whisper-tiny-khmer-aug-v3

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: 0.2718
• Wer: 65.2181

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
0.8309	1.0	793	0.4093	85.4873
0.3554	2.0	1586	0.2979	76.7634
0.2652	3.0	2379	0.2675	78.9201
0.2227	4.0	3172	0.2514	73.6014
0.1932	5.0	3965	0.2501	69.7908
0.1698	6.0	4758	0.2514	69.5476
0.1507	7.0	5551	0.2505	74.2014
0.1357	8.0	6344	0.2539	67.0991
0.1241	9.0	7137	0.2615	69.8719
0.1126	10.0	7930	0.2718	65.2181

Framework versions

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