zephyr-7b-dpo-lora-pubmedqa-selfgen-ultrafeedback-old

This model is a fine-tuned version of EllieS/zephyr-7b-dpo-lora-pubmedqa-selfgen-old on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:

• Loss: 0.5601
• Rewards/chosen: -0.4096
• Rewards/rejected: -0.9163
• Rewards/accuracies: 0.7000
• Rewards/margins: 0.5067
• Logps/rejected: -348.5905
• Logps/chosen: -331.9001
• Logits/rejected: -1.5161
• Logits/chosen: -1.5640

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-06
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• total_eval_batch_size: 4
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• 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
0.5074	0.39	3000	0.5837	-0.4893	-0.8886	0.6940	0.3993	-345.8194	-339.8665	-1.8683	-1.8925
0.5332	0.79	6000	0.5604	-0.4169	-0.9240	0.7040	0.5071	-349.3569	-332.6285	-1.5156	-1.5626

Framework versions

• PEFT 0.7.1
• Transformers 4.36.2
• Pytorch 2.1.2+cu121
• Datasets 2.14.6
• Tokenizers 0.15.2