piyushmaharana/outcomes-whisper-tiny-v1.1

OutcomesAI-Whisper-tiny-v1.2

This model is a fine-tuned version of piyushmaharana/outcomes-whisper-tiny-v1 on the big-transcript-pronounce dataset. It achieves the following results on the evaluation set:

• Loss: 0.0602
• Wer: 2.8200

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
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0401	12.5	100	0.0998	5.4230
0.0006	25.0	200	0.0777	6.5076
0.0003	37.5	300	0.0723	4.1215
0.0002	50.0	400	0.0691	3.4707
0.0001	62.5	500	0.0669	3.2538
0.0001	75.0	600	0.0656	3.0369
0.0001	87.5	700	0.0646	3.2538
0.0001	100.0	800	0.0635	3.4707
0.0001	112.5	900	0.0628	3.4707
0.0001	125.0	1000	0.0624	3.4707
0.0001	137.5	1100	0.0619	3.4707
0.0001	150.0	1200	0.0614	2.8200
0.0001	162.5	1300	0.0613	3.2538
0.0	175.0	1400	0.0609	3.2538
0.0	187.5	1500	0.0607	3.2538
0.0	200.0	1600	0.0606	3.2538
0.0	212.5	1700	0.0604	3.2538
0.0	225.0	1800	0.0605	3.4707
0.0	237.5	1900	0.0603	3.2538
0.0	250.0	2000	0.0602	2.8200

Framework versions

• Transformers 4.44.0
• Pytorch 2.4.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1