whisper-tiny-khmer-aug-kcc

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: 0.6705
• Wer: 46.2523

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
2.0089	1.0	810	0.8377	64.9564
0.8348	2.0	1620	0.6317	51.0793
0.6384	3.0	2430	0.5752	51.3541
0.5227	4.0	3240	0.5643	48.8592
0.4323	5.0	4050	0.5749	48.2771
0.3649	6.0	4860	0.5818	46.6970
0.3092	7.0	5670	0.6012	46.0317
0.2611	8.0	6480	0.6331	46.0643
0.2239	9.0	7290	0.6683	46.4222
0.1963	10.0	8100	0.6705	46.2523

Framework versions

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