Mehtap/base_12

base Turkish Whisper (bTW)

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

• Loss: 1.0576
• Wer: 1.1825
• Cer: 1.0651

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
• gradient_accumulation_steps: 4
• 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: 1000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
1.6978	3.33	100	1.3610	0.7852	0.4184
0.6547	6.66	200	0.8659	0.7226	0.4379
0.3805	9.99	300	0.8060	0.7256	0.4330
0.1886	13.33	400	0.8382	0.6395	0.4164
0.0745	16.66	500	0.9106	0.8185	0.6747
0.0303	19.99	600	0.9697	0.8509	0.5685
0.0139	23.33	700	1.0096	0.8773	0.6483
0.0069	26.66	800	1.0367	1.2781	1.2923
0.0054	29.99	900	1.0518	1.2363	1.1066
0.0048	33.33	1000	1.0576	1.1825	1.0651

Framework versions

• Transformers 4.26.0
• Pytorch 1.12.0+cu102
• Datasets 2.9.0
• Tokenizers 0.13.2