zephyr-7b-dpo-lora

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

• Loss: 0.3553
• Rewards/chosen: -0.8622
• Rewards/rejected: -3.1235
• Rewards/accuracies: 0.8281
• Rewards/margins: 2.2613
• Logps/rejected: -204.2707
• Logps/chosen: -282.4587
• Logits/rejected: -2.6699
• Logits/chosen: -2.7156

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: 2e-06
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• total_train_batch_size: 128
• total_eval_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.2024	1.0	485	0.4197	-0.3974	-1.8930	0.8086	1.4956	-191.9660	-277.8107	-2.7272	-2.7680
0.1305	2.0	970	0.3694	-0.7584	-2.8597	0.8242	2.1013	-201.6330	-281.4208	-2.6866	-2.7306
0.109	3.0	1455	0.3553	-0.8622	-3.1235	0.8281	2.2613	-204.2707	-282.4587	-2.6699	-2.7156

Framework versions

• Transformers 4.35.0
• Pytorch 2.1.1+cu121
• Datasets 2.14.6
• Tokenizers 0.14.1