mistral-7b-dpo-full-wo-kqa_silver_wogold-ep3

This model is a fine-tuned version of mistralai/Mistral-7B-v0.1 on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:

• Loss: 0.2946
• Rewards/chosen: -2.2047
• Rewards/rejected: -5.9058
• Rewards/accuracies: 0.8793
• Rewards/margins: 3.7012
• Logps/rejected: -1521.8148
• Logps/chosen: -750.3969
• Logits/rejected: -3.0916
• Logits/chosen: -3.2432

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: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 2
• 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: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 1

Training results

Training Loss	Epoch	Step	Logits/chosen	Logits/rejected	Logps/chosen	Logps/rejected	Validation Loss	Rewards/accuracies	Rewards/chosen	Rewards/margins	Rewards/rejected
0.1362	0.65	100	-3.2790	-3.1123	-721.5319	-1477.7874	0.3206	0.8707	-1.9160	3.5495	-5.4655

Framework versions

• Transformers 4.39.0.dev0
• Pytorch 2.1.2
• Datasets 2.14.6
• Tokenizers 0.15.2