bert_uncased_L-4_H-128_A-2_mrpc

This model is a fine-tuned version of google/bert_uncased_L-4_H-128_A-2 on the GLUE MRPC dataset. It achieves the following results on the evaluation set:

• Loss: 0.5327
• Accuracy: 0.7574
• F1: 0.8401
• Combined Score: 0.7987

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
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Combined Score
0.6437	1.0	15	0.6181	0.6838	0.8122	0.7480
0.6197	2.0	30	0.6047	0.6912	0.8158	0.7535
0.595	3.0	45	0.5877	0.6985	0.8161	0.7573
0.582	4.0	60	0.5687	0.7279	0.8284	0.7782
0.5617	5.0	75	0.5594	0.7279	0.8295	0.7787
0.5409	6.0	90	0.5550	0.7132	0.8208	0.7670
0.5213	7.0	105	0.5417	0.7255	0.8245	0.7750
0.4968	8.0	120	0.5530	0.7328	0.8310	0.7819
0.4741	9.0	135	0.5580	0.7353	0.8333	0.7843
0.4545	10.0	150	0.5390	0.7549	0.8397	0.7973
0.4366	11.0	165	0.5327	0.7574	0.8401	0.7987
0.4206	12.0	180	0.5350	0.7598	0.8424	0.8011
0.397	13.0	195	0.5649	0.7549	0.8447	0.7998
0.3873	14.0	210	0.5602	0.7623	0.8482	0.8052
0.3725	15.0	225	0.5622	0.7525	0.8399	0.7962
0.3506	16.0	240	0.5588	0.7525	0.8374	0.7949

Framework versions

• Transformers 4.46.3
• Pytorch 2.2.1+cu118
• Datasets 2.17.0
• Tokenizers 0.20.3