sjs40031316/whisper-small-hi

Whisper Small Hi - i hope you work

This model is a fine-tuned version of openai/whisper-small on the kspon speech dataset. It achieves the following results on the evaluation set:

• Loss: 1.2062
• Wer: 51.2535
• Cer: 29.8633

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: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.0062	17.5439	1000	0.9555	50.6964	30.3190
0.0013	35.0877	2000	1.0772	52.3677	31.2303
0.0002	52.6316	3000	1.1549	71.9591	38.1002
0.0002	70.1754	4000	1.1925	71.2163	37.4343
0.0001	87.7193	5000	1.2062	51.2535	29.8633

Framework versions

• Transformers 4.44.2
• Pytorch 2.8.0+cu126
• Datasets 2.20.0
• Tokenizers 0.19.1