USS-reward-model-exp6-grl-only

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: 4.4503
• Mse: 0.1857
• Mae: 0.3335
• R2: 0.0374
• Spearman Correlation: 0.2772

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
11.3890	1.0	97	0.1401	0.1999	0.3675	-0.0365	-0.0219
1.5946	2.0	194	0.1645	0.2008	0.3874	-0.0410	0.1877
12.8941	3.0	291	3.9913	1.2837	1.0346	-5.6549	0.1595
47.4786	4.0	388	5.2619	0.3131	0.4203	-0.6231	0.1512
52.6400	5.0	485	5.2644	0.6928	0.7324	-2.5916	0.2834
47.2109	6.0	582	4.4503	0.1857	0.3335	0.0374	0.2772
40.5784	7.0	679	3.8972	0.3032	0.4734	-0.5721	0.2260
34.6866	8.0	776	3.3503	0.2189	0.3854	-0.1347	0.3143
30.7190	9.0	873	3.0725	0.2011	0.3632	-0.0425	0.2707
28.8300	10.0	970	2.9724	0.1929	0.3394	-0.0003	0.2642

Framework versions

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