beratcmn/whisper-base-tr

Whisper Base TR

This model is a fine-tuned version of openai/whisper-base on the Common Voice 13 Turkish 30% dataset. It achieves the following results on the evaluation set:

• Loss: 0.4968
• Wer: 41.2122

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: 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
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.3817	0.5	33	0.5206	42.0632
0.2896	1.0	66	0.5182	44.3036
0.4421	1.5	99	0.5153	43.3137
0.187	2.0	132	0.5079	42.1501
0.2459	2.5	165	0.5001	41.7506
0.2297	3.0	198	0.4968	41.2122

Framework versions

• Transformers 4.35.0
• Pytorch 2.0.0
• Datasets 2.14.6
• Tokenizers 0.14.1