bert-18

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: 8.3340

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: 0.0001
• 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
11.2713	0.09	5	12.1606
11.199	0.18	10	11.9115
10.6074	0.27	15	11.6709
10.5475	0.36	20	11.4407
10.3761	0.45	25	11.2173
10.2166	0.55	30	11.0033
9.4143	0.64	35	10.7983
9.8307	0.73	40	10.6034
9.3026	0.82	45	10.4169
9.0636	0.91	50	10.2387
8.7689	1.0	55	10.0700
8.7969	1.09	60	9.9094
8.7596	1.18	65	9.7588
8.8433	1.27	70	9.6152
8.3576	1.36	75	9.4808
8.6226	1.45	80	9.3540
8.3176	1.55	85	9.2346
8.2174	1.64	90	9.1231
8.0514	1.73	95	9.0198
8.0813	1.82	100	8.9240
7.6971	1.91	105	8.8362
7.865	2.0	110	8.7562
7.7614	2.09	115	8.6834
7.6525	2.18	120	8.6179
7.7074	2.27	125	8.5593
7.7802	2.36	130	8.5073
7.4788	2.45	135	8.4625
7.6863	2.55	140	8.4245
7.3113	2.64	145	8.3934
7.6127	2.73	150	8.3692
7.471	2.82	155	8.3509
7.4979	2.91	160	8.3393
7.5977	3.0	165	8.3340

Framework versions

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