distilbert_add_GLUE_Experiment_logit_kd_qqp_256

This model is a fine-tuned version of distilbert-base-uncased on the GLUE QQP dataset. It achieves the following results on the evaluation set:

• Loss: 0.6586
• Accuracy: 0.6554
• F1: 0.1310
• Combined Score: 0.3932

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
• train_batch_size: 256
• eval_batch_size: 256
• seed: 10
• distributed_type: multi-GPU
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Combined Score
0.8355	1.0	1422	0.8004	0.6318	0.0	0.3159
0.7677	2.0	2844	0.7488	0.6318	0.0	0.3159
0.7048	3.0	4266	0.6935	0.6318	0.0	0.3159
0.6449	4.0	5688	0.6875	0.6337	0.0106	0.3221
0.6082	5.0	7110	0.6688	0.6354	0.0205	0.3279
0.5829	6.0	8532	0.6854	0.6386	0.0393	0.3389
0.5637	7.0	9954	0.6707	0.6522	0.1155	0.3839
0.5502	8.0	11376	0.6752	0.6522	0.1145	0.3833
0.5389	9.0	12798	0.6677	0.6561	0.1348	0.3954
0.5304	10.0	14220	0.6693	0.6622	0.1659	0.4140
0.5234	11.0	15642	0.6728	0.6511	0.1082	0.3797
0.5175	12.0	17064	0.6812	0.6554	0.1303	0.3928
0.5127	13.0	18486	0.6644	0.6540	0.1235	0.3888
0.5085	14.0	19908	0.6605	0.6622	0.1677	0.4149
0.505	15.0	21330	0.6647	0.6570	0.1391	0.3980
0.502	16.0	22752	0.6667	0.6528	0.1170	0.3849
0.499	17.0	24174	0.6586	0.6554	0.1310	0.3932
0.497	18.0	25596	0.6589	0.6597	0.1535	0.4066
0.4947	19.0	27018	0.6715	0.6599	0.1535	0.4067
0.4928	20.0	28440	0.6631	0.6535	0.1202	0.3868
0.4907	21.0	29862	0.6690	0.6651	0.1796	0.4224
0.4891	22.0	31284	0.6603	0.6652	0.1830	0.4241

Framework versions

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