metadata

library_name: transformers
license: gemma
base_model: google/gemma-7b
tags:
  - alignment-handbook
  - trl
  - orpo
  - generated_from_trainer
  - trl
  - orpo
  - generated_from_trainer
datasets:
  - argilla/dpo-mix-7k
model-index:
  - name: gemma-7b-borpo
    results: []

gemma-7b-borpo

This model is a fine-tuned version of google/gemma-7b on the argilla/dpo-mix-7k dataset. It achieves the following results on the evaluation set:

Loss: 1.5984
Rewards/chosen: -0.0575
Rewards/rejected: -0.0699
Rewards/accuracies: 0.5899
Rewards/margins: 0.0124
Logps/rejected: -1.3977
Logps/chosen: -1.1506
Logits/rejected: 270.9628
Logits/chosen: 299.8625
Nll Loss: 1.5312
Log Odds Ratio: -0.6761
Log Odds Chosen: 0.3679

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-06
train_batch_size: 2
eval_batch_size: 1
seed: 42
distributed_type: multi-GPU
num_devices: 4
gradient_accumulation_steps: 4
total_train_batch_size: 32
total_eval_batch_size: 4
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: inverse_sqrt
lr_scheduler_warmup_steps: 100
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	Nll Loss	Log Odds Ratio	Log Odds Chosen
1.4516	0.9968	157	1.4765	-0.0513	-0.0577	0.5468	0.0064	-1.1547	-1.0260	293.8872	321.9495	1.4282	-0.6924	0.1911
1.0587	2.0	315	1.4250	-0.0502	-0.0595	0.5468	0.0093	-1.1904	-1.0035	296.0850	323.6012	1.3729	-0.6901	0.2723
0.5897	2.9905	471	1.5984	-0.0575	-0.0699	0.5899	0.0124	-1.3977	-1.1506	270.9628	299.8625	1.5312	-0.6761	0.3679

Framework versions

Transformers 4.44.2
Pytorch 2.4.0+cu121
Datasets 3.0.0
Tokenizers 0.19.1