DrishtiSharma/dolphin-2.1-mistral-7b-dpo-ultrafeedback-binarized-preferences-ipo

doplhin-2.1-mistral-7b-dpo-ultrafeedback-binarized-preferences-ipo

This model is a fine-tuned version of cognitivecomputations/dolphin-2.1-mistral-7b on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 13.6404
• Rewards/chosen: -0.4693
• Rewards/rejected: -0.7026
• Rewards/accuracies: 0.8234
• Rewards/margins: 0.2333
• Logps/rejected: -9.0933
• Logps/chosen: -6.2746
• Logits/rejected: -0.8214
• Logits/chosen: -0.8422

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-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• 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
17.7871	0.25	700	16.4082	-0.2243	-0.3706	0.7903	0.1464	-5.7735	-3.8245	-1.8423	-1.8837
13.4212	0.51	1400	14.5490	-0.4924	-0.7383	0.8092	0.2459	-9.4501	-6.5058	-0.9174	-0.9510
13.2665	0.76	2100	13.6404	-0.4693	-0.7026	0.8234	0.2333	-9.0933	-6.2746	-0.8214	-0.8422

Framework versions

• PEFT 0.8.2
• Transformers 4.37.2
• Pytorch 2.1.2+cu121
• Datasets 2.16.1
• Tokenizers 0.15.2