zephyr-7b-dpo-lora

This model is a fine-tuned version of mistralai/Mistral-7B-v0.1 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5381
• Rewards/chosen: -0.0977
• Rewards/rejected: -0.7746
• Rewards/accuracies: 0.7183
• Rewards/margins: 0.6768
• Logps/rejected: -237.3503
• Logps/chosen: -283.6626
• Logits/rejected: -1.8216
• Logits/chosen: -1.9102

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: 16
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• 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: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.5546	1.0	968	0.5678	-0.0091	-0.4982	0.7054	0.4891	-234.5862	-282.7758	-1.8508	-1.9394
0.5491	2.0	1936	0.5438	-0.0836	-0.7251	0.7192	0.6414	-236.8553	-283.5217	-1.8279	-1.9162
0.5463	3.0	2904	0.5381	-0.0977	-0.7746	0.7183	0.6768	-237.3503	-283.6626	-1.8216	-1.9102

Framework versions

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