model_hh_shp4_400

This model is a fine-tuned version of meta-llama/Llama-2-7b-chat-hf on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 4.9693
  • Rewards/chosen: -8.6216
  • Rewards/rejected: -8.5158
  • Rewards/accuracies: 0.5100
  • Rewards/margins: -0.1058
  • Logps/rejected: -223.9328
  • Logps/chosen: -248.9690
  • Logits/rejected: -0.7143
  • Logits/chosen: -0.6784

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: 0.0005
  • train_batch_size: 4
  • eval_batch_size: 1
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 16
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_steps: 100
  • training_steps: 1000

Training results

Training Loss Epoch Step Validation Loss Rewards/chosen Rewards/rejected Rewards/accuracies Rewards/margins Logps/rejected Logps/chosen Logits/rejected Logits/chosen
0.0234 4.0 100 1.9125 1.9279 1.7933 0.6000 0.1345 -212.4782 -237.2473 -0.4999 -0.5089
0.0617 8.0 200 3.6324 1.3369 2.0986 0.5100 -0.7617 -212.1390 -237.9039 -0.8276 -0.7974
0.013 12.0 300 4.9916 -0.1751 0.9862 0.5100 -1.1614 -213.3750 -239.5840 -0.6475 -0.6290
0.0 16.0 400 4.9706 -8.6036 -8.4956 0.5200 -0.1080 -223.9103 -248.9490 -0.7141 -0.6786
0.0 20.0 500 4.9635 -8.6143 -8.5440 0.5 -0.0704 -223.9641 -248.9609 -0.7130 -0.6773
0.0 24.0 600 4.9862 -8.5996 -8.4640 0.5100 -0.1356 -223.8752 -248.9445 -0.7137 -0.6781
0.0 28.0 700 4.9701 -8.6180 -8.5034 0.5 -0.1147 -223.9190 -248.9650 -0.7143 -0.6782
0.0 32.0 800 4.9508 -8.5621 -8.5009 0.5100 -0.0612 -223.9163 -248.9029 -0.7140 -0.6782
0.0 36.0 900 4.9729 -8.6333 -8.5082 0.5 -0.1252 -223.9243 -248.9819 -0.7143 -0.6783
0.0 40.0 1000 4.9693 -8.6216 -8.5158 0.5100 -0.1058 -223.9328 -248.9690 -0.7143 -0.6784

Framework versions

  • PEFT 0.10.0
  • Transformers 4.39.3
  • Pytorch 2.2.2+cu121
  • Datasets 2.18.0
  • Tokenizers 0.15.2
Downloads last month
3
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for guoyu-zhang/model_hh_shp4_400

Adapter
(1208)
this model