twitter-roberta-base-WNUT

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

• Loss: 0.1938
• Precision: 0.7045
• Recall: 0.6304
• F1: 0.6654
• Accuracy: 0.9640

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: 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	0.3912	0.0	0.0	0.0	0.9205
No log	0.93	50	0.2847	0.25	0.0024	0.0047	0.9209
No log	1.39	75	0.2449	0.5451	0.3469	0.4240	0.9426
No log	1.85	100	0.1946	0.6517	0.4856	0.5565	0.9492
No log	2.31	125	0.1851	0.6921	0.5646	0.6219	0.9581
No log	2.78	150	0.1672	0.6867	0.5873	0.6331	0.9594
No log	3.24	175	0.1675	0.6787	0.5837	0.6277	0.9615
No log	3.7	200	0.1644	0.6765	0.6328	0.6539	0.9638
No log	4.17	225	0.1672	0.6997	0.6495	0.6737	0.9640
No log	4.63	250	0.1652	0.6915	0.6435	0.6667	0.9649
No log	5.09	275	0.1882	0.7067	0.6053	0.6521	0.9629
No log	5.56	300	0.1783	0.7128	0.6352	0.6717	0.9645
No log	6.02	325	0.1813	0.7011	0.6172	0.6565	0.9639
No log	6.48	350	0.1804	0.7139	0.6447	0.6776	0.9647
No log	6.94	375	0.1902	0.7218	0.6268	0.6709	0.9641
No log	7.41	400	0.1883	0.7106	0.6316	0.6688	0.9641
No log	7.87	425	0.1862	0.7067	0.6340	0.6683	0.9643
No log	8.33	450	0.1882	0.7053	0.6328	0.6671	0.9639
No log	8.8	475	0.1919	0.7055	0.6304	0.6658	0.9638
0.1175	9.26	500	0.1938	0.7045	0.6304	0.6654	0.9640
0.1175	9.72	525	0.1880	0.7025	0.6411	0.6704	0.9646

Framework versions

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