arkanalexei/whisper-tiny-sundanese-pretrained2

BisiX: Sundanese Whisper (Reset Params)

This model is a fine-tuned version of openai/whisper-tiny.en on the SU ID ASR dataset. It achieves the following results on the evaluation set:

• Loss: 10.7367
• Wer: 100.0
• Cer: 100.0

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: 64
• eval_batch_size: 64
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• training_steps: 1000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
10.8526	1.1765	100	10.8491	100.9708	81.1835
10.8392	2.3529	200	10.8336	1808.6292	800.3308
10.8194	3.5294	300	10.8165	2677.0607	1196.5347
10.8034	4.7059	400	10.7992	207.8742	83.6906
10.7876	5.8824	500	10.7815	148.4494	78.2801
10.7684	7.0588	600	10.7662	110.9303	77.7046
10.7567	8.2353	700	10.7537	93.9326	75.0677
10.7473	9.4118	800	10.7442	100.0	100.0
10.7389	10.5882	900	10.7386	100.0	100.0
10.7366	11.7647	1000	10.7367	100.0	100.0

Framework versions

• Transformers 4.45.1
• Pytorch 2.4.1+cu124
• Datasets 3.0.1
• Tokenizers 0.20.0