Ranjit/odia_whisper_small_v3.0

Ranjit/whisper_small_35k_or

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

• Loss: 0.2584
• Wer: 13.8614

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.98) and epsilon=1e-06
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 200
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0022	5.59	1000	0.2405	13.9995
0.0005	11.17	2000	0.2584	13.8614
0.0002	16.76	3000	0.2683	16.5598
0.0	22.35	4000	0.2907	15.0380
0.0	27.93	5000	0.3085	14.2035