gopikachu/whisper-large-hi

Whisper Medium Hi - Gopika

This model is a fine-tuned version of openai/whisper-medium on the Konnakol - MASC dataset. It achieves the following results on the evaluation set:

• Loss: 1.3999
• Wer Sentence: 0.7709
• Cer Sentence: 0.6099
• Wer Word: 0.8476
• Cer Word: 0.4796
• Wer Char: 0.5988
• Cer Char: 0.4003

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: 5e-05
• train_batch_size: 32
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Sentence	Cer Sentence	Wer Word	Cer Word	Wer Char	Cer Char
0.0078	0.4348	10	1.3529	0.7381	0.6003	0.8273	0.4644	0.5883	0.3815
0.0113	0.8696	20	1.3522	0.7494	0.6057	0.8318	0.4690	0.5933	0.3835
0.0021	1.3043	30	1.3684	0.7675	0.6147	0.8521	0.4797	0.6025	0.3952
0.0067	1.7391	40	1.3809	0.7607	0.6067	0.8488	0.4732	0.5935	0.3878
0.0047	2.1739	50	1.3999	0.7709	0.6099	0.8476	0.4796	0.5988	0.4003

Framework versions

• Transformers 4.41.1
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1