zephyr-7b-dpo-maximal

This model is a fine-tuned version of alignment-handbook/zephyr-7b-sft-full on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3380
• Rewards/chosen: -0.1339
• Rewards/rejected: -3.0976
• Rewards/accuracies: 0.8790
• Rewards/margins: 2.9637
• Logps/rejected: -275.9525
• Logps/chosen: -285.9466
• Logits/rejected: -2.1375
• Logits/chosen: -2.2908

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
• total_train_batch_size: 32
• total_eval_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• 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
0.3619	0.26	500	0.3822	0.1843	-2.0970	0.8651	2.2812	-265.9466	-282.7652	-2.1994	-2.3618
0.396	0.52	1000	0.3747	-0.7559	-3.2293	0.8730	2.4733	-277.2696	-292.1672	-2.1335	-2.2927
0.3618	0.78	1500	0.3452	-0.4962	-3.2836	0.875	2.7874	-277.8134	-289.5698	-2.1794	-2.3280

Framework versions

• Transformers 4.36.1
• Pytorch 2.0.1+cu117
• Datasets 2.16.1
• Tokenizers 0.15.0