bert-16

This model is a fine-tuned version of deepset/bert-base-cased-squad2 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 10.2625

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: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss
11.3013	0.09	5	12.2645
11.351	0.18	10	12.1394
10.8708	0.27	15	12.0185
10.9286	0.36	20	11.9023
10.8502	0.45	25	11.7895
10.806	0.55	30	11.6807
10.0507	0.64	35	11.5771
10.6057	0.73	40	11.4770
10.1224	0.82	45	11.3818
9.9648	0.91	50	11.2900
9.7027	1.0	55	11.2029
9.8434	1.09	60	11.1193
9.8999	1.18	65	11.0406
10.0717	1.27	70	10.9654
9.5874	1.36	75	10.8939
9.9629	1.45	80	10.8264
9.6759	1.55	85	10.7628
9.6001	1.64	90	10.7023
9.4276	1.73	95	10.6462
9.5867	1.82	100	10.5940
9.1448	1.91	105	10.5456
9.4274	2.0	110	10.5012
9.3621	2.09	115	10.4610
9.1924	2.18	120	10.4240
9.318	2.27	125	10.3911
9.4253	2.36	130	10.3618
9.0476	2.45	135	10.3361
9.3817	2.55	140	10.3147
8.9517	2.64	145	10.2968
9.4026	2.73	150	10.2828
9.1891	2.82	155	10.2723
9.1735	2.91	160	10.2654
9.3527	3.0	165	10.2625

Framework versions

• Transformers 4.34.1
• Pytorch 2.0.1+cu118
• Datasets 2.14.5
• Tokenizers 0.14.1