7.35	1000	0.5538	41.5453

Framework versions

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