zephyr-7b-gemma-dpo

This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.5041
• Rewards/chosen: -0.7250
• Rewards/rejected: -1.7776
• Rewards/accuracies: 0.7660
• Rewards/margins: 1.0526
• Logps/rejected: -446.8852
• Logps/chosen: -382.4696
• Logits/rejected: 314.8608
• Logits/chosen: 314.4090

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: 1
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 16
• total_train_batch_size: 128
• total_eval_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.5055	0.9858	52	0.5194	-0.1748	-1.0342	0.7447	0.8594	-432.0167	-371.4648	314.8033	314.6528
0.2167	1.9716	104	0.5041	-0.7250	-1.7776	0.7660	1.0526	-446.8852	-382.4696	314.8608	314.4090

Framework versions

• Transformers 4.45.2
• Pytorch 2.5.1+rocm6.2
• Datasets 3.5.0
• Tokenizers 0.20.3