Modelpublic

YeRyeongLee/albert-base-v2-finetuned-filtered-0609

sourceHugging Faceapache-2.0updated 4y agoView on Hugging Face

0likes59downloads

Model Card

albert-base-v2-finetuned-filtered-0609

This model is a fine-tuned version of albert-base-v2 on an unknown dataset. It achieves the following results on the evaluation set:

—Loss: 0.2062
—Accuracy: 0.9723
—Precision: 0.9724
—Recall: 0.9723
—F1: 0.9723

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
—trainbatchsize: 8
—evalbatchsize: 8
—seed: 42
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—lrschedulerwarmup_steps: 1000
—num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
0.2688	1.0	3180	0.2282	0.9560	0.9577	0.9560	0.9562
0.2268	2.0	6360	0.1909	0.9638	0.9640	0.9638	0.9638
0.1831	3.0	9540	0.2590	0.9572	0.9584	0.9572	0.9572
0.1588	4.0	12720	0.1752	0.9673	0.9678	0.9673	0.9673
0.0972	5.0	15900	0.1868	0.9695	0.9696	0.9695	0.9695
0.0854	6.0	19080	0.2042	0.9701	0.9707	0.9701	0.9702
0.0599	7.0	22260	0.1793	0.9748	0.9749	0.9748	0.9749
0.0389	8.0	25440	0.1996	0.9742	0.9743	0.9742	0.9742
0.0202	9.0	28620	0.2188	0.9723	0.9726	0.9723	0.9724
0.0152	10.0	31800	0.2062	0.9723	0.9724	0.9723	0.9723

Framework versions

—Transformers 4.19.2
—Pytorch 1.9.1+cu111
—Datasets 1.16.1
—Tokenizers 0.12.1