deberta-v3-base-finetuned-finance-text-classification

This model is a fine-tuned version of microsoft/deberta-v3-base on the sentence_50Agree financial-phrasebank + Kaggle Dataset, a dataset consisting of 4840 Financial News categorised by sentiment (negative, neutral, positive). The Kaggle dataset includes Covid-19 sentiment data and can be found here: sentiment-classification-selflabel-dataset.

It achieves the following results on the evaluation set:

• Loss: 0.7687
• Accuracy: 0.8913
• F1: 0.8912
• Precision: 0.8927
• Recall: 0.8913

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: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
No log	1.0	285	0.4187	0.8399	0.8407	0.8687	0.8399
0.5002	2.0	570	0.3065	0.8755	0.8733	0.8781	0.8755
0.5002	3.0	855	0.4148	0.8775	0.8775	0.8778	0.8775
0.1937	4.0	1140	0.4249	0.8696	0.8699	0.8719	0.8696
0.1937	5.0	1425	0.5121	0.8834	0.8824	0.8831	0.8834
0.0917	6.0	1710	0.6113	0.8775	0.8779	0.8839	0.8775
0.0917	7.0	1995	0.7296	0.8775	0.8776	0.8793	0.8775
0.0473	8.0	2280	0.7034	0.8953	0.8942	0.8964	0.8953
0.0275	9.0	2565	0.6995	0.8834	0.8836	0.8846	0.8834
0.0275	10.0	2850	0.7736	0.8755	0.8755	0.8789	0.8755
0.0186	11.0	3135	0.7173	0.8814	0.8814	0.8840	0.8814
0.0186	12.0	3420	0.7659	0.8854	0.8852	0.8873	0.8854
0.0113	13.0	3705	0.8415	0.8854	0.8855	0.8907	0.8854
0.0113	14.0	3990	0.7577	0.8953	0.8951	0.8966	0.8953
0.0074	15.0	4275	0.7687	0.8913	0.8912	0.8927	0.8913

Framework versions

• Transformers 4.19.2
• Pytorch 1.11.0+cu113
• Datasets 2.2.2
• Tokenizers 0.12.1