roberta-base-finetuned-ner-agglo-twitter

This model is a fine-tuned version of ArBert/roberta-base-finetuned-ner on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6645
• Precision: 0.6885
• Recall: 0.7665
• F1: 0.7254

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: 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: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1
No log	1.0	245	0.2820	0.6027	0.7543	0.6700
No log	2.0	490	0.2744	0.6308	0.7864	0.7000
0.2301	3.0	735	0.2788	0.6433	0.7637	0.6984
0.2301	4.0	980	0.3255	0.6834	0.7221	0.7022
0.1153	5.0	1225	0.3453	0.6686	0.7439	0.7043
0.1153	6.0	1470	0.3988	0.6797	0.7420	0.7094
0.0617	7.0	1715	0.4711	0.6702	0.7259	0.6969
0.0617	8.0	1960	0.4904	0.6904	0.7505	0.7192
0.0328	9.0	2205	0.5088	0.6591	0.7713	0.7108
0.0328	10.0	2450	0.5709	0.6468	0.7788	0.7067
0.019	11.0	2695	0.5570	0.6642	0.7533	0.7059
0.019	12.0	2940	0.5574	0.6899	0.7656	0.7258
0.0131	13.0	3185	0.5858	0.6952	0.7609	0.7265
0.0131	14.0	3430	0.6239	0.6556	0.7826	0.7135
0.0074	15.0	3675	0.5931	0.6825	0.7599	0.7191
0.0074	16.0	3920	0.6364	0.6785	0.7580	0.7161
0.005	17.0	4165	0.6437	0.6855	0.7580	0.7199
0.005	18.0	4410	0.6610	0.6779	0.7599	0.7166
0.0029	19.0	4655	0.6625	0.6853	0.7656	0.7232
0.0029	20.0	4900	0.6645	0.6885	0.7665	0.7254

Framework versions

• Transformers 4.16.2
• Pytorch 1.10.0+cu111
• Datasets 1.18.3
• Tokenizers 0.11.0