wang24

This model is a fine-tuned version of airesearch/wangchanberta-base-att-spm-uncased on the Legal dataset. This model change Attention Head and Hidden Layer to 24 It achieves the following results on the evaluation set:

• Loss: 0.1510

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss
0.1871	1.0	52	0.1757
0.1893	2.0	104	0.1812
0.1822	3.0	156	0.1843
0.1864	4.0	208	0.1575
0.1767	5.0	260	0.1862
0.147	6.0	312	0.1431
0.14	7.0	364	0.1509
0.1164	8.0	416	0.1426
0.1027	9.0	468	0.1499
0.1195	10.0	520	0.1510

Framework versions

• Transformers 4.39.3
• Pytorch 2.1.2
• Datasets 2.18.0
• Tokenizers 0.15.2