hBERTv1_new_pretrain_48_KD_qqp

This model is a fine-tuned version of gokuls/bert_12_layer_model_v1_complete_training_new_48_KD on the GLUE QQP dataset. It achieves the following results on the evaluation set:

• Loss: 0.4106
• Accuracy: 0.8117
• F1: 0.7245
• Combined Score: 0.7681

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: 4e-05
• train_batch_size: 128
• eval_batch_size: 128
• 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

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Combined Score
0.5016	1.0	2843	0.4533	0.7776	0.6963	0.7370
0.4125	2.0	5686	0.4433	0.8023	0.7269	0.7646
0.3574	3.0	8529	0.4106	0.8117	0.7245	0.7681
0.3134	4.0	11372	0.4395	0.8208	0.7461	0.7834
0.279	5.0	14215	0.4975	0.8236	0.7627	0.7931
0.248	6.0	17058	0.5527	0.8129	0.7066	0.7598
0.2215	7.0	19901	0.4814	0.8209	0.7697	0.7953
0.1998	8.0	22744	0.4820	0.8272	0.7702	0.7987

Framework versions

• Transformers 4.30.2
• Pytorch 1.14.0a0+410ce96
• Datasets 2.12.0
• Tokenizers 0.13.3