andrewbawitlung/whisper-large-v3-mizonal2-lus-v25.03.06

Whisper Large v3 Turbo

This model is a fine-tuned version of openai/whisper-large-v3 on the MiZonal v2.0.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2560
• Wer: 14.4917

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
0.36	0.66	500	0.3358	26.2739
0.1887	1.31	1000	0.2487	20.3331
0.1663	1.97	1500	0.2124	17.9302
0.1002	2.62	2000	0.2090	17.4994
0.0573	3.28	2500	0.2081	16.2068
0.0547	3.93	3000	0.2140	16.0080
0.0291	4.59	3500	0.2229	16.1654
0.0134	5.24	4000	0.2326	15.3782
0.0138	5.9	4500	0.2297	15.1131
0.0047	6.55	5000	0.2485	14.9888
0.0018	7.21	5500	0.2529	14.8645
0.0014	7.86	6000	0.2560	14.4917

Framework versions

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