whisper-base-khmer-aug-v2

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

• Loss: 0.3641
• Wer: 74.0879

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: 0.0001
• 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: constant
• lr_scheduler_warmup_steps: 1000
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.948	0.9989	669	0.5716	93.8220
0.4722	1.9993	1339	0.4280	89.0871
0.3698	2.9996	2009	0.3862	84.2711
0.3161	4.0	2679	0.3621	83.6549
0.2794	4.9989	3348	0.3442	77.5742
0.2473	5.9993	4018	0.3291	81.9523
0.2231	6.9996	4688	0.3443	77.3958
0.2023	8.0	5358	0.3464	76.8445
0.1835	8.9989	6027	0.3557	75.7419
0.1654	9.9888	6690	0.3641	74.0879

Framework versions

• Transformers 4.44.0
• Pytorch 2.3.1
• Datasets 2.21.0
• Tokenizers 0.19.1