irish-whisper-MCV_augmentation

This model is a fine-tuned version of openai/whisper-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 4.9504
• Wer: 81.7722

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: 0.0001
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• 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: 1000
• training_steps: 60000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0217	161.2927	5000	3.2474	96.2025
0.0153	322.5854	10000	3.8866	83.7975
0.0136	483.8780	15000	4.1084	78.4810
0.0048	645.1626	20000	4.5027	86.3291
0.0058	806.4553	25000	4.7173	83.2911
0.0094	967.7480	30000	4.9078	83.2911
0.0060	1129.0325	35000	4.7303	85.8228
0.0067	1290.3252	40000	4.7119	85.3165
0.0022	1451.6179	45000	4.7075	86.3291
0.0001	1612.9106	50000	5.0320	86.3291
0.0009	1774.1951	55000	4.9213	81.0127
0.0020	1935.4878	60000	4.9504	81.7722

Framework versions

• Transformers 5.3.0.dev0
• Pytorch 2.9.0+cu126
• Datasets 4.0.0
• Tokenizers 0.22.2