qthuan2604/BARTPho_Syllable_Restore_Diacritics_Vietnamese

PhoBERT_Encoder_Vietnamese

This model is a fine-tuned version of vinai/bartpho-syllable on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2662
• Cer: 0.3885
• Wer: 0.5551
• Char Acc: 0.4075

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: 3e-05
• train_batch_size: 24
• eval_batch_size: 96
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 96
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer	Wer	Char Acc
2.5328	0.99	91	0.5275	0.4963	0.7764	0.2969
0.5873	1.99	183	0.3111	0.4205	0.6115	0.3692
0.4335	2.98	273	0.2662	0.3885	0.5551	0.4075
0.3957	3.99	332	0.2027	0.358	0.4833	0.464
0.273	5	416	0.1413	0.3217	0.4135	0.5601
0.1961	5.98	498	0.1237	0.3102	0.3896	0.5871
0.1583	6.99	581	0.112	0.303	0.3743	0.6107
0.2442	8	665	0.1039	0.2987	0.3637	0.6237
0.098	8.98	747	0.0982	0.2977	0.3579	0.6309
0.077	9.99	830	0.0964	0.2959	0.3536	0.6365
0.0674	10.99	913	0.0971	0.2943	0.3505	0.6398
0.0621	11.99	996	0.0956	0.2951	0.3507	0.6382
0.0567	13	1080	0.0966	0.2933	0.3465	0.6439
0.051	13.98	1162	0.0966	0.2923	0.3452	0.648

Framework versions

• Transformers 4.43.3
• Pytorch 2.6.0+cu124
• Datasets 3.1.0
• Tokenizers 0.19.1