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_harmless_gpt3_20000_gemma2b_shuffleTrue_extractchosenTrue
    results: []

RM-HH-AllMix_harmless_gpt3_20000_gemma2b_shuffleTrue_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: 0.5277
Accuracy: 0.7168

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
0.8315	0.04	250	0.7587	0.5337
0.7136	0.08	500	0.6394	0.6303
0.6091	0.13	750	0.6058	0.6544
0.6005	0.17	1000	0.5916	0.6634
0.5844	0.21	1250	0.5839	0.6743
0.5823	0.25	1500	0.5729	0.6796
0.5845	0.29	1750	0.5629	0.6815
0.5726	0.33	2000	0.5599	0.6833
0.5564	0.38	2250	0.5675	0.6886
0.5681	0.42	2500	0.5550	0.6912
0.5713	0.46	2750	0.5367	0.6897
0.5403	0.5	3000	0.5392	0.6980
0.5299	0.54	3250	0.5502	0.7029
0.5397	0.59	3500	0.5411	0.7025
0.5629	0.63	3750	0.5377	0.7048
0.5307	0.67	4000	0.5290	0.7119
0.5154	0.71	4250	0.5322	0.7104
0.5307	0.75	4500	0.5363	0.7123
0.5414	0.79	4750	0.5320	0.7161
0.5444	0.84	5000	0.5269	0.7194
0.4831	0.88	5250	0.5325	0.7183
0.528	0.92	5500	0.5281	0.7187
0.527	0.96	5750	0.5277	0.7168

Framework versions

PEFT 0.10.0
Transformers 4.38.2
Pytorch 2.1.2+cu121
Datasets 2.18.0
Tokenizers 0.15.2