xlm-roberta-base_arb_corr_2e-05

This model is a fine-tuned version of FacebookAI/xlm-roberta-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0228
• Spearman Corr: 0.7757

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: 32
• eval_batch_size: 128
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Spearman Corr
No log	0.85	200	0.0226	0.7708
No log	1.69	400	0.0232	0.7707
0.0026	2.54	600	0.0237	0.7725
0.0026	3.38	800	0.0242	0.7723
0.0025	4.23	1000	0.0240	0.7696
0.0025	5.07	1200	0.0243	0.7682
0.0025	5.92	1400	0.0239	0.7711
0.0022	6.77	1600	0.0230	0.7693
0.0022	7.61	1800	0.0228	0.7706
0.0019	8.46	2000	0.0234	0.7720
0.0019	9.3	2200	0.0234	0.7747
0.0018	10.15	2400	0.0226	0.7689
0.0018	10.99	2600	0.0219	0.7703
0.0018	11.84	2800	0.0237	0.7730
0.002	12.68	3000	0.0235	0.7675
0.002	13.53	3200	0.0238	0.7741
0.0018	14.38	3400	0.0226	0.7782
0.0018	15.22	3600	0.0221	0.7741
0.0017	16.07	3800	0.0230	0.7728
0.0017	16.91	4000	0.0227	0.7749
0.0017	17.76	4200	0.0229	0.7762
0.0015	18.6	4400	0.0224	0.7755
0.0015	19.45	4600	0.0242	0.7781
0.0014	20.3	4800	0.0228	0.7761
0.0014	21.14	5000	0.0228	0.7757

Framework versions

• Transformers 4.37.2
• Pytorch 2.2.0+cu121
• Datasets 2.17.0
• Tokenizers 0.15.2