rishabhjain16/whisper_l2_to_cv_sq

Whisper large V2 to CV Albanian

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

• Loss: 0.7918
• Wer: 34.6232

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
0.0515	9.0	500	0.6733	42.4847
0.0101	18.01	1000	0.6810	37.5967
0.0074	27.01	1500	0.7185	38.0855
0.0009	37.0	2000	0.6987	35.5193
0.0002	46.0	2500	0.7393	35.0305
0.0001	55.01	3000	0.7603	35.0305
0.0001	64.01	3500	0.7739	34.8676
0.0001	74.0	4000	0.7832	34.8269
0.0001	83.0	4500	0.7895	34.9084
0.0001	92.01	5000	0.7918	34.6232

Framework versions

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