xlm-roberta-base_latin_kin-amh-eng_train_spearman_corr

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: 0.0290
• Spearman Corr: 0.7237

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	0.59	200	0.0279	0.6205
No log	1.17	400	0.0265	0.6575
No log	1.76	600	0.0380	0.6844
0.0373	2.35	800	0.0307	0.7164
0.0373	2.93	1000	0.0340	0.6776
0.0373	3.52	1200	0.0312	0.7050
0.0244	4.11	1400	0.0298	0.7134
0.0244	4.69	1600	0.0285	0.7190
0.0244	5.28	1800	0.0278	0.7217
0.0244	5.87	2000	0.0298	0.7006
0.0185	6.45	2200	0.0258	0.7114
0.0185	7.04	2400	0.0295	0.7164
0.0185	7.62	2600	0.0333	0.7204
0.0144	8.21	2800	0.0300	0.7201
0.0144	8.8	3000	0.0326	0.7290
0.0144	9.38	3200	0.0292	0.7114
0.0144	9.97	3400	0.0289	0.7052
0.0109	10.56	3600	0.0336	0.7125
0.0109	11.14	3800	0.0307	0.7147
0.0109	11.73	4000	0.0308	0.7174
0.0086	12.32	4200	0.0285	0.7217
0.0086	12.9	4400	0.0288	0.7083
0.0086	13.49	4600	0.0290	0.7237

Framework versions

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