whisper-tiny-khmer-aug

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: 0.2697
• Wer: 68.1206

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
0.8969	1.0	670	0.4289	91.3572
0.3822	2.0	1340	0.3025	79.3741
0.292	3.0	2010	0.2727	85.6819
0.2439	4.0	2680	0.2637	74.1365
0.2124	5.0	3350	0.2548	70.2124
0.1844	6.0	4020	0.2606	79.7470
0.1651	7.0	4690	0.2505	68.5909
0.1472	8.0	5360	0.2637	67.9261
0.1344	9.0	6030	0.2672	66.2234
0.1194	10.0	6700	0.2697	68.1206

Framework versions

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