twitter-roberta-base-dec2021-WNUT

This model is a fine-tuned version of cardiffnlp/twitter-roberta-base-dec2021 on the wnut_17 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2152
• Precision: 0.7112
• Recall: 0.6244
• F1: 0.6650
• Accuracy: 0.9643

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: 64
• eval_batch_size: 1024
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	0.46	25	0.2818	0.0982	0.0383	0.0551	0.9241
No log	0.93	50	0.2158	0.6181	0.4569	0.5254	0.9480
No log	1.39	75	0.1930	0.6682	0.5347	0.5940	0.9555
No log	1.85	100	0.1728	0.6583	0.5646	0.6079	0.9594
No log	2.31	125	0.1787	0.7050	0.5718	0.6314	0.9619
No log	2.78	150	0.2051	0.6979	0.5251	0.5993	0.9587
No log	3.24	175	0.1755	0.7172	0.5945	0.6501	0.9621
No log	3.7	200	0.1720	0.6943	0.6304	0.6608	0.9645
No log	4.17	225	0.1873	0.7203	0.6316	0.6730	0.9646
No log	4.63	250	0.1781	0.6934	0.6196	0.6545	0.9638
No log	5.09	275	0.1953	0.7040	0.6172	0.6577	0.9631
No log	5.56	300	0.1953	0.7223	0.6316	0.6739	0.9642
No log	6.02	325	0.1839	0.7008	0.6471	0.6729	0.9648
No log	6.48	350	0.1995	0.716	0.6423	0.6772	0.9650
No log	6.94	375	0.2056	0.7251	0.6184	0.6675	0.9640
No log	7.41	400	0.2044	0.7065	0.6220	0.6616	0.9640
No log	7.87	425	0.2042	0.7201	0.6400	0.6776	0.9650
No log	8.33	450	0.2247	0.7280	0.6244	0.6722	0.9638
No log	8.8	475	0.2060	0.7064	0.6447	0.6742	0.9649
0.0675	9.26	500	0.2152	0.7112	0.6244	0.6650	0.9643
0.0675	9.72	525	0.2086	0.7070	0.6495	0.6771	0.9650

Framework versions

• Transformers 4.20.1
• Pytorch 1.12.0
• Datasets 2.3.2
• Tokenizers 0.12.1