metadata

library_name: transformers
license: gemma
base_model: google/gemma-7b
tags:
  - trl
  - orpo
  - generated_from_trainer
model-index:
  - name: gemma-7b-borpo-low-quality-v4
    results: []

gemma-7b-borpo-low-quality-v4

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

Loss: 1.9080
Rewards/chosen: -0.6019
Rewards/rejected: -0.7507
Rewards/accuracies: 0.6259
Rewards/margins: 0.1488
Logps/rejected: -1.5015
Logps/chosen: -1.2038
Logits/rejected: 247.6263
Logits/chosen: 282.0085
Nll Loss: 1.5545
Log Odds Ratio: -0.6477
Log Odds Chosen: 0.4230

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
1.8238	0.9955	167	1.8176	-0.5378	-0.6319	0.5468	0.0941	-1.2637	-1.0755	293.2744	322.6454	1.4783	-0.6631	0.2616
1.3092	1.9970	335	1.7560	-0.5202	-0.6309	0.5324	0.1106	-1.2617	-1.0405	279.0659	309.3900	1.4054	-0.6637	0.3224
0.6827	2.9866	501	1.9080	-0.6019	-0.7507	0.6259	0.1488	-1.5015	-1.2038	247.6263	282.0085	1.5545	-0.6477	0.4230

Framework versions

Transformers 4.44.2
Pytorch 2.4.1+cu121
Datasets 3.0.0
Tokenizers 0.19.1