llama2-7b-sft-dpo-lora-20231201-52

This model is a fine-tuned version of meta-llama/Llama-2-7b-hf on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.6869
• Rewards/chosen: 0.0289
• Rewards/rejected: 0.0137
• Rewards/accuracies: 0.5675
• Rewards/margins: 0.0152
• Logps/rejected: -288.6960
• Logps/chosen: -359.9693
• Logits/rejected: -0.3198
• Logits/chosen: -0.1736

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
0.6911	1.0	121	0.6918	0.0084	0.0083	0.5238	0.0001	-288.7494	-360.1742	-0.3206	-0.1740
0.6892	2.0	242	0.6888	0.0170	0.0139	0.4841	0.0032	-288.6942	-360.0880	-0.3202	-0.1740
0.6867	3.0	363	0.6869	0.0289	0.0137	0.5675	0.0152	-288.6960	-359.9693	-0.3198	-0.1736

Framework versions

• Transformers 4.35.0
• Pytorch 2.1.1
• Datasets 2.14.6
• Tokenizers 0.14.1