arun100/whisper-base-tr-1

Whisper Base Turkish

This model is a fine-tuned version of openai/whisper-base on the mozilla-foundation/common_voice_16_0 tr dataset. It achieves the following results on the evaluation set:

• Loss: 0.4280
• Wer: 30.3627

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-06
• 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
0.4162	1.03	500	0.5339	35.4346
0.3776	2.06	1000	0.4788	33.4892
0.3238	4.02	1500	0.4568	31.9497
0.2714	5.06	2000	0.4469	31.4277
0.3232	7.02	2500	0.4386	31.0991
0.2324	8.05	3000	0.4353	30.7406
0.2953	10.01	3500	0.4306	30.6035
0.2878	11.04	4000	0.4292	30.4278
0.3077	13.01	4500	0.4286	30.4155
0.2914	14.04	5000	0.4280	30.3627

Framework versions

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