xlm-roberta-large-xnli-anli-v4.0

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

• Loss: 0.5122
• F1 Macro: 0.8080
• F1 Micro: 0.8090
• Accuracy Balanced: 0.8085
• Accuracy: 0.8090
• Precision Macro: 0.8076
• Recall Macro: 0.8085
• Precision Micro: 0.8090
• Recall Micro: 0.8090

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: 9e-06
• train_batch_size: 8
• eval_batch_size: 64
• seed: 40
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.06
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	F1 Macro	F1 Micro	Accuracy Balanced	Accuracy	Precision Macro	Recall Macro	Precision Micro	Recall Micro
0.3655	1.69	200	0.4396	0.8166	0.8171	0.8185	0.8171	0.8163	0.8185	0.8171	0.8171

eval result

Datasets	asadfgglie/nli-zh-tw-all/test	asadfgglie/BanBan_2024-10-17-facial_expressions-nli/test	eval_dataset	test_dataset
eval_loss	0.507	0.871	0.486	0.512
eval_f1_macro	0.81	0.575	0.82	0.808
eval_f1_micro	0.811	0.58	0.821	0.809
eval_accuracy_balanced	0.811	0.584	0.821	0.808
eval_accuracy	0.811	0.58	0.821	0.809
eval_precision_macro	0.81	0.59	0.819	0.808
eval_recall_macro	0.811	0.584	0.821	0.808
eval_precision_micro	0.811	0.58	0.821	0.809
eval_recall_micro	0.811	0.58	0.821	0.809
eval_runtime	50.981	0.641	10.403	40.014
eval_samples_per_second	166.729	1476.463	163.413	169.942
eval_steps_per_second	2.609	23.411	2.595	2.674
Size of dataset	8500	946	1700	6800

Framework versions

• Transformers 4.33.3
• Pytorch 2.5.1+cu121
• Datasets 2.14.7
• Tokenizers 0.13.3