Modelpublic

tmnam20/mdeberta-v3-base-qnli-100

sourceHugging Facemitupdated 3y agoView on Hugging Face

0likes27downloads

Model Card

mdeberta-v3-base-qnli-100

This model is a fine-tuned version of microsoft/mdeberta-v3-base on the tmnam20/VieGLUE/QNLI dataset. It achieves the following results on the evaluation set:

—Loss: 0.2906
—Accuracy: 0.8975

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: 2e-05
—trainbatchsize: 32
—evalbatchsize: 16
—seed: 100
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.3773	0.15	500	0.3870	0.8431
0.3547	0.31	1000	0.3175	0.8658
0.3385	0.46	1500	0.2986	0.8739
0.342	0.61	2000	0.2787	0.8845
0.3003	0.76	2500	0.3075	0.8726
0.3298	0.92	3000	0.2781	0.8807
0.2475	1.07	3500	0.2695	0.8942
0.2441	1.22	4000	0.2615	0.8940
0.249	1.37	4500	0.2548	0.8958
0.2261	1.53	5000	0.2588	0.8946
0.2348	1.68	5500	0.2587	0.8982
0.2626	1.83	6000	0.2581	0.8982
0.2463	1.99	6500	0.2520	0.8964
0.1768	2.14	7000	0.2795	0.8951
0.1768	2.29	7500	0.3069	0.8942
0.1752	2.44	8000	0.2783	0.8971
0.1687	2.6	8500	0.2900	0.8995
0.163	2.75	9000	0.2828	0.8969
0.1547	2.9	9500	0.2873	0.8980

Framework versions

—Transformers 4.35.2
—Pytorch 2.2.0.dev20231203+cu121
—Datasets 2.15.0
—Tokenizers 0.15.0