zephyr-7b-dpo-full-gpt_consistent-3

This model is a fine-tuned version of alignment-handbook/zephyr-7b-sft-full on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4898
• Rewards/chosen: -0.9477
• Rewards/rejected: -1.9210
• Rewards/accuracies: 0.7543
• Rewards/margins: 0.9733
• Logps/rejected: -438.6234
• Logps/chosen: -379.8649
• Logits/rejected: 2.2425
• Logits/chosen: 0.9259

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-07
• train_batch_size: 8
• eval_batch_size: 8
• seed: 55
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 2
• total_train_batch_size: 128
• total_eval_batch_size: 64
• 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	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.5718	0.2294	100	0.5525	-0.3811	-0.9418	0.6940	0.5607	-340.7015	-323.2031	-1.1398	-1.3666
0.5281	0.4587	200	0.5102	-0.6195	-1.4237	0.7457	0.8042	-388.8913	-347.0378	1.0061	-0.1205
0.5213	0.6881	300	0.4939	-0.9485	-1.8771	0.7543	0.9286	-434.2354	-379.9385	1.7281	0.3985
0.5037	0.9174	400	0.4898	-0.9477	-1.9210	0.7543	0.9733	-438.6234	-379.8649	2.2425	0.9259

Framework versions

• Transformers 4.44.0.dev0
• Pytorch 2.1.2
• Datasets 2.20.0
• Tokenizers 0.19.1