USS-reward-model-WRS_alpha0.5

This model is a fine-tuned version of answerdotai/ModernBERT-large on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1442
• Mse: 0.2014
• Mae: 0.3569
• R2: -0.0442
• Spearman Correlation: 0.3045

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: 5e-05
• train_batch_size: 2
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• gradient_accumulation_steps: 10
• total_train_batch_size: 20
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Mse	Mae	R2	Spearman Correlation
153.0692	1.0	97	0.2201	1.4526	1.1032	-6.5307	0.2503
9.3095	2.0	194	0.1442	0.2014	0.3569	-0.0442	0.3045
4.9540	3.0	291	0.1609	0.2101	0.3679	-0.0891	0.2504
4.3671	4.0	388	0.1331	0.2639	0.4187	-0.3682	0.3116
2.4743	5.0	485	0.2161	0.2424	0.3931	-0.2565	0.2496
1.6148	6.0	582	0.1284	0.4024	0.5261	-1.0864	0.2314
0.7143	7.0	679	0.1154	0.2833	0.4374	-0.4686	0.2660
0.4222	8.0	776	0.1248	0.2246	0.3798	-0.1644	0.2440
0.3196	9.0	873	0.1145	0.2390	0.3950	-0.2390	0.2442
0.1041	10.0	970	0.1226	0.2381	0.3921	-0.2345	0.2592

Framework versions

• Transformers 5.9.0
• Pytorch 2.12.0+cu130
• Datasets 4.8.5
• Tokenizers 0.22.2