whisper-khmer-base-v5

This model is a fine-tuned version of PhanithLIM/whisper-khmer-base-v4 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1001
• Wer: 59.8788

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: 1e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: constant
• lr_scheduler_warmup_steps: 1000
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2476	1.0	61	0.1605	77.2121
0.1921	2.0	122	0.1410	72.8485
0.1687	3.0	183	0.1286	68.6061
0.1554	4.0	244	0.1222	68.0606
0.1433	5.0	305	0.1187	67.9394
0.1348	6.0	366	0.1105	64.8485
0.1284	7.0	427	0.1107	64.0606
0.1188	8.0	488	0.1069	63.6364
0.114	9.0	549	0.1047	62.4242
0.1098	9.8430	600	0.1001	59.8788

Framework versions

• Transformers 4.51.3
• Pytorch 2.5.1+cu121
• Datasets 3.6.0
• Tokenizers 0.21.0