whisper-tiny-khmer-aug-v5

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

• Loss: 0.2692
• Wer: 65.8991

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: 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: constant
• lr_scheduler_warmup_steps: 1000
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.8233	1.0	793	0.4089	85.6008
0.3463	2.0	1586	0.2983	79.4389
0.2618	3.0	2379	0.2675	83.5739
0.2207	4.0	3172	0.2512	76.1635
0.1929	5.0	3965	0.2446	69.3692
0.1709	6.0	4758	0.2454	70.6502
0.1526	7.0	5551	0.2439	74.9635
0.1382	8.0	6344	0.2515	66.4018
0.1241	9.0	7137	0.2559	66.8721
0.1138	10.0	7930	0.2692	65.8991

Framework versions

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