nash_dpo_iter_2

This model is a fine-tuned version of YYYYYYibo/nash_dpo_iter_1 on the updated and the original datasets. It achieves the following results on the evaluation set:

• Loss: 0.6403
• Rewards/chosen: -0.3108
• Rewards/rejected: -0.4420
• Rewards/accuracies: 0.6160
• Rewards/margins: 0.1313
• Logps/rejected: -334.0379
• Logps/chosen: -329.4806
• Logits/rejected: -2.3393
• Logits/chosen: -2.4452

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	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6563	0.64	100	0.6403	-0.3108	-0.4420	0.6160	0.1313	-334.0379	-329.4806	-2.3393	-2.4452

Framework versions

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