xlmroberta_mixup_back

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

• Loss: 0.0611
• Spearman Corr: 0.2466

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: 36
• eval_batch_size: 144
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 72
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Spearman Corr
No log	0.11	20	0.0701	-0.0323
No log	0.21	40	0.0613	0.0336
No log	0.32	60	0.0755	0.0627
No log	0.43	80	0.1200	0.1902
No log	0.53	100	0.0961	0.1906
No log	0.64	120	0.0973	0.1694
No log	0.75	140	0.0918	0.2296
No log	0.86	160	0.0614	0.2128
No log	0.96	180	0.0607	0.2090
No log	1.07	200	0.0581	0.2252
No log	1.18	220	0.0651	0.2340
No log	1.28	240	0.0556	0.2590
No log	1.39	260	0.1081	0.2593
No log	1.5	280	0.0724	0.2920
No log	1.6	300	0.1213	0.2862
No log	1.71	320	0.0617	0.2879
No log	1.82	340	0.0774	0.2620
No log	1.93	360	0.0571	0.2518
0.0972	2.03	380	0.0565	0.2471
0.0972	2.14	400	0.0531	0.2608
0.0972	2.25	420	0.0728	0.2815
0.0972	2.35	440	0.0733	0.2658
0.0972	2.46	460	0.0672	0.2942
0.0972	2.57	480	0.2626	0.2995
0.0972	2.67	500	0.0733	0.3036
0.0972	2.78	520	0.0647	0.2741
0.0972	2.89	540	0.0594	0.2604
0.0972	2.99	560	0.0594	0.2439
0.0972	3.1	580	0.0685	0.2463
0.0972	3.21	600	0.0779	0.2668
0.0972	3.32	620	0.0559	0.2760
0.0972	3.42	640	0.0845	0.2826
0.0972	3.53	660	0.1213	0.2937
0.0972	3.64	680	0.0748	0.2798
0.0972	3.74	700	0.0835	0.2542
0.0972	3.85	720	0.0565	0.2417
0.0972	3.96	740	0.0763	0.2658
0.064	4.06	760	0.0750	0.2533
0.064	4.17	780	0.0611	0.2466

Framework versions

• Transformers 4.37.2
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.1