arbml/whisper-medium-ar

openai/whisper-medium

This model is a fine-tuned version of openai/whisper-medium on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.8488
• Wer: 16.5882

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: 32
• 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: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2963	0.1	1000	0.9115	27.3641
0.2676	0.2	2000	0.8796	24.1024
0.3166	0.3	3000	0.8467	20.1700
0.2797	0.4	4000	0.8756	29.4889
0.2302	0.5	5000	0.8523	19.6414
0.2803	0.6	6000	0.8715	19.7413
0.2794	0.7	7000	0.8548	18.6840
0.2173	0.8	8000	0.8543	17.9019
0.217	0.9	9000	0.8518	16.3840
0.1718	1.0	10000	0.8488	16.5882

Framework versions

• Transformers 4.26.0.dev0
• Pytorch 1.13.0+cu117
• Datasets 2.7.1.dev0
• Tokenizers 0.13.2