model_hh_usp3_200

This model is a fine-tuned version of meta-llama/Llama-2-7b-chat-hf on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.8202
• Rewards/chosen: -6.0974
• Rewards/rejected: -8.9863
• Rewards/accuracies: 0.6300
• Rewards/margins: 2.8889
• Logps/rejected: -120.5816
• Logps/chosen: -117.1597
• Logits/rejected: -0.0796
• Logits/chosen: -0.0123

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: 0.0005
• train_batch_size: 4
• eval_batch_size: 1
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 100
• training_steps: 1000

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.0	8.0	100	1.8186	-5.9435	-8.8369	0.6300	2.8934	-120.4156	-116.9887	-0.0816	-0.0143
0.0	16.0	200	1.8025	-6.0024	-8.9192	0.6200	2.9168	-120.5071	-117.0542	-0.0812	-0.0142
0.0	24.0	300	1.7997	-6.0563	-8.9882	0.6300	2.9319	-120.5837	-117.1141	-0.0806	-0.0133
0.0	32.0	400	1.8278	-6.0899	-8.9796	0.6300	2.8898	-120.5742	-117.1513	-0.0794	-0.0122
0.0	40.0	500	1.8304	-6.1315	-9.0197	0.6300	2.8882	-120.6187	-117.1977	-0.0796	-0.0128
0.0	48.0	600	1.8260	-6.0887	-8.9905	0.6300	2.9018	-120.5863	-117.1501	-0.0804	-0.0131
0.0	56.0	700	1.8303	-6.1106	-8.9877	0.6300	2.8771	-120.5832	-117.1744	-0.0801	-0.0132
0.0	64.0	800	1.8139	-6.0927	-8.9885	0.6300	2.8957	-120.5840	-117.1545	-0.0786	-0.0117
0.0	72.0	900	1.8263	-6.1113	-8.9822	0.6200	2.8709	-120.5770	-117.1752	-0.0789	-0.0120
0.0	80.0	1000	1.8202	-6.0974	-8.9863	0.6300	2.8889	-120.5816	-117.1597	-0.0796	-0.0123

Framework versions

• PEFT 0.10.0
• Transformers 4.39.1
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2