hBERTv1_new_pretrain_w_init_48_mrpc

This model is a fine-tuned version of gokuls/bert_12_layer_model_v1_complete_training_new_wt_init_48 on the GLUE MRPC dataset. It achieves the following results on the evaluation set:

• Loss: 0.6229
• Accuracy: 0.6838
• F1: 0.8122
• Combined Score: 0.7480

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
0.6607	1.0	29	0.6262	0.6838	0.8122	0.7480
0.6421	2.0	58	0.6368	0.6838	0.8122	0.7480
0.6411	3.0	87	0.6258	0.6838	0.8122	0.7480
0.6406	4.0	116	0.6422	0.6838	0.8122	0.7480
0.6364	5.0	145	0.6263	0.6838	0.8122	0.7480
0.6322	6.0	174	0.6253	0.6838	0.8122	0.7480
0.6398	7.0	203	0.6289	0.6838	0.8122	0.7480
0.6363	8.0	232	0.6267	0.6838	0.8122	0.7480
0.6374	9.0	261	0.6375	0.6838	0.8122	0.7480
0.6374	10.0	290	0.6248	0.6838	0.8122	0.7480
0.638	11.0	319	0.6262	0.6838	0.8122	0.7480
0.6353	12.0	348	0.6236	0.6838	0.8122	0.7480
0.6338	13.0	377	0.6263	0.6838	0.8122	0.7480
0.637	14.0	406	0.6250	0.6838	0.8122	0.7480
0.6375	15.0	435	0.6229	0.6838	0.8122	0.7480
0.7037	16.0	464	0.6438	0.6838	0.8122	0.7480
0.6198	17.0	493	0.6242	0.6961	0.8038	0.7499
0.5847	18.0	522	0.6260	0.6740	0.7742	0.7241
0.4983	19.0	551	0.7174	0.7034	0.8158	0.7596
0.4245	20.0	580	0.7737	0.6789	0.7828	0.7308

Framework versions

• Transformers 4.29.2
• Pytorch 1.14.0a0+410ce96
• Datasets 2.12.0
• Tokenizers 0.13.3