zephyr-7b-dpo-lora

This model is a fine-tuned version of mistralai/Mistral-7B-v0.1 on the None dataset. It achieves the following results on the evaluation set:

• Loss: -7.5416
• Rewards/chosen: -29.9194
• Rewards/rejected: -39.8539
• Rewards/accuracies: 0.6151
• Rewards/margins: 9.9345
• Logps/rejected: -633.4722
• Logps/chosen: -588.9970
• Logits/rejected: -1.0751
• Logits/chosen: -1.2630

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-07
• train_batch_size: 8
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• total_train_batch_size: 64
• total_eval_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
-0.2007	1.0	969	-0.0988	-1.1416	-2.4993	0.6746	1.3577	-259.9262	-301.2188	-1.9876	-2.0976
-2.3739	2.0	1938	-3.0140	-12.9185	-17.8885	0.6587	4.9699	-413.8172	-418.9880	-1.4397	-1.5909
-5.7169	3.0	2907	-7.5416	-29.9194	-39.8539	0.6151	9.9345	-633.4722	-588.9970	-1.0751	-1.2630

Framework versions

• Transformers 4.35.0
• Pytorch 2.1.1+cu121
• Datasets 2.14.6
• Tokenizers 0.14.1