zephyr-7b-dpo-lora-1

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: 0.6542
• Rewards/chosen: 0.0641
• Rewards/rejected: -0.0250
• Rewards/accuracies: 0.6765
• Rewards/margins: 0.0891
• Logps/rejected: -222.6298
• Logps/chosen: -264.3456
• Logits/rejected: -2.1600
• Logits/chosen: -2.2261

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: 2
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• gradient_accumulation_steps: 32
• total_train_batch_size: 64
• 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
0.6768	0.99	96	0.6764	0.0306	-0.0056	0.6380	0.0363	-222.4367	-264.6802	-2.1612	-2.2277
0.6591	1.99	193	0.6596	0.0553	-0.0203	0.6775	0.0756	-222.5832	-264.4333	-2.1606	-2.2268
0.6511	2.97	288	0.6542	0.0641	-0.0250	0.6765	0.0891	-222.6298	-264.3456	-2.1600	-2.2261

Framework versions

• Transformers 4.35.0
• Pytorch 2.1.0+cu121
• Datasets 2.14.6
• Tokenizers 0.14.1