iulik-pisik/vreme_model_small

Whisper Small Romanian - Vreme

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

• Loss: 0.1873
• Wer: 7.9942

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• 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
0.0141	6.02	1000	0.1417	9.2067
0.0007	12.05	2000	0.1657	8.4464
0.0002	18.07	3000	0.1833	8.0353
0.0001	24.1	4000	0.1873	7.9942

Framework versions

• Transformers 4.38.2
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2