Llama-3.1-8B-Instruct-SAA-400

This model is a fine-tuned version of meta-llama/Llama-3.1-8B-Instruct on the bct_non_cot_dpo_400 dataset. It achieves the following results on the evaluation set:

• Loss: 0.1181
• Rewards/chosen: -0.0081
• Rewards/rejected: -0.0551
• Rewards/accuracies: 0.8000
• Rewards/margins: 0.0470
• Logps/rejected: -0.5510
• Logps/chosen: -0.0815
• Logits/rejected: -0.3587
• Logits/chosen: -0.3145
• Sft Loss: 0.0122
• Odds Ratio Loss: 1.0589

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-06
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 10.0

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen	Sft Loss	Odds Ratio Loss
1.0959	2.2222	50	0.8812	-0.0833	-0.1278	0.8000	0.0444	-1.2775	-0.8334	-0.4147	-0.3542	0.0993	7.8184
0.2448	4.4444	100	0.1774	-0.0141	-0.0615	0.8000	0.0474	-0.6147	-0.1406	-0.3962	-0.3451	0.0187	1.5871
0.1229	6.6667	150	0.1202	-0.0083	-0.0555	0.7750	0.0472	-0.5554	-0.0834	-0.3636	-0.3187	0.0124	1.0785
0.1265	8.8889	200	0.1181	-0.0081	-0.0551	0.8000	0.0470	-0.5510	-0.0815	-0.3587	-0.3145	0.0122	1.0589

Framework versions

• PEFT 0.12.0
• Transformers 4.45.2
• Pytorch 2.3.0
• Datasets 2.19.0
• Tokenizers 0.20.0