Modelpublic

gokuls/bert_12_layer_model_v2_complete_training_new_96

sourceHugging Faceupdated 3y agoView on Hugging Face

0likes12downloads

Model Card

bert12layermodelv2completetrainingnew96

This model is a fine-tuned version of gokuls/bert_12_layer_model_v2_complete_training_new_72 on the None dataset. It achieves the following results on the evaluation set:

—Loss: 2.6222
—Accuracy: 0.5276

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: 1e-05
—trainbatchsize: 48
—evalbatchsize: 48
—seed: 10
—distributed_type: multi-GPU
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—lrschedulerwarmup_steps: 10000
—num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
3.0388	0.08	10000	2.9330	0.4848
2.9843	0.16	20000	2.8849	0.4916
2.9373	0.25	30000	2.8368	0.4984
2.9099	0.33	40000	2.8001	0.5033
2.8637	0.41	50000	2.7610	0.5085
2.8368	0.49	60000	2.7313	0.5126
2.7988	0.57	70000	2.7014	0.5167
2.7719	0.66	80000	2.6760	0.5199
2.7385	0.74	90000	2.6492	0.5234
2.7161	0.82	100000	2.6222	0.5276

Framework versions

—Transformers 4.30.1
—Pytorch 1.14.0a0+410ce96
—Datasets 2.12.0
—Tokenizers 0.13.3