telugu_wav2vec_optimizer_ablation_adamw

This model is a fine-tuned version of nik1509/telugu_wav2vecmedium_basemodel on the None dataset. It achieves the following results on the evaluation set:

• Loss: 579.8488
• Wer: 0.3969
• Cer: 0.1684

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
278.5511	1.0	1152	462.5067	0.4090	0.1689
272.0582	2.0	2304	467.9001	0.4053	0.1674
290.6093	3.0	3456	479.2049	0.4016	0.1665
281.503	4.0	4608	483.6781	0.4034	0.1661
281.271	5.0	5760	490.5450	0.4014	0.1659
291.9046	6.0	6912	508.9962	0.3929	0.1650
268.282	7.0	8064	510.8926	0.3975	0.1664
264.0682	8.0	9216	521.1376	0.3964	0.1667
259.9121	9.0	10368	518.7831	0.3947	0.1669
248.439	10.0	11520	535.6939	0.4004	0.1682
228.4703	11.0	12672	546.0759	0.3980	0.1677
273.53	12.0	13824	565.0540	0.3971	0.1672
218.3814	13.0	14976	567.6315	0.3970	0.1680
235.1851	14.0	16128	569.3064	0.3971	0.1686
241.3145	15.0	17280	579.8488	0.3969	0.1684

Framework versions

• Transformers 4.53.0
• Pytorch 2.8.0.dev20250609+cu118
• Datasets 3.6.0
• Tokenizers 0.21.2