Modelpublic

gokuls/mobilebert_sa_GLUE_Experiment_logit_kd_mnli_128

sourceHugging Faceapache-2.0updated 4y agoView on Hugging Face

0likes21downloads

Model Card

mobilebertsaGLUEExperimentlogitkdmnli_128

This model is a fine-tuned version of google/mobilebert-uncased on the GLUE MNLI dataset. It achieves the following results on the evaluation set:

—Loss: 1.2689
—Accuracy: 0.5950

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: 128
—evalbatchsize: 128
—seed: 10
—distributed_type: multi-GPU
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.6825	1.0	3068	1.4581	0.5256
1.4941	2.0	6136	1.3516	0.5680
1.4199	3.0	9204	1.3259	0.5712
1.3747	4.0	12272	1.3024	0.5856
1.34	5.0	15340	1.2875	0.5931
1.3087	6.0	18408	1.2730	0.5928
1.2769	7.0	21476	1.2845	0.5916
1.246	8.0	24544	1.2750	0.5965
1.2166	9.0	27612	1.2651	0.6020
1.1883	10.0	30680	1.2773	0.6043
1.1604	11.0	33748	1.2555	0.6011
1.1329	12.0	36816	1.2792	0.5991
1.1074	13.0	39884	1.2891	0.5986
1.0812	14.0	42952	1.2889	0.5947
1.0577	15.0	46020	1.2871	0.5970
1.0338	16.0	49088	1.3296	0.6026

Framework versions

—Transformers 4.26.0
—Pytorch 1.14.0a0+410ce96
—Datasets 2.9.0
—Tokenizers 0.13.2