ModelEcho

This model is a fine-tuned version of gigant/whisper-medium-romanian on the arrow dataset. It achieves the following results on the evaluation set:

• Loss: 0.1863
• Wer: 10.5222

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: 2e-05
• train_batch_size: 16
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• 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: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.3049	0.2453	500	0.2264	12.3599
0.2764	0.4907	1000	0.2071	11.3205
0.2792	0.7360	1500	0.1992	11.0228
0.2504	0.9814	2000	0.1945	10.7824
0.2424	1.2267	2500	0.1918	10.6759
0.249	1.4720	3000	0.1896	10.6031
0.244	1.7174	3500	0.1880	10.5701
0.2304	1.9627	4000	0.1868	10.4709
0.2326	2.2080	4500	0.1866	10.5371
0.2204	2.4534	5000	0.1863	10.5222

Framework versions

• PEFT 0.16.0
• Transformers 4.53.2
• Pytorch 2.7.1+cu118
• Datasets 4.0.0
• Tokenizers 0.21.2