predict-perception-bert-blame-object

This model is a fine-tuned version of dbmdz/bert-base-italian-xxl-cased on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.5837
• Rmse: 0.5589
• Rmse Blame::a Un oggetto: 0.5589
• Mae: 0.3862
• Mae Blame::a Un oggetto: 0.3862
• R2: 0.2884
• R2 Blame::a Un oggetto: 0.2884
• Cos: 0.3913
• Pair: 0.0
• Rank: 0.5
• Neighbors: 0.5024
• Rsa: nan

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: 1e-05
• train_batch_size: 20
• eval_batch_size: 8
• seed: 1996
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Rmse	Rmse Blame::a Un oggetto	Mae	Mae Blame::a Un oggetto	R2	R2 Blame::a Un oggetto	Cos	Pair	Rank	Neighbors	Rsa
1.0603	1.0	15	0.8503	0.6745	0.6745	0.4386	0.4386	-0.0365	-0.0365	0.1304	0.0	0.5	0.5197	nan
0.9662	2.0	30	0.8510	0.6748	0.6748	0.4548	0.4548	-0.0374	-0.0374	0.0435	0.0	0.5	0.4840	nan
0.9438	3.0	45	0.7622	0.6386	0.6386	0.4541	0.4541	0.0709	0.0709	0.0435	0.0	0.5	0.4635	nan
0.9096	4.0	60	0.8301	0.6665	0.6665	0.4305	0.4305	-0.0119	-0.0119	0.0435	0.0	0.5	0.3499	nan
0.8383	5.0	75	0.7306	0.6252	0.6252	0.3814	0.3814	0.1094	0.1094	0.3043	0.0	0.5	0.5098	nan
0.7828	6.0	90	0.7434	0.6307	0.6307	0.4005	0.4005	0.0937	0.0937	0.3043	0.0	0.5	0.4335	nan
0.7028	7.0	105	0.7218	0.6214	0.6214	0.4090	0.4090	0.1202	0.1202	0.3913	0.0	0.5	0.4470	nan
0.6661	8.0	120	0.7434	0.6307	0.6307	0.4042	0.4042	0.0938	0.0938	0.3913	0.0	0.5	0.4470	nan
0.578	9.0	135	0.7719	0.6426	0.6426	0.3975	0.3975	0.0591	0.0591	0.3913	0.0	0.5	0.4470	nan
0.544	10.0	150	0.7117	0.6171	0.6171	0.4126	0.4126	0.1324	0.1324	0.2174	0.0	0.5	0.3489	nan
0.4638	11.0	165	0.6683	0.5980	0.5980	0.3952	0.3952	0.1853	0.1853	0.3043	0.0	0.5	0.3989	nan
0.3998	12.0	180	0.6772	0.6019	0.6019	0.4201	0.4201	0.1745	0.1745	0.3043	0.0	0.5	0.3989	nan
0.3403	13.0	195	0.6576	0.5932	0.5932	0.4237	0.4237	0.1984	0.1984	0.2174	0.0	0.5	0.3491	nan
0.2839	14.0	210	0.6281	0.5797	0.5797	0.4208	0.4208	0.2344	0.2344	0.2174	0.0	0.5	0.3491	nan
0.2619	15.0	225	0.6254	0.5785	0.5785	0.3752	0.3752	0.2376	0.2376	0.3913	0.0	0.5	0.5756	nan
0.2175	16.0	240	0.6074	0.5701	0.5701	0.3985	0.3985	0.2596	0.2596	0.3043	0.0	0.5	0.4142	nan
0.1884	17.0	255	0.6045	0.5687	0.5687	0.4036	0.4036	0.2631	0.2631	0.3913	0.0	0.5	0.5024	nan
0.1797	18.0	270	0.6038	0.5684	0.5684	0.3914	0.3914	0.2640	0.2640	0.3913	0.0	0.5	0.5024	nan
0.1316	19.0	285	0.6199	0.5759	0.5759	0.4078	0.4078	0.2443	0.2443	0.3913	0.0	0.5	0.5024	nan
0.1429	20.0	300	0.6119	0.5722	0.5722	0.3954	0.3954	0.2540	0.2540	0.3913	0.0	0.5	0.5024	nan
0.1202	21.0	315	0.6193	0.5756	0.5756	0.3987	0.3987	0.2451	0.2451	0.3913	0.0	0.5	0.5024	nan
0.1159	22.0	330	0.6218	0.5768	0.5768	0.3995	0.3995	0.2420	0.2420	0.3913	0.0	0.5	0.5024	nan
0.1027	23.0	345	0.6207	0.5763	0.5763	0.4100	0.4100	0.2433	0.2433	0.3043	0.0	0.5	0.4142	nan
0.1006	24.0	360	0.5646	0.5496	0.5496	0.3687	0.3687	0.3117	0.3117	0.3913	0.0	0.5	0.5024	nan
0.0902	25.0	375	0.5582	0.5465	0.5465	0.3714	0.3714	0.3196	0.3196	0.3913	0.0	0.5	0.5024	nan
0.0901	26.0	390	0.5650	0.5498	0.5498	0.3704	0.3704	0.3112	0.3112	0.3913	0.0	0.5	0.5024	nan
0.0937	27.0	405	0.5713	0.5529	0.5529	0.3735	0.3735	0.3036	0.3036	0.3913	0.0	0.5	0.5024	nan
0.0812	28.0	420	0.5773	0.5558	0.5558	0.3759	0.3759	0.2962	0.2962	0.3913	0.0	0.5	0.5024	nan
0.0911	29.0	435	0.5818	0.5579	0.5579	0.3832	0.3832	0.2908	0.2908	0.3913	0.0	0.5	0.5024	nan
0.082	30.0	450	0.5837	0.5589	0.5589	0.3862	0.3862	0.2884	0.2884	0.3913	0.0	0.5	0.5024	nan

Framework versions

• Transformers 4.16.2
• Pytorch 1.10.2+cu113
• Datasets 1.18.3
• Tokenizers 0.11.0