wavlm-base-ug-combined-2de

This model is a fine-tuned version of microsoft/wavlm-base on the AJIKADEV/UGANDAN-ENGLISH - NA dataset. It achieves the following results on the evaluation set:

• Loss: 0.1440
• Wer: 0.1389

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: 0.0003
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000.0
• training_steps: 30000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2685	2.4117	3000	0.2399	0.2679
0.1763	4.8235	6000	0.1947	0.2208
0.1175	7.2348	9000	0.1777	0.2065
0.0892	9.6466	12000	0.1667	0.1852
0.0686	12.0579	15000	0.1585	0.1739
0.0479	14.4696	18000	0.1461	0.1634
0.039	16.8814	21000	0.1453	0.1538
0.0284	19.2927	24000	0.1471	0.1473
0.021	21.7045	27000	0.1443	0.1417
0.0163	24.1158	30000	0.1440	0.1388

Framework versions

• Transformers 5.0.0.dev0
• Pytorch 2.9.1+cu128
• Datasets 3.6.0
• Tokenizers 0.22.1