zephyr-7b-dpo-full

This model is a fine-tuned version of alignment-handbook/zephyr-7b-sft-full on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:

• Loss: 0.8483
• Rewards/chosen: 0.7019
• Rewards/rejected: -1.3716
• Rewards/accuracies: 0.6786
• Rewards/margins: 2.0734
• Logps/rejected: -263.1501
• Logps/chosen: -283.4079
• Logits/rejected: -2.6313
• Logits/chosen: -2.6617

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: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• total_eval_batch_size: 32
• 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
0.9481	0.42	200	0.9533	-0.6644	-2.6641	0.7163	1.9997	-264.4427	-284.7741	-2.6754	-2.7043
1.0499	0.84	400	0.8542	0.5163	-1.5982	0.6766	2.1145	-263.3767	-283.5935	-2.6278	-2.6586

Framework versions

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