aligner-v1-llama3-01

This model is a fine-tuned version of meta-llama/Meta-Llama-3-8B on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.4361
• Rewards/chosen: -0.0331
• Rewards/rejected: -0.0349
• Rewards/accuracies: 0.8333
• Rewards/margins: 0.0018
• Logps/rejected: -0.3493
• Logps/chosen: -0.3313
• Logits/rejected: -1.5592
• Logits/chosen: -1.5485
• Nll Loss: 1.3699
• Log Odds Ratio: -0.6618
• Log Odds Chosen: 0.0646

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
2.8628	0.2105	15	2.7068	-0.1219	-0.1183	0.0	-0.0036	-1.1830	-1.2191	-1.8565	-1.8340	2.6349	-0.7190	-0.0509
2.1044	0.4211	30	2.0553	-0.0702	-0.0687	0.1667	-0.0015	-0.6871	-0.7024	-1.6352	-1.6218	1.9845	-0.7082	-0.0296
1.6915	0.6316	45	1.6323	-0.0431	-0.0436	0.8333	0.0006	-0.4364	-0.4305	-1.6833	-1.6715	1.5639	-0.6842	0.0185
1.4279	0.8421	60	1.4361	-0.0331	-0.0349	0.8333	0.0018	-0.3493	-0.3313	-1.5592	-1.5485	1.3699	-0.6618	0.0646

Framework versions

• PEFT 0.10.0
• Transformers 4.40.1
• Pytorch 2.2.1+cu121
• Datasets 2.19.0
• Tokenizers 0.19.1