andrewbawitlung/whisper-medium-mizonal2-lus-v25.03.06

Whisper Medium

This model is a fine-tuned version of openai/whisper-medium on the MiZonal v2.0.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2789
• Wer: 15.3451

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: 500
• training_steps: 6000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4096	0.66	500	0.3767	28.0139
0.1929	1.31	1000	0.2698	20.9877
0.1757	1.97	1500	0.2310	19.7199
0.089	2.62	2000	0.2347	19.1234
0.0405	3.28	2500	0.2445	18.8334
0.0423	3.93	3000	0.2450	17.3585
0.0188	4.59	3500	0.2535	18.1622
0.0077	5.24	4000	0.2613	16.8862
0.0074	5.9	4500	0.2661	16.1240
0.0027	6.55	5000	0.2710	15.3865
0.001	7.21	5500	0.2763	15.3948
0.001	7.86	6000	0.2789	15.3451

Framework versions

• Transformers 4.37.2
• Pytorch 2.3.1+cu121
• Datasets 2.16.1
• Tokenizers 0.15.1