twhin-bert-large-finetuned-fintwitter-classification

This model is a fine-tuned version of Twitter/twhin-bert-large on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6936
• Accuracy: 0.8903
• F1: 0.8903
• Precision: 0.8903
• Recall: 0.8903

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: 128
• eval_batch_size: 128
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
No log	1.0	75	0.5087	0.7835	0.7594	0.7846	0.7835
No log	2.0	150	0.3652	0.8685	0.8696	0.8735	0.8685
No log	3.0	225	0.3452	0.8723	0.8739	0.8772	0.8723
No log	4.0	300	0.3332	0.8823	0.8830	0.8840	0.8823
No log	5.0	375	0.3618	0.8907	0.8909	0.8912	0.8907
No log	6.0	450	0.3995	0.8802	0.8807	0.8814	0.8802
0.2712	7.0	525	0.4459	0.8886	0.8888	0.8889	0.8886
0.2712	8.0	600	0.5018	0.8903	0.8897	0.8894	0.8903
0.2712	9.0	675	0.5457	0.8836	0.8844	0.8860	0.8836
0.2712	10.0	750	0.5562	0.8723	0.8725	0.8750	0.8723
0.2712	11.0	825	0.6080	0.8827	0.8827	0.8828	0.8827
0.2712	12.0	900	0.5886	0.8853	0.8852	0.8851	0.8853
0.2712	13.0	975	0.6021	0.8874	0.8878	0.8890	0.8874
0.0383	14.0	1050	0.5960	0.8853	0.8849	0.8846	0.8853
0.0383	15.0	1125	0.6393	0.8815	0.8819	0.8827	0.8815
0.0383	16.0	1200	0.6803	0.8844	0.8852	0.8865	0.8844
0.0383	17.0	1275	0.6839	0.8915	0.8915	0.8914	0.8915
0.0383	18.0	1350	0.6968	0.8907	0.8901	0.8899	0.8907
0.0383	19.0	1425	0.6971	0.8865	0.8870	0.8878	0.8865
0.0122	20.0	1500	0.6936	0.8903	0.8903	0.8903	0.8903

Framework versions

• Transformers 4.25.1
• Pytorch 1.13.0+cu116
• Datasets 2.8.0
• Tokenizers 0.13.2