openhermes-mistral-dpo-gptq

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: 0.6599
• Rewards/chosen: 0.0397
• Rewards/rejected: -0.0752
• Rewards/accuracies: 0.9375
• Rewards/margins: 0.1149
• Logps/rejected: -164.5962
• Logps/chosen: -292.6904
• Logits/rejected: -2.6901
• Logits/chosen: -2.3670

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
• 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
0.6819	0.01	10	0.6600	0.0491	-0.0050	1.0	0.0540	-163.8940	-292.5971	-2.6930	-2.3675
0.7106	0.01	20	0.6787	0.0460	0.0162	0.5625	0.0298	-163.6827	-292.6277	-2.6971	-2.3713
0.6487	0.01	30	0.6889	0.0454	-0.0002	0.8125	0.0456	-163.8460	-292.6334	-2.6960	-2.3700
0.5981	0.02	40	0.6718	0.0307	-0.0583	0.9375	0.0890	-164.4272	-292.7806	-2.6928	-2.3685
0.6573	0.03	50	0.6599	0.0397	-0.0752	0.9375	0.1149	-164.5962	-292.6904	-2.6901	-2.3670

Framework versions

• Transformers 4.35.2
• Pytorch 2.0.1+cu117
• Datasets 2.17.0
• Tokenizers 0.15.1