predict-perception-bert-focus-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: 0.2271
• Rmse: 0.5965
• Rmse Focus::a Su un oggetto: 0.5965
• Mae: 0.4372
• Mae Focus::a Su un oggetto: 0.4372
• R2: 0.4957
• R2 Focus::a Su un oggetto: 0.4957
• Cos: 0.6522
• Pair: 0.0
• Rank: 0.5
• Neighbors: 0.6622
• 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 Focus::a Su un oggetto	Mae	Mae Focus::a Su un oggetto	R2	R2 Focus::a Su un oggetto	Cos	Pair	Rank	Neighbors	Rsa
1.0371	1.0	15	0.4358	0.8263	0.8263	0.7132	0.7132	0.0323	0.0323	0.3043	0.0	0.5	0.3510	nan
0.9574	2.0	30	0.4420	0.8321	0.8321	0.7175	0.7175	0.0186	0.0186	0.3043	0.0	0.5	0.4627	nan
0.9137	3.0	45	0.4208	0.8119	0.8119	0.6955	0.6955	0.0657	0.0657	0.3913	0.0	0.5	0.3928	nan
0.8465	4.0	60	0.3356	0.7251	0.7251	0.6237	0.6237	0.2548	0.2548	0.5652	0.0	0.5	0.6247	nan
0.6864	5.0	75	0.2876	0.6712	0.6712	0.5624	0.5624	0.3616	0.3616	0.5652	0.0	0.5	0.6247	nan
0.5804	6.0	90	0.3148	0.7022	0.7022	0.5577	0.5577	0.3011	0.3011	0.5652	0.0	0.5	0.6247	nan
0.4983	7.0	105	0.4068	0.7983	0.7983	0.6606	0.6606	0.0968	0.0968	0.3913	0.0	0.5	0.4519	nan
0.3584	8.0	120	0.2567	0.6342	0.6342	0.4883	0.4883	0.4300	0.4300	0.5652	0.0	0.5	0.6247	nan
0.2771	9.0	135	0.2130	0.5777	0.5777	0.4193	0.4193	0.5270	0.5270	0.6522	0.0	0.5	0.6622	nan
0.2135	10.0	150	0.2522	0.6285	0.6285	0.4572	0.4572	0.4401	0.4401	0.6522	0.0	0.5	0.6622	nan
0.1654	11.0	165	0.2662	0.6457	0.6457	0.4603	0.4603	0.4090	0.4090	0.6522	0.0	0.5	0.6622	nan
0.1554	12.0	180	0.2459	0.6207	0.6207	0.4778	0.4778	0.4540	0.4540	0.6522	0.0	0.5	0.6622	nan
0.1195	13.0	195	0.2385	0.6113	0.6113	0.4618	0.4618	0.4704	0.4704	0.5652	0.0	0.5	0.5693	nan
0.1046	14.0	210	0.2296	0.5997	0.5997	0.4544	0.4544	0.4903	0.4903	0.6522	0.0	0.5	0.6622	nan
0.089	15.0	225	0.2520	0.6283	0.6283	0.4974	0.4974	0.4404	0.4404	0.6522	0.0	0.5	0.6622	nan
0.083	16.0	240	0.2297	0.5998	0.5998	0.4635	0.4635	0.4901	0.4901	0.5652	0.0	0.5	0.5610	nan
0.0701	17.0	255	0.2207	0.5879	0.5879	0.4442	0.4442	0.5101	0.5101	0.6522	0.0	0.5	0.6622	nan
0.0585	18.0	270	0.2397	0.6128	0.6128	0.4617	0.4617	0.4678	0.4678	0.6522	0.0	0.5	0.6622	nan
0.0652	19.0	285	0.2284	0.5981	0.5981	0.4449	0.4449	0.4929	0.4929	0.6522	0.0	0.5	0.6622	nan
0.059	20.0	300	0.2491	0.6247	0.6247	0.4599	0.4599	0.4469	0.4469	0.6522	0.0	0.5	0.6622	nan
0.0464	21.0	315	0.2306	0.6010	0.6010	0.4373	0.4373	0.4880	0.4880	0.6522	0.0	0.5	0.6622	nan
0.0529	22.0	330	0.2370	0.6093	0.6093	0.4480	0.4480	0.4738	0.4738	0.6522	0.0	0.5	0.6622	nan
0.0555	23.0	345	0.2361	0.6082	0.6082	0.4474	0.4474	0.4757	0.4757	0.6522	0.0	0.5	0.6622	nan
0.0447	24.0	360	0.2283	0.5980	0.5980	0.4399	0.4399	0.4932	0.4932	0.6522	0.0	0.5	0.6622	nan
0.046	25.0	375	0.2259	0.5948	0.5948	0.4413	0.4413	0.4985	0.4985	0.6522	0.0	0.5	0.6622	nan
0.0379	26.0	390	0.2263	0.5953	0.5953	0.4402	0.4402	0.4977	0.4977	0.6522	0.0	0.5	0.6622	nan
0.0438	27.0	405	0.2270	0.5963	0.5963	0.4378	0.4378	0.4961	0.4961	0.6522	0.0	0.5	0.6622	nan
0.0354	28.0	420	0.2211	0.5886	0.5886	0.4379	0.4379	0.5090	0.5090	0.6522	0.0	0.5	0.6622	nan
0.0363	29.0	435	0.2269	0.5962	0.5962	0.4362	0.4362	0.4961	0.4961	0.6522	0.0	0.5	0.6622	nan
0.0451	30.0	450	0.2271	0.5965	0.5965	0.4372	0.4372	0.4957	0.4957	0.6522	0.0	0.5	0.6622	nan

Framework versions

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