Llama-2-7b-hf-DPO-Filtered-0.2-version-2

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: 1.0005
  • Rewards/chosen: -14.6192
  • Rewards/rejected: -15.8788
  • Rewards/accuracies: 0.6500
  • Rewards/margins: 1.2596
  • Logps/rejected: -200.2418
  • Logps/chosen: -187.6859
  • Logits/rejected: -1.3992
  • Logits/chosen: -1.4011

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-05
  • train_batch_size: 1
  • eval_batch_size: 1
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_steps: 10
  • 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
0.7942 0.3007 289 0.6999 -0.9241 -0.9313 0.4500 0.0072 -50.7666 -50.7346 -0.3565 -0.3616
0.2572 0.6015 578 0.7096 -5.2469 -5.5281 0.5 0.2812 -96.7347 -93.9629 -0.7591 -0.7630
1.0719 0.9022 867 0.9637 -8.1192 -8.8822 0.6500 0.7630 -130.2761 -122.6863 -1.0380 -1.0448
7.5957 1.2029 1156 0.8582 -11.6385 -12.5781 0.75 0.9396 -167.2348 -157.8790 -1.2656 -1.2691
1.1754 1.5036 1445 1.0008 -14.3623 -14.8965 0.6500 0.5342 -190.4189 -185.1169 -1.2682 -1.2719
4.1259 1.8044 1734 0.8550 -12.9397 -14.0616 0.6500 1.1219 -182.0699 -170.8906 -1.3336 -1.3374
0.0022 2.1051 2023 0.9542 -13.3905 -14.4785 0.6500 1.0880 -186.2391 -175.3986 -1.3731 -1.3758
0.0207 2.4058 2312 0.9893 -14.5572 -15.8387 0.6500 1.2815 -199.8413 -187.0660 -1.3984 -1.4023
0.0015 2.7066 2601 1.0005 -14.6192 -15.8788 0.6500 1.2596 -200.2418 -187.6859 -1.3992 -1.4011

Framework versions

  • PEFT 0.10.0
  • Transformers 4.40.2
  • Pytorch 2.3.0+cu121
  • Datasets 2.19.1
  • Tokenizers 0.19.1
Downloads last month
1
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for LBK95/Llama-2-7b-hf-DPO-Filtered-0.2-version-2

Adapter
(2329)
this model