Modelpublic

th-nuernberg/gelectra-large-german-counseling-gecco

sourceHugging Facemitupdated 3y agoView on Hugging Face

0likes12downloads

Model Card

gecco-german-counseling-gelectra-large

This model is a fine-tuned version of deepset/gelectra-large on the None dataset. It achieves the following results on the evaluation set:

—Loss: 1.3342
—Accuracy: 0.7290
—F1: 0.4663

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: 64
—evalbatchsize: 64
—seed: 42
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—num_epochs: 16

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
3.6182	1.0	20	3.4739	0.1710	0.0391
3.3872	2.0	40	3.1759	0.2742	0.0806
3.1164	3.0	60	2.8804	0.3774	0.1428
2.8264	4.0	80	2.5862	0.5290	0.2440
2.5488	5.0	100	2.3090	0.5871	0.2816
2.2554	6.0	120	2.0804	0.6194	0.3196
2.0363	7.0	140	1.9051	0.6452	0.3435
1.8361	8.0	160	1.7684	0.6677	0.3593
1.5928	9.0	180	1.6380	0.7	0.4038
1.5584	10.0	200	1.5519	0.7097	0.4142
1.3548	11.0	220	1.4832	0.7194	0.4325
1.2585	12.0	240	1.4253	0.7226	0.4530
1.209	13.0	260	1.3874	0.7194	0.4512
1.0846	14.0	280	1.3526	0.7323	0.4667
1.0705	15.0	300	1.3377	0.7258	0.4643
1.0137	16.0	320	1.3342	0.7290	0.4663

Framework versions

—Transformers 4.35.1
—Pytorch 1.10.1+cu111
—Datasets 2.14.7
—Tokenizers 0.14.1