USS-reward-model-grl

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: 4.3828
• Mse: 0.3399
• Mae: 0.4563
• R2: -0.7622
• Spearman Correlation: 0.1634

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
134.0139	1.0	97	0.2249	0.5842	0.6534	-2.0289	0.2181
16.3257	2.0	194	0.2289	0.7627	0.6692	-2.9542	0.1503
5.1921	3.0	291	0.1993	0.4319	0.5319	-1.2392	0.1107
3.6719	4.0	388	0.2842	0.4943	0.5859	-1.5625	0.1886
8.7617	5.0	485	3.2637	0.7597	0.7331	-2.9384	0.2087
49.3081	6.0	582	5.8898	1.9898	1.3009	-9.3156	0.2507
53.3273	7.0	679	5.3314	0.4956	0.5869	-1.5694	0.2168
48.4105	8.0	776	4.8294	0.8525	0.8221	-3.4196	0.1920
45.6567	9.0	873	4.4777	0.4814	0.5543	-1.4957	0.1756
47.8308	10.0	970	4.3828	0.3399	0.4563	-0.7622	0.1634

Framework versions

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