PhanithLIM/whisper-tindy-aug-26dec

small-tindy-aug-26dec

This model is a fine-tuned version of openai/whisper-tiny on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2720
• Wer: 82.4497

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
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: constant
• lr_scheduler_warmup_steps: 1000
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.4654	1.0	272	1.2543	113.9215
1.0788	2.0	544	0.8559	105.8815
0.7043	3.0	816	0.5695	97.9140
0.5241	4.0	1088	0.4696	94.4734
0.4376	5.0	1360	0.4099	90.9894
0.3814	6.0	1632	0.3720	90.8953
0.3414	7.0	1904	0.3490	87.5489
0.3127	8.0	2176	0.3282	85.8395
0.29	9.0	2448	0.3132	85.5425
0.2691	10.0	2720	0.3018	84.6516
0.2525	11.0	2992	0.2949	84.6733
0.2365	12.0	3264	0.2878	85.1586
0.2235	13.0	3536	0.2787	82.1744
0.2117	14.0	3808	0.2770	81.7181
0.2004	15.0	4080	0.2720	82.4497

Framework versions

• Transformers 4.45.2
• Pytorch 2.4.0
• Datasets 3.2.0
• Tokenizers 0.20.3