Note from me: This is a fine tuned Phi-2 on Argilla-provided Intel Orca DPO pairs. It's run with the default settings, just with the batch sized at 2 instead of 1. The below was automatically generated by the trainer. It cost about $2.50 to train on RunPod.

phi2-lora-distilabel-intel-orca-dpo-pairs

This model is a fine-tuned version of microsoft/phi-2 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.4467
• Rewards/chosen: -0.0981
• Rewards/rejected: -1.3106
• Rewards/accuracies: 0.8410
• Rewards/margins: 1.2125
• Logps/rejected: -228.4777
• Logps/chosen: -209.0628
• Logits/rejected: 0.4528
• Logits/chosen: 0.2946

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: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 250
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.5578	0.78	250	0.4467	-0.0981	-1.3106	0.8410	1.2125	-228.4777	-209.0628	0.4528	0.2946

Framework versions

• PEFT 0.9.0
• Transformers 4.38.2
• Pytorch 2.1.0+cu118
• Datasets 2.18.0
• Tokenizers 0.15.2