xlsr

This model is a fine-tuned version of S-Sethisak/xlsr-khmer-fleur-ex02 on the fleurs dataset. It achieves the following results on the evaluation set:

• Loss: 0.8011
• Wer: 0.6776

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: 6.25e-06
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 800
• training_steps: 8000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.1893	0.1434	400	1.3998	1.0
1.7694	0.2867	800	0.9742	0.9704
1.7196	0.4301	1200	0.8980	0.7788
1.8691	0.5735	1600	0.8685	0.7422
1.8432	0.7168	2000	0.8528	0.7295
1.8607	0.8602	2400	0.8395	0.7231
1.7744	1.0036	2800	0.8338	0.7122
1.6846	1.1470	3200	0.8259	0.7024
1.7989	1.2903	3600	0.8297	0.6974
1.5462	1.4337	4000	0.8212	0.6938
1.6145	1.5771	4400	0.8214	0.6908
1.4987	1.7204	4800	0.8172	0.6854
1.5861	1.8638	5200	0.8185	0.6835
1.6129	2.0072	5600	0.8144	0.6810
1.6523	2.1505	6000	0.8170	0.6788
1.5069	2.2939	6400	0.8116	0.6793
1.5815	2.4373	6800	0.8113	0.6780
1.4807	2.5806	7200	0.8069	0.6768
1.6869	2.7240	7600	0.8024	0.6777
1.712	2.8674	8000	0.8011	0.6776

Framework versions

• Transformers 4.52.4
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1