zephyr-7b-dpo-full

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: 0.0629
• Rewards/chosen: -1.4532
• Rewards/rejected: -2.1723
• Rewards/accuracies: 0.7148
• Rewards/margins: 0.7191
• Logps/rejected: -474.5827
• Logps/chosen: -402.3561
• Logits/rejected: -2.3616
• Logits/chosen: -2.3768

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: 1
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 2
• total_train_batch_size: 128
• total_eval_batch_size: 64
• 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
0.1256	0.21	100	0.1140	-0.5631	-0.9563	0.6953	0.3933	-352.9850	-313.3454	-2.7051	-2.7197
0.08	0.42	200	0.0859	-0.8468	-1.4618	0.7266	0.6149	-403.5297	-341.7239	-2.5427	-2.5568
0.0612	0.63	300	0.0656	-1.3631	-2.0381	0.7070	0.6750	-461.1611	-393.3499	-2.4106	-2.4258
0.0676	0.84	400	0.0629	-1.4532	-2.1723	0.7148	0.7191	-474.5827	-402.3561	-2.3616	-2.3768

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.2+cu121
• Datasets 2.14.6
• Tokenizers 0.14.1