zephyr-7b-gpo-gen-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: 0.0550
• Rewards/chosen: -0.0251
• Rewards/rejected: -0.0231
• Rewards/accuracies: 0.3875
• Rewards/margins: -0.0020
• Logps/rejected: -278.0226
• Logps/chosen: -291.8019
• Logits/rejected: -1.7909
• Logits/chosen: -1.9487

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
0.3754	0.08	100	0.0537	0.0	0.0	0.0	0.0	-254.9398	-266.6976	-1.8067	-1.9618
0.3556	0.16	200	0.0537	0.0	0.0	0.0	0.0	-254.9398	-266.6976	-1.8067	-1.9618
0.3556	0.24	300	0.0537	0.0	0.0	0.0	0.0	-254.9398	-266.6976	-1.8067	-1.9618
0.3606	0.32	400	0.0537	0.0	0.0	0.0	0.0	-254.9398	-266.6976	-1.8067	-1.9618
0.3606	0.4	500	0.0586	-0.0343	-0.0263	0.3125	-0.0079	-281.2869	-300.9843	-1.7627	-1.9202
0.3408	0.48	600	0.0587	-0.0387	-0.0304	0.3120	-0.0083	-285.3777	-305.4413	-1.7361	-1.8917
0.3359	0.56	700	0.0587	-0.0387	-0.0304	0.3095	-0.0083	-285.3294	-305.3720	-1.7363	-1.8920
0.3507	0.64	800	0.0569	-0.0251	-0.0199	0.3215	-0.0052	-274.8357	-291.8357	-1.8172	-1.9784
0.3926	0.72	900	0.0550	-0.0245	-0.0224	0.3840	-0.0021	-277.3842	-291.2067	-1.7982	-1.9565
0.3655	0.8	1000	0.0549	-0.0254	-0.0235	0.3860	-0.0019	-278.4594	-292.0937	-1.7905	-1.9482
0.3682	0.88	1100	0.0549	-0.0253	-0.0234	0.3850	-0.0020	-278.3317	-292.0442	-1.7919	-1.9497
0.3531	0.96	1200	0.0550	-0.0251	-0.0231	0.3910	-0.0020	-278.0787	-291.8378	-1.7915	-1.9493

Framework versions

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