predika-org/ayira-large-v2-2

Ayira Large V2.2

This model is a fine-tuned version of predika-org/whisper-large-v2-ht on the Haitian Creole ASR Dataset dataset. It achieves the following results on the evaluation set:

• Loss: 0.2141
• Wer: 5.4775

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: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH 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: 500
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0162	2.2676	1000	0.1964	6.5939
0.0062	4.5351	2000	0.1997	5.6896
0.0008	6.8027	3000	0.2064	5.6514
0.0003	9.0703	4000	0.2141	5.4775

Framework versions

• Transformers 4.52.4
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1