reward_modeling_anthropic_hh_rm1e-4

This model is a fine-tuned version of facebook/opt-350m on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6931
• Accuracy: 0.7339

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: 0.0001
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 1.0

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.7053	0.1087	500	0.6931	0.6148
0.6926	0.2174	1000	0.6931	0.6260
0.6912	0.3262	1500	0.6931	0.6737
0.6923	0.4349	2000	0.6931	0.6653
0.6946	0.5436	2500	0.6931	0.6698
0.6888	0.6523	3000	0.6931	0.6973
0.6963	0.7610	3500	0.6931	0.7138
0.689	0.8698	4000	0.6931	0.7124
0.6942	0.9785	4500	0.6931	0.7339

Framework versions

• Transformers 4.40.2
• Pytorch 2.4.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1