arun100/whisper-base-hi-2

Whisper Base Hindi

This model is a fine-tuned version of openai/whisper-base on the mozilla-foundation/common_voice_16_0 hi dataset. It achieves the following results on the evaluation set:

• Loss: 0.4679
• Wer: 28.6490

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-06
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• 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
0.6425	6.01	500	0.7025	41.4477
0.3973	13.0	1000	0.5367	33.9692
0.3125	19.01	1500	0.4927	31.4458
0.2848	26.0	2000	0.4739	30.1037
0.2201	32.01	2500	0.4675	29.4859
0.2257	39.01	3000	0.4637	28.9933
0.1837	46.0	3500	0.4657	28.9140
0.1897	52.01	4000	0.4658	28.7450
0.1764	59.0	4500	0.4676	28.7178
0.1681	65.01	5000	0.4679	28.6490

Framework versions

• Transformers 4.37.0.dev0
• Pytorch 2.1.2+cu121
• Datasets 2.16.2.dev0
• Tokenizers 0.15.0