tajroberto-ner

This model is a fine-tuned version of muhtasham/RoBERTa-tg on the wikiann dataset. It achieves the following results on the evaluation set:

• Loss: 0.9408
• Precision: 0.3155
• Recall: 0.5673
• F1: 0.4055
• Accuracy: 0.8360

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: 2e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 200

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	2.0	50	0.7710	0.0532	0.1827	0.0824	0.6933
No log	4.0	100	0.5901	0.0847	0.25	0.1265	0.7825
No log	6.0	150	0.5226	0.2087	0.4615	0.2874	0.8186
No log	8.0	200	0.5041	0.2585	0.5096	0.3430	0.8449
No log	10.0	250	0.5592	0.2819	0.5096	0.3630	0.8499
No log	12.0	300	0.5725	0.3032	0.5481	0.3904	0.8558
No log	14.0	350	0.6433	0.3122	0.5673	0.4027	0.8508
No log	16.0	400	0.6744	0.3543	0.5962	0.4444	0.8553
No log	18.0	450	0.7617	0.3353	0.5577	0.4188	0.8335
0.2508	20.0	500	0.7608	0.3262	0.5865	0.4192	0.8419
0.2508	22.0	550	0.8483	0.3224	0.5673	0.4111	0.8494
0.2508	24.0	600	0.8370	0.3275	0.5385	0.4073	0.8439
0.2508	26.0	650	0.8652	0.3410	0.5673	0.4260	0.8394
0.2508	28.0	700	0.9441	0.3409	0.5769	0.4286	0.8216
0.2508	30.0	750	0.9228	0.3333	0.5577	0.4173	0.8439
0.2508	32.0	800	0.9175	0.3430	0.5673	0.4275	0.8355
0.2508	34.0	850	0.9603	0.3073	0.5288	0.3887	0.8340
0.2508	36.0	900	0.9417	0.3240	0.5577	0.4099	0.8370
0.2508	38.0	950	0.9408	0.3155	0.5673	0.4055	0.8360

Framework versions

• Transformers 4.21.2
• Pytorch 1.12.1+cu113
• Datasets 2.4.0
• Tokenizers 0.12.1