dpo-selective-mixdata

This model was trained from scratch on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5805
• Rewards/chosen: -4.7292
• Rewards/rejected: -5.2763
• Rewards/accuracies: 0.6934
• Rewards/margins: 0.5471
• Logps/rejected: -654.5243
• Logps/chosen: -590.7578
• Logits/rejected: 6.2956
• Logits/chosen: 6.4467

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: 4
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• total_eval_batch_size: 32
• 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
0.5481	0.27	500	0.6089	-2.6822	-3.1236	0.6705	0.4414	-439.2521	-386.0565	3.9671	4.1604
0.5519	0.53	1000	0.5867	-4.2523	-4.7597	0.6894	0.5074	-602.8671	-543.0739	5.1974	5.3486
0.5597	0.8	1500	0.5821	-4.7906	-5.3218	0.6959	0.5311	-659.0733	-596.9037	6.4644	6.6294

Framework versions

• Transformers 4.36.2
• Pytorch 2.1.2
• Datasets 2.14.6
• Tokenizers 0.15.0