ind_roberta

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

• Loss: 0.3951
• Accuracy@en: 0.9367
• F1@en: 0.9341
• Precision@en: 0.9360
• Recall@en: 0.9324
• Loss@en: 0.3951

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: 4
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 15

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy@en	F1@en	Precision@en	Recall@en	Loss@en
0.1854	1.0	375	0.4027	0.9033	0.8994	0.9012	0.8979	0.4027
0.203	2.0	750	0.4013	0.89	0.8877	0.8845	0.8944	0.4013
0.1282	3.0	1125	0.6106	0.89	0.8883	0.8858	0.8983	0.6106
0.0811	4.0	1500	0.3951	0.9367	0.9341	0.9360	0.9324	0.3951
0.0425	5.0	1875	0.4764	0.93	0.9282	0.9250	0.9333	0.4764
0.005	6.0	2250	0.5299	0.9367	0.9343	0.9349	0.9337	0.5299
0.0147	7.0	2625	0.5200	0.93	0.9285	0.9249	0.9359	0.5200
0.0182	8.0	3000	0.5532	0.9267	0.9242	0.9232	0.9253	0.5532
0.0125	9.0	3375	0.5398	0.9367	0.9346	0.9331	0.9363	0.5398
0.0171	10.0	3750	0.5157	0.9367	0.9349	0.9321	0.9389	0.5157
0.0109	11.0	4125	0.6538	0.92	0.9184	0.9149	0.9261	0.6538
0.0054	12.0	4500	0.5676	0.93	0.9281	0.9253	0.9320	0.5676
0.0047	13.0	4875	0.6763	0.9167	0.9146	0.9114	0.9195	0.6763
0.0076	14.0	5250	0.6970	0.9133	0.9109	0.9084	0.9141	0.6970
0.0066	15.0	5625	0.6947	0.9167	0.9146	0.9114	0.9195	0.6947

Framework versions

• Transformers 4.17.0
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2