Holarissun's picture
End of training
4d60e43 verified
metadata
license: gemma
library_name: peft
tags:
  - trl
  - reward-trainer
  - generated_from_trainer
metrics:
  - accuracy
base_model: google/gemma-2b
model-index:
  - name: RM-HH-Mix_harmless_gpt3_20000_gemma2b_shuffleFalse_extractchosenTrue
    results: []

RM-HH-Mix_harmless_gpt3_20000_gemma2b_shuffleFalse_extractchosenTrue

This model is a fine-tuned version of google/gemma-2b on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.1257
  • Accuracy: 0.9465

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: 1.41e-05
  • train_batch_size: 1
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 4
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 1.0

Training results

Training Loss Epoch Step Validation Loss Accuracy
0.8425 0.06 250 0.8667 0.5565
0.7359 0.11 500 0.5129 0.774
0.6491 0.17 750 0.3182 0.8645
0.6171 0.22 1000 0.2427 0.904
0.5956 0.28 1250 0.1885 0.925
0.5504 0.33 1500 0.1771 0.928
0.5778 0.39 1750 0.1663 0.931
0.574 0.44 2000 0.1533 0.937
0.614 0.5 2250 0.1523 0.9355
0.5568 0.56 2500 0.1427 0.9395
0.5474 0.61 2750 0.1300 0.9435
0.5179 0.67 3000 0.1308 0.944
0.5643 0.72 3250 0.1231 0.947
0.5704 0.78 3500 0.1262 0.9465
0.5348 0.83 3750 0.1275 0.946
0.5388 0.89 4000 0.1256 0.947
0.5579 0.94 4250 0.1257 0.9465

Framework versions

  • PEFT 0.9.0
  • Transformers 4.38.2
  • Pytorch 2.1.2
  • Datasets 2.18.0
  • Tokenizers 0.15.2