starchat2-15b-dpo-v0.1

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

• Loss: 0.4343
• Rewards/chosen: -1.1086
• Rewards/rejected: -3.1181
• Rewards/accuracies: 0.7785
• Rewards/margins: 2.0095
• Logps/rejected: -330.6774
• Logps/chosen: -321.5190
• Logits/rejected: -1.9859
• Logits/chosen: -2.0084

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: 2
• eval_batch_size: 4
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 8
• total_train_batch_size: 128
• 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: 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
0.4811	0.9996	574	0.4443	-0.7694	-2.4115	0.7658	1.6421	-316.5446	-314.7338	-1.9984	-2.0208
0.4589	1.9991	1148	0.4343	-1.1086	-3.1181	0.7785	2.0095	-330.6774	-321.5190	-1.9859	-2.0084

Framework versions

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