USS-reward-model-weighted

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: 0.0176
• Mse: 0.1932
• Mae: 0.3318
• R2: -0.0014
• Spearman Correlation: 0.2246

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: 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
0.4138	1.0	97	0.0174	0.1963	0.3690	-0.0177	0.1412
0.2178	2.0	194	0.0105	0.2337	0.3839	-0.2114	0.3311
0.0998	3.0	291	0.0114	0.2363	0.3869	-0.2250	0.3403
0.0549	4.0	388	0.0115	0.2234	0.3868	-0.1580	0.2939
0.0461	5.0	485	0.0186	0.2199	0.3629	-0.1401	0.2499
0.0267	6.0	582	0.0151	0.2173	0.3714	-0.1264	0.3104
0.0170	7.0	679	0.0159	0.2028	0.3450	-0.0514	0.3342
0.0127	8.0	776	0.0178	0.1956	0.3362	-0.0142	0.1951
0.0071	9.0	873	0.0170	0.1941	0.3347	-0.0060	0.2219
0.0022	10.0	970	0.0176	0.1932	0.3318	-0.0014	0.2246

Framework versions

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