Roberta_combo_1.6lakh

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: 1.4440
• Accuracy: 0.62
• Auc: 0.608
• Precision: 0.581
• Recall: 0.814
• F1: 0.678
• F1-macro: 0.608
• F1-micro: 0.62
• F1-weighted: 0.607

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 2
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Auc	Precision	Recall	F1	F1-macro	F1-micro	F1-weighted
0.5271	0.1935	1000	1.0367	0.595	0.57	0.567	0.751	0.646	0.587	0.595	0.586
0.439	0.3870	2000	1.2336	0.594	0.576	0.558	0.832	0.668	0.573	0.594	0.571
0.4166	0.5805	3000	1.3620	0.606	0.57	0.569	0.814	0.67	0.591	0.606	0.59
0.3997	0.7740	4000	1.3161	0.61	0.59	0.575	0.787	0.664	0.599	0.61	0.598
0.391	0.9675	5000	1.3359	0.613	0.596	0.573	0.833	0.679	0.596	0.613	0.595
0.3618	1.1610	6000	1.4006	0.617	0.603	0.578	0.817	0.677	0.603	0.617	0.602
0.3564	1.3545	7000	1.3645	0.619	0.612	0.584	0.782	0.669	0.611	0.619	0.61
0.352	1.5480	8000	1.4672	0.617	0.601	0.582	0.785	0.668	0.608	0.617	0.606
0.3465	1.7415	9000	1.4327	0.62	0.609	0.586	0.771	0.666	0.613	0.62	0.612
0.3413	1.9350	10000	1.4440	0.62	0.608	0.581	0.814	0.678	0.608	0.62	0.607

Framework versions

• Transformers 4.55.2
• Pytorch 2.6.0+cu124
• Datasets 4.0.0
• Tokenizers 0.21.4