Hugging Face's logo

Holarissun
/
RM-HH-AllMix_helpful_gpt3_loraR64_20000_gemma2b_shuffleTrue_extractchosenFalse

PEFT
Safetensors
trl
reward-trainer
Generated from Trainer
Model card Files Community
RM-HH-AllMix_helpful_gpt3_loraR64_20000_gemma2b_shuffleTrue_extractchosenFalse / README.md
Holarissun's picture
Holarissun
End of training
2715002 verified
preview code
|
raw history blame contribute delete
No virus
4.48 kB
metadata
license: gemma
library_name: peft
tags:
  - trl
  - reward-trainer
  - generated_from_trainer
base_model: google/gemma-2b
metrics:
  - accuracy
model-index:
  - name: >-
      RM-HH-AllMix_helpful_gpt3_loraR64_20000_gemma2b_shuffleTrue_extractchosenFalse
    results: []

RM-HH-AllMix_helpful_gpt3_loraR64_20000_gemma2b_shuffleTrue_extractchosenFalse

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: 0.5220
  • Accuracy: 0.7437

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: 2.0

Training results

Training Loss Epoch Step Validation Loss Accuracy
0.7063 0.04 250 0.6784 0.5939
0.6441 0.08 500 0.6032 0.6613
0.582 0.13 750 0.5617 0.6921
0.5045 0.17 1000 0.5495 0.6985
0.5345 0.21 1250 0.5444 0.7034
0.53 0.25 1500 0.5522 0.7076
0.5325 0.29 1750 0.5550 0.7061
0.5145 0.33 2000 0.5596 0.7121
0.5156 0.38 2250 0.5480 0.7143
0.4995 0.42 2500 0.5477 0.7181
0.5329 0.46 2750 0.5350 0.7207
0.5037 0.5 3000 0.5472 0.7196
0.5417 0.54 3250 0.5233 0.7249
0.5179 0.59 3500 0.5230 0.7256
0.5264 0.63 3750 0.5196 0.7286
0.4931 0.67 4000 0.5267 0.7279
0.5114 0.71 4250 0.5202 0.7317
0.4735 0.75 4500 0.5238 0.7332
0.4902 0.79 4750 0.5294 0.7332
0.5483 0.84 5000 0.5165 0.7343
0.548 0.88 5250 0.5070 0.7350
0.4918 0.92 5500 0.5115 0.7384
0.5079 0.96 5750 0.5108 0.7369
0.49 1.0 6000 0.5127 0.7388
0.5161 1.05 6250 0.5103 0.7392
0.4573 1.09 6500 0.5226 0.7369
0.4973 1.13 6750 0.5208 0.7358
0.5163 1.17 7000 0.5135 0.7373
0.4857 1.21 7250 0.5188 0.7381
0.4996 1.25 7500 0.5200 0.7384
0.5029 1.3 7750 0.5185 0.7388
0.4983 1.34 8000 0.5177 0.7384
0.4718 1.38 8250 0.5186 0.7392
0.4723 1.42 8500 0.5204 0.7381
0.5238 1.46 8750 0.5143 0.7403
0.4613 1.51 9000 0.5178 0.7384
0.517 1.55 9250 0.5212 0.7377
0.495 1.59 9500 0.5181 0.7407
0.4865 1.63 9750 0.5191 0.7418
0.4799 1.67 10000 0.5231 0.7414
0.4546 1.71 10250 0.5241 0.7426
0.4673 1.76 10500 0.5256 0.7433
0.4598 1.8 10750 0.5259 0.7448
0.5035 1.84 11000 0.5245 0.7444
0.5113 1.88 11250 0.5236 0.7433
0.4821 1.92 11500 0.5230 0.7433
0.5071 1.97 11750 0.5220 0.7437

Framework versions

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