test2

This model is a fine-tuned version of TheBloke/OpenHermes-2-Mistral-7B-GPTQ on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2724
• Rewards/chosen: 0.0617
• Rewards/rejected: -1.6629
• Rewards/accuracies: 1.0
• Rewards/margins: 1.7245
• Logps/rejected: -396.0331
• Logps/chosen: -110.0373
• Logits/rejected: -3.2686
• Logits/chosen: -3.3982

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: 0.0002
• train_batch_size: 1
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 2
• training_steps: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6305	0.01	10	0.4258	0.0681	-0.7348	1.0	0.8029	-386.7522	-109.9732	-3.2678	-3.4032
0.3418	0.02	20	0.2724	0.0617	-1.6629	1.0	1.7245	-396.0331	-110.0373	-3.2686	-3.3982

Framework versions

• PEFT 0.9.0
• Transformers 4.38.2
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2