zephyr-7b-gemma-dpo

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

• Loss: 0.4989
• Rewards/chosen: 0.7674
• Rewards/rejected: -0.0150
• Rewards/accuracies: 0.8021
• Rewards/margins: 0.7824
• Logps/rejected: -393.8280
• Logps/chosen: -330.8377
• Logits/rejected: -1.5388
• Logits/chosen: -1.5352

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: 8
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• total_eval_batch_size: 64
• 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
0.5208	0.9858	52	0.5289	0.6786	0.0849	0.8021	0.5938	-391.8310	-332.6122	-1.5499	-1.5480
0.362	1.9716	104	0.4989	0.7674	-0.0150	0.8021	0.7824	-393.8280	-330.8377	-1.5388	-1.5352

Framework versions

• Transformers 4.45.2
• Pytorch 2.5.1+rocm6.2
• Datasets 3.2.0
• Tokenizers 0.20.3