finbert-tone-finetuned-finance-text-classification

This model is a fine-tuned version of yiyanghkust/finbert-tone on the nickmuchi/financial-text-combo-classification dataset which is a combined dataset of financial_phrasebank,FinanceInc/auditor_sentiment and zeroshot/twitter-financial-news-sentiment. It achieves the following results on the evaluation set:

• Loss: 0.6645
• Accuracy: 0.9097
• F1: 0.9102
• Precision: 0.9110
• Recall: 0.9097

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	141	0.3934	0.8431	0.8427	0.8456	0.8431
No log	2.0	282	0.3214	0.8843	0.8843	0.8867	0.8843
No log	3.0	423	0.3302	0.8882	0.8902	0.8965	0.8882
0.4444	4.0	564	0.3611	0.8980	0.8993	0.9026	0.8980
0.4444	5.0	705	0.4006	0.8975	0.8987	0.9014	0.8975
0.4444	6.0	846	0.4517	0.9037	0.9043	0.9057	0.9037
0.4444	7.0	987	0.5324	0.9027	0.9035	0.9057	0.9027
0.0406	8.0	1128	0.5308	0.9063	0.9074	0.9098	0.9063
0.0406	9.0	1269	0.5586	0.9081	0.9084	0.9089	0.9081
0.0406	10.0	1410	0.5783	0.9076	0.9080	0.9086	0.9076
0.0121	11.0	1551	0.5741	0.9115	0.9116	0.9121	0.9115
0.0121	12.0	1692	0.6288	0.9104	0.9108	0.9115	0.9104
0.0121	13.0	1833	0.6328	0.9050	0.9059	0.9078	0.9050
0.0121	14.0	1974	0.6887	0.9042	0.9054	0.9088	0.9042
0.0063	15.0	2115	0.6345	0.9086	0.9094	0.9109	0.9086
0.0063	16.0	2256	0.6545	0.9102	0.9103	0.9108	0.9102
0.0063	17.0	2397	0.6585	0.9086	0.9092	0.9103	0.9086
0.0033	18.0	2538	0.6676	0.9081	0.9087	0.9098	0.9081
0.0033	19.0	2679	0.6614	0.9110	0.9113	0.9119	0.9110
0.0033	20.0	2820	0.6645	0.9097	0.9102	0.9110	0.9097

Framework versions

• Transformers 4.25.1
• Pytorch 1.13.1+cu116
• Datasets 2.8.0
• Tokenizers 0.13.2