rishabhjain16/whisper_l2_to_kaggle_sq

Whisper large V2 to Kaggle Albanian

This model is a fine-tuned version of openai/whisper-large-v2 on the rishabhjain16/kaggle_albanian default dataset. It achieves the following results on the evaluation set:

• Loss: 0.3393
• Wer: 25.1606

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: 16
• 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: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.3422	0.1	500	0.5929	44.6536
0.4805	0.2	1000	0.4738	37.5890
0.3886	0.3	1500	0.4295	34.2173
0.3929	0.4	2000	0.3847	30.9552
0.2195	1.09	2500	0.3645	28.2136
0.1844	1.19	3000	0.3627	27.7056
0.1873	1.29	3500	0.3520	27.1403
0.1695	1.39	4000	0.3335	26.1816
0.089	2.08	4500	0.3623	26.1094
0.072	2.18	5000	0.3393	25.1606

Framework versions

• Transformers 4.37.2
• Pytorch 1.14.0a0+44dac51
• Datasets 2.17.1
• Tokenizers 0.15.2