zephyr-7b-gpo-update3-i1

This model is a fine-tuned version of DUAL-GPO/zephyr-7b-gpo-update3-i0 on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:

• Loss: 0.0574
• Rewards/chosen: -0.0003
• Rewards/rejected: 0.0038
• Rewards/accuracies: 0.3765
• Rewards/margins: -0.0041
• Logps/rejected: -254.1840
• Logps/chosen: -266.7596
• Logits/rejected: -1.8151
• Logits/chosen: -1.9709

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
• gradient_accumulation_steps: 2
• total_train_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
0.0013	0.4	100	0.0537	0.0	0.0	0.0	0.0	-254.9398	-266.6976	-1.8067	-1.9618
0.0013	0.8	200	0.0575	-0.0013	0.0029	0.3800	-0.0041	-254.3691	-266.9557	-1.8139	-1.9695

Framework versions

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