whisper-medium-nyagen-combined-model

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

• Loss: 0.2561
• Wer: 0.1952

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Use 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
1.1602	0.5326	200	0.5182	0.3747
0.5456	1.0639	400	0.3445	0.2552
0.5516	1.5965	600	0.2903	0.2413
0.224	2.1278	800	0.2817	0.2384
0.2413	2.6605	1000	0.2561	0.1952
0.1036	3.1917	1200	0.2583	0.1904
0.1135	3.7244	1400	0.2637	0.2120
0.057	4.2557	1600	0.2731	0.2097

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0