distilhubert-finetuned-grade

This model is a fine-tuned version of ntu-spml/distilhubert on the PQVD dataset. It achieves the following results on the evaluation set:

• Loss: 1.7494
• Accuracy: 0.6703

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: 5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use 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_ratio: 0.1
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.6335	1.0	92	0.5993	0.6593
0.4618	2.0	184	0.6011	0.6703
0.3811	3.0	276	0.7119	0.5604
0.4686	4.0	368	0.7788	0.6593
0.2984	5.0	460	0.8130	0.6044
0.2004	6.0	552	0.8343	0.6484
0.3806	7.0	644	0.9339	0.6264
0.1813	8.0	736	1.1104	0.5055
0.1335	9.0	828	1.1915	0.6813
0.2548	10.0	920	1.2242	0.6703
0.5109	11.0	1012	1.4604	0.6923
0.155	12.0	1104	1.5649	0.6593
0.1678	13.0	1196	1.6666	0.6703
0.1302	14.0	1288	1.7023	0.6813
0.0068	15.0	1380	1.7494	0.6703

Framework versions

• Transformers 4.49.0
• Pytorch 2.6.0+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0