arun100/whisper-base-uk-2

Whisper Base Ukrainian

This model is a fine-tuned version of arun100/whisper-base-uk-1 on the google/fleurs uk_ua dataset. It achieves the following results on the evaluation set:

• Loss: 0.4710
• Wer: 33.5630

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: 2.5e-07
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• 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
0.2683	95.0	1000	0.4710	33.5630
0.142	190.0	2000	0.4714	33.8344
0.0871	285.0	3000	0.4782	33.9596
0.0656	380.0	4000	0.4830	33.7230
0.0595	476.0	5000	0.4847	33.7161

Framework versions

• Transformers 4.37.0.dev0
• Pytorch 2.1.2+cu121
• Datasets 2.16.2.dev0
• Tokenizers 0.15.0