nash_dpo_doff_no_golden_iter_1

This model is a fine-tuned version of alignment-handbook/zephyr-7b-sft-full on the updated and the original datasets. It achieves the following results on the evaluation set:

• Logits/chosen: -2.7444
• Logits/rejected: -2.6646
• Logps/chosen: -288.1525
• Logps/rejected: -266.5563
• Loss: 0.6749
• Rewards/accuracies: 0.6100
• Rewards/chosen: -0.0411
• Rewards/margins: 0.0493
• Rewards/rejected: -0.0904

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
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 16
• total_train_batch_size: 128
• total_eval_batch_size: 8
• 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	Logits/chosen	Logits/rejected	Logps/chosen	Logps/rejected	Validation Loss	Rewards/accuracies	Rewards/chosen	Rewards/margins	Rewards/rejected
0.6794	0.65	100	-2.7444	-2.6646	-288.1525	-266.5563	0.6749	0.6100	-0.0411	0.0493	-0.0904

Framework versions

• PEFT 0.7.1
• Transformers 4.36.2
• Pytorch 2.1.2+cu121
• Datasets 2.14.6
• Tokenizers 0.15.2