large-v2-no-bg-v1

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

• Loss: 0.6263

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-06
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.01
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
1.6361	142.8571	500	1.6697
1.257	285.7143	1000	1.2417
0.9481	428.5714	1500	0.8995
0.8397	571.4286	2000	0.8044
0.7741	714.2857	2500	0.7487
0.7336	857.1429	3000	0.7035
0.6932	1000.0	3500	0.6689
0.6626	1142.8571	4000	0.6449
0.6501	1285.7143	4500	0.6310
0.6413	1428.5714	5000	0.6263

Framework versions

• PEFT 0.10.0
• Transformers 4.41.0.dev0
• Pytorch 2.2.2+cu121
• Datasets 2.19.0
• Tokenizers 0.19.1