geninhu/whisper-large-v2-az

Whisper Large v2 Azerbaijani

This model is a fine-tuned version of openai/whisper-large-v2 on the mozilla-foundation/common_voice_11_0 az dataset. It achieves the following results on the evaluation set:

• Loss: 0.9435
• Wer: 38.4615

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• 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
0.0	999.0	1000	0.8373	39.6450
0.0	1999.0	2000	0.9435	38.4615
0.0	2999.0	3000	1.0010	43.1953
0.0	3999.0	4000	1.0380	44.3787
0.0	4999.0	5000	1.0529	43.7870

Framework versions

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