bertweet-base_regression_7_seed13_EN

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

• Loss: 1.0892
• Mse: 5.5584
• Rmse: 2.3576
• Mae: 1.3807
• R2: 0.2207
• F1: 0.7757
• Precision: 0.7780
• Recall: 0.7797
• Accuracy: 0.7797

Model description

This is the best-performing REGRESSION model for English irony detection. The model was fine-tuned both a mix of English and Dutch tweets. The model predicts one numerical value indicating irony likelihood, where 0 is not ironic and 6 is ironic.

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: 5e-06
• 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
• lr_scheduler_warmup_steps: 200
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Mse	Rmse	Mae	R2	F1	Precision	Recall	Accuracy
1.7333	0.4630	100	1.8475	9.7883	3.1286	2.2877	-0.4094	0.4570	0.3669	0.6057	0.6057
1.6952	0.9259	200	1.7889	8.8708	2.9784	2.2442	-0.2773	0.4570	0.3669	0.6057	0.6057
1.6175	1.3889	300	1.6295	7.6123	2.7590	2.0223	-0.0961	0.4570	0.3669	0.6057	0.6057
1.4401	1.8519	400	1.4962	6.6368	2.5762	1.8601	0.0444	0.4570	0.3669	0.6057	0.6057
1.2553	2.3148	500	1.3949	5.9003	2.4291	1.7518	0.1504	0.4570	0.3669	0.6057	0.6057
1.2296	2.7778	600	1.3520	5.9339	2.4360	1.6730	0.1456	0.4570	0.3669	0.6057	0.6057
1.0909	3.2407	700	1.2565	5.3251	2.3076	1.5831	0.2332	0.4570	0.3669	0.6057	0.6057
1.0031	3.7037	800	1.2159	4.7598	2.1817	1.5709	0.3146	0.4570	0.3669	0.6057	0.6057
0.9833	4.1667	900	1.1544	4.6141	2.1480	1.5031	0.3356	0.7296	0.8018	0.7572	0.7572
0.825	4.6296	1000	1.1512	5.0019	2.2365	1.4608	0.2798	0.7757	0.7943	0.7859	0.7859
0.8187	5.0926	1100	1.1150	4.9111	2.2161	1.4352	0.2928	0.7815	0.7849	0.7859	0.7859
0.7138	5.5556	1200	1.0724	4.8492	2.2021	1.3871	0.3018	0.7766	0.7791	0.7807	0.7807
0.6706	6.0185	1300	1.0560	4.9024	2.2141	1.3650	0.2941	0.7786	0.7823	0.7833	0.7833
0.6112	6.4815	1400	1.0594	5.0772	2.2533	1.3694	0.2689	0.7750	0.7759	0.7781	0.7781
0.5906	6.9444	1500	1.0611	5.1421	2.2676	1.3794	0.2596	0.7736	0.7734	0.7755	0.7755
0.5597	7.4074	1600	1.0286	5.0419	2.2454	1.3290	0.2740	0.7839	0.7879	0.7885	0.7885
0.5422	7.8704	1700	1.0531	5.2061	2.2817	1.3596	0.2504	0.7672	0.7678	0.7702	0.7702
0.5255	8.3333	1800	1.0478	5.2565	2.2927	1.3372	0.2431	0.7811	0.7853	0.7859	0.7859
0.5116	8.7963	1900	1.0544	5.2090	2.2823	1.3546	0.2500	0.7721	0.7733	0.7755	0.7755
0.5213	9.2593	2000	1.0423	5.1715	2.2741	1.3341	0.2554	0.7819	0.7846	0.7859	0.7859
0.4999	9.7222	2100	1.0566	5.2819	2.2982	1.3481	0.2395	0.7721	0.7733	0.7755	0.7755

Framework versions

• Transformers 4.40.2
• Pytorch 2.1.2
• Datasets 2.18.0
• Tokenizers 0.19.1