meg51/whisper-medium-hi

Whisper Medium Hi -megha sharma

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

• Loss: 0.3496
• Wer: 18.2839

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: 8
• 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: 500
• training_steps: 9000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0311	3.3898	1000	0.2263	20.5584
0.0109	6.7797	2000	0.2669	20.4412
0.0026	10.1695	3000	0.2998	19.6408
0.0018	13.5593	4000	0.2947	18.8208
0.0002	16.9492	5000	0.3169	18.8012
0.0	20.3390	6000	0.3288	18.2058
0.0	23.7288	7000	0.3401	18.1667
0.0	27.1186	8000	0.3469	18.2741
0.0	30.5085	9000	0.3496	18.2839

Framework versions

• Transformers 4.43.3
• Pytorch 2.4.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1