distilbert_sa_GLUE_Experiment_logit_kd_qqp_192

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: 0.7047
• Accuracy: 0.6401
• F1: 0.0484
• Combined Score: 0.3442

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
0.824	1.0	1422	0.7761	0.6318	0.0	0.3159
0.7581	2.0	2844	0.7544	0.6318	0.0	0.3159
0.7333	3.0	4266	0.7480	0.6318	0.0	0.3159
0.713	4.0	5688	0.7358	0.6318	0.0	0.3159
0.695	5.0	7110	0.7279	0.6365	0.0280	0.3323
0.6783	6.0	8532	0.7193	0.6370	0.0313	0.3341
0.6642	7.0	9954	0.7148	0.6332	0.0083	0.3208
0.6509	8.0	11376	0.7138	0.6409	0.0536	0.3472
0.639	9.0	12798	0.7084	0.6385	0.0398	0.3391
0.6287	10.0	14220	0.7117	0.6418	0.0591	0.3504
0.6195	11.0	15642	0.7064	0.6379	0.0363	0.3371
0.6101	12.0	17064	0.7079	0.6444	0.0725	0.3584
0.6023	13.0	18486	0.7059	0.6495	0.1028	0.3762
0.5948	14.0	19908	0.7103	0.6418	0.0577	0.3497
0.589	15.0	21330	0.7077	0.6415	0.0567	0.3491
0.5827	16.0	22752	0.7047	0.6401	0.0484	0.3442
0.5769	17.0	24174	0.7051	0.6409	0.0525	0.3467
0.572	18.0	25596	0.7090	0.6492	0.1017	0.3754
0.5673	19.0	27018	0.7078	0.6398	0.0467	0.3433
0.563	20.0	28440	0.7065	0.6427	0.0630	0.3529
0.5591	21.0	29862	0.7084	0.6464	0.0839	0.3651

Framework versions

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