Mistral-7B-Instruct-v0.3-dpo-lora_lr1e-5_2ep

This model is a fine-tuned version of mistralai/Mistral-7B-Instruct-v0.3 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3109
• Rewards/chosen: 0.7129
• Rewards/rejected: -1.3545
• Rewards/accuracies: 0.8795
• Rewards/margins: 2.0674
• Logps/rejected: -373.3019
• Logps/chosen: -422.7253
• Logits/rejected: -0.2832
• Logits/chosen: -0.1040

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: 1e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.4052	1.0	103	0.3249	0.5808	-1.2724	0.8795	1.8532	-372.4805	-424.0459	-0.3089	-0.1139
0.1566	2.0	206	0.3109	0.7129	-1.3545	0.8795	2.0674	-373.3019	-422.7253	-0.2832	-0.1040

Framework versions

• PEFT 0.11.1
• Transformers 4.41.1
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1