zephyr-7b-dpo-lora

This model is a fine-tuned version of HuggingFaceH4/zephyr-7b-beta on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6843
• Rewards/chosen: 0.0440
• Rewards/rejected: 0.0071
• Rewards/accuracies: 0.5
• Rewards/margins: 0.0369
• Logps/rejected: -132.8740
• Logps/chosen: -190.5722
• Logits/rejected: -2.2999
• Logits/chosen: -2.2747

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: 2
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 32
• total_train_batch_size: 512
• 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: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6931	0.55	1	0.6931	0.0	0.0	0.0	0.0	-132.9451	-191.0126	-2.3015	-2.2762
0.6931	1.66	3	0.6928	0.0185	-0.0111	0.5	0.0296	-133.0566	-190.8279	-2.3016	-2.2755
0.6931	2.76	5	0.6843	0.0440	0.0071	0.5	0.0369	-132.8740	-190.5722	-2.2999	-2.2747

Framework versions

• Transformers 4.35.0
• Pytorch 2.1.0
• Datasets 2.14.6
• Tokenizers 0.14.1