orpo-phi3

This model is a fine-tuned version of microsoft/Phi-3-mini-4k-instruct on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.2224
• Rewards/chosen: -0.1152
• Rewards/rejected: -0.1838
• Rewards/accuracies: 1.0
• Rewards/margins: 0.0685
• Logps/rejected: -1.8375
• Logps/chosen: -1.1521
• Logits/rejected: -1.5266
• Logits/chosen: -0.4575
• Nll Loss: 1.1880
• Log Odds Ratio: -0.3436
• Log Odds Chosen: 0.8917

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: 8e-06
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 10
• 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	Nll Loss	Log Odds Ratio	Log Odds Chosen
4.0204	0.24	3	1.2224	-0.1152	-0.1838	1.0	0.0685	-1.8375	-1.1521	-1.5266	-0.4575	1.1880	-0.3436	0.8917
2.2028	0.48	6	1.2224	-0.1152	-0.1838	1.0	0.0685	-1.8375	-1.1521	-1.5266	-0.4575	1.1880	-0.3436	0.8917
4.7852	0.72	9	1.2224	-0.1152	-0.1838	1.0	0.0685	-1.8375	-1.1521	-1.5266	-0.4575	1.1880	-0.3436	0.8917
3.1295	0.96	12	1.2224	-0.1152	-0.1838	1.0	0.0685	-1.8375	-1.1521	-1.5266	-0.4575	1.1880	-0.3436	0.8917

Framework versions

• PEFT 0.11.1
• Transformers 4.41.1
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1