hanasim/breeze-dsw-tiny-id

Breeze DSW Indonesian - tiny

This model is a fine-tuned version of openai/whisper-tiny on the mozilla-foundation/common_voice_16_0 id dataset. It achieves the following results on the evaluation set:

• Loss: 0.7090
• Wer: 43.4447

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: 32
• eval_batch_size: 16
• seed: 42
• distributed_type: multi-GPU
• 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
0.99	0.1	100	0.8486	54.2460
0.7896	1.04	200	0.7578	48.3899
0.4164	1.14	300	0.7388	49.3284
0.5456	2.09	400	0.7178	45.7954
0.4761	3.03	500	0.7109	45.2158
0.2674	3.13	600	0.7007	44.8431
0.3628	4.08	700	0.7026	44.2497
0.2565	5.02	800	0.7085	44.5073
0.2147	5.12	900	0.7090	43.4447
0.28	6.06	1000	0.7134	43.9553

Framework versions

• Transformers 4.37.0.dev0
• Pytorch 2.1.2+cu121
• Datasets 2.16.2.dev0
• Tokenizers 0.15.0