Modelpublic

abdiharyadi/deberta-v3-large-ft-icar-a-v0.9

sourceHugging Facemitupdated 1y agoView on Hugging Face

0likes1downloads

Model Card

deberta-v3-large-ft-icar-a-v0.9

This model is a fine-tuned version of microsoft/deberta-v3-large on an unknown dataset. It achieves the following results on the evaluation set:

—Loss: 0.6068
—Accuracy: 0.9234
—Precision: 0.9256
—Recall: 0.9053
—F1: 0.9134

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-06
—trainbatchsize: 1
—evalbatchsize: 1
—seed: 42
—gradientaccumulationsteps: 3
—totaltrainbatch_size: 3
—optimizer: Use OptimizerNames.ADAMWTORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizerargs=No additional optimizer arguments
—lrschedulertype: linear
—num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
2.5395	1.0	871	0.6154	0.7596	0.6169	0.7051	0.6282
1.9256	2.0	1742	0.5633	0.9096	0.9215	0.8815	0.8938
1.455	3.0	2613	0.6652	0.9066	0.8990	0.8936	0.8957
1.1127	4.0	3484	0.6068	0.9234	0.9256	0.9053	0.9134
0.8832	5.0	4355	0.6356	0.9173	0.9076	0.9042	0.9055
0.6882	6.0	5226	0.7137	0.9158	0.9124	0.9071	0.9093
0.5421	7.0	6097	0.7485	0.9234	0.9236	0.9148	0.9186
0.3706	8.0	6968	0.7840	0.9234	0.9229	0.9077	0.9133
0.3018	9.0	7839	0.9213	0.9127	0.9041	0.9143	0.9067
0.2344	10.0	8710	0.9114	0.9204	0.9151	0.9170	0.9154
0.1504	11.0	9581	0.8238	0.9234	0.9191	0.9180	0.9175
0.163	12.0	10452	0.9299	0.9127	0.8989	0.9187	0.9056
0.1608	13.0	11323	0.8935	0.9188	0.9146	0.9177	0.9151
0.1165	14.0	12194	1.1499	0.9096	0.9023	0.9150	0.9066
0.0733	15.0	13065	1.0307	0.9158	0.9060	0.9165	0.9086
0.1196	16.0	13936	1.1132	0.9051	0.8965	0.8922	0.8935
0.0975	17.0	14807	1.0897	0.9142	0.9117	0.8938	0.9006
0.1164	18.0	15678	1.0514	0.9096	0.8975	0.9135	0.9027
0.1125	19.0	16549	1.1011	0.9051	0.8971	0.8963	0.8951

Framework versions

—Transformers 4.52.4
—Pytorch 2.6.0+cu124
—Datasets 4.0.0
—Tokenizers 0.21.2