unraveled-7b-dpo-lora

This model is a fine-tuned version of NousResearch/Yarn-Mistral-7b-128k on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5895
• Rewards/chosen: 0.1439
• Rewards/rejected: -0.1833
• Rewards/accuracies: 0.6880
• Rewards/margins: 0.3272
• Logps/rejected: -221.8329
• Logps/chosen: -266.1414
• Logits/rejected: -1.9675
• Logits/chosen: -2.0859

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: 4
• gradient_accumulation_steps: 32
• total_train_batch_size: 256
• total_eval_batch_size: 16
• 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
0.6313	1.0	242	0.6318	0.1228	-0.0304	0.6600	0.1532	-220.3036	-266.3521	-1.9863	-2.1062
0.6013	2.0	484	0.5983	0.1484	-0.1334	0.6760	0.2819	-221.3338	-266.0959	-1.9723	-2.0914
0.5889	3.0	726	0.5895	0.1439	-0.1833	0.6880	0.3272	-221.8329	-266.1414	-1.9675	-2.0859

Framework versions

• Transformers 4.35.0
• Pytorch 2.1.0+cu118
• Datasets 2.14.6
• Tokenizers 0.14.1