zephyr-dpo-qlora-uf6k-5e-6

This model is a fine-tuned version of alignment-handbook/zephyr-7b-sft-full on the generation/UF6konly dataset. It achieves the following results on the evaluation set:

• Loss: 0.6104
• Rewards/chosen: -0.1894
• Rewards/rejected: -0.4534
• Rewards/accuracies: 0.6870
• Rewards/margins: 0.2640
• Rewards/margins Max: 1.0323
• Rewards/margins Min: -0.4642
• Rewards/margins Std: 0.5046
• Logps/rejected: -303.9143
• Logps/chosen: -303.5331
• Logits/rejected: -2.7412
• Logits/chosen: -2.7732

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: 4
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• total_eval_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: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Rewards/margins Max	Rewards/margins Min	Rewards/margins Std	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6517	0.3	100	0.6560	-0.0102	-0.1055	0.6900	0.0954	0.4105	-0.2027	0.2020	-269.1304	-285.6097	-2.7441	-2.7791
0.5795	0.61	200	0.6177	-0.2139	-0.4530	0.6820	0.2391	0.9738	-0.4425	0.4777	-303.8808	-305.9859	-2.7382	-2.7701
0.5791	0.91	300	0.6103	-0.2010	-0.4650	0.6860	0.2641	1.0353	-0.4649	0.5062	-305.0838	-304.6903	-2.7402	-2.7722

Framework versions

• PEFT 0.7.1
• Transformers 4.39.0.dev0
• Pytorch 2.1.2+cu121
• Datasets 2.14.6
• Tokenizers 0.15.2