irish-whisper-combined-data-no-aug

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: 1.0523
• Wer: 31.9210

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
0.1161	17.6684	5000	0.8361	37.3581
0.0502	35.3360	10000	0.9223	36.8504
0.0434	53.0035	15000	0.9615	37.0073
0.0296	70.6720	20000	0.9900	37.5335
0.0149	88.3395	25000	1.0153	36.3427
0.0109	106.0071	30000	1.0371	36.7211
0.0101	123.6755	35000	1.0194	35.5949
0.0001	141.3431	40000	0.9988	34.5888
0.0000	159.0106	45000	1.0226	34.1734
0.0000	176.6790	50000	1.0323	33.1026
0.0000	194.3466	55000	1.0440	32.0318
0.0000	212.0141	60000	1.0523	31.9210

Framework versions

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