whisper-small-hyper-tuned-v3

This model is a fine-tuned version of Sekiraw/whisper-small-hyper-tuned-v2 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2305
• Wer: 0.3964

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: 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: 1000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1243	0.0330	100	0.2124	0.3780
0.1045	0.0660	200	0.2246	0.3971
0.0936	0.0990	300	0.2496	0.4044
0.0883	0.1320	400	0.2906	0.4809
0.0964	0.1650	500	0.3260	0.4525
0.0974	0.1980	600	0.3168	0.4578
4.0232	0.2309	700	3.4281	1.0
2.0352	0.2639	800	0.3124	0.4796
0.1273	0.2969	900	0.2447	0.4228
0.1489	0.3299	1000	0.2305	0.3964

Framework versions

• Transformers 4.45.1
• Pytorch 2.1.0+cu118
• Datasets 3.0.1
• Tokenizers 0.20.0