mistral-dpo

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: 0.6931
• Rewards/chosen: 0.0156
• Rewards/rejected: 0.0156
• Rewards/accuracies: 0.0
• Rewards/margins: 0.0
• Logps/rejected: -165.4536
• Logps/chosen: -165.4536
• Logits/rejected: -1.7890
• Logits/chosen: -1.7890

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: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6931	0.01	10	0.6931	0.0246	0.0246	0.0	0.0	-165.3638	-165.3638	-1.7889	-1.7889
0.6931	0.01	20	0.6931	0.0133	0.0133	0.0	0.0	-165.4770	-165.4770	-1.7895	-1.7895
0.6931	0.01	30	0.6931	0.0060	0.0060	0.0	0.0	-165.5494	-165.5494	-1.7923	-1.7923
0.6931	0.02	40	0.6931	0.0063	0.0063	0.0	0.0	-165.5464	-165.5464	-1.7897	-1.7897
0.6931	0.03	50	0.6931	0.0156	0.0156	0.0	0.0	-165.4536	-165.4536	-1.7890	-1.7890

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0