falcon-7b-dpo-lora

This model is a fine-tuned version of tiiuae/falcon-7b on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6909
• Rewards/chosen: 0.0159
• Rewards/rejected: 0.0096
• Rewards/accuracies: 0.3175
• Rewards/margins: 0.0063
• Logps/rejected: -88.4353
• Logps/chosen: -109.9771
• Logits/rejected: -14.9384
• Logits/chosen: -14.8499

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: 32
• total_train_batch_size: 512
• total_eval_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.6927	1.0	121	0.6929	0.0035	0.0015	0.2857	0.0020	-88.5168	-110.1007	-14.9346	-14.8473
0.6915	2.0	242	0.6917	0.0116	0.0071	0.3056	0.0045	-88.4605	-110.0199	-14.9351	-14.8469
0.6913	3.0	363	0.6909	0.0159	0.0096	0.3175	0.0063	-88.4353	-109.9771	-14.9384	-14.8499

Framework versions

• Transformers 4.35.0
• Pytorch 2.1.2+cu121
• Datasets 2.14.6
• Tokenizers 0.14.1