selfbiorag-7b-dpo-full-wo-kqa_silver_wogold-ep3

This model is a fine-tuned version of dmis-lab/selfbiorag_7b on the HuggingFaceH4/ultrafeedback_binarized dataset. It achieves the following results on the evaluation set:

• Loss: 0.6392
• Rewards/chosen: 0.1232
• Rewards/rejected: -0.0030
• Rewards/accuracies: 0.7527
• Rewards/margins: 0.1262
• Logps/rejected: -171.6258
• Logps/chosen: -150.9050
• Logits/rejected: -1.5645
• Logits/chosen: -1.7964

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: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 2
• 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
0.6608	0.25	100	0.6631	0.1074	0.0395	0.7107	0.0680	-167.3843	-152.4830	-1.5362	-1.7612
0.6271	0.51	200	0.6474	0.1331	0.0272	0.7455	0.1060	-168.6118	-149.9109	-1.5243	-1.7495
0.61	0.76	300	0.6403	0.1251	0.0020	0.7554	0.1232	-171.1355	-150.7145	-1.5597	-1.7911

Framework versions

• Transformers 4.39.0.dev0
• Pytorch 2.1.2
• Datasets 2.14.6
• Tokenizers 0.15.2