zephyr-dpo-timedial-selfgen-mix

This model is a fine-tuned version of EllieS/zephyr-dpo-timedial on the EllieS/timedial_selfgen dataset. It achieves the following results on the evaluation set:

• Loss: 0.0236
• Rewards/chosen: 0.5600
• Rewards/rejected: -3.3438
• Rewards/accuracies: 1.0
• Rewards/margins: 3.9038
• Logps/rejected: -542.8434
• Logps/chosen: -74.4535
• Logits/rejected: -3.2030
• Logits/chosen: -2.7110

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-06
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• gradient_accumulation_steps: 2
• total_train_batch_size: 4
• 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
0.039	0.35	100	0.0525	0.5119	-2.4372	1.0	2.9492	-452.1829	-79.2593	-3.2065	-2.7171
0.0106	0.7	200	0.0236	0.5600	-3.3438	1.0	3.9038	-542.8434	-74.4535	-3.2030	-2.7110

Framework versions

• PEFT 0.7.1
• Transformers 4.39.0.dev0
• Pytorch 2.1.2+cu121
• Datasets 2.14.6
• Tokenizers 0.15.2