metadata

license: apache-2.0
library_name: peft
tags:
  - alignment-handbook
  - generated_from_trainer
  - trl
  - dpo
  - generated_from_trainer
datasets:
  - sablo/HelpSteer_binarized
base_model: sablo/sablo-pebble-mistral
model-index:
  - name: sablo-pebble-mistral-dpo-lora-HelpSteer_binarized
    results: []

sablo-pebble-mistral-dpo-lora-HelpSteer_binarized

This model is a fine-tuned version of sablo/sablo-pebble-mistral on the sablo/HelpSteer_binarized dataset. It achieves the following results on the evaluation set:

Loss: 0.5212
Rewards/chosen: -2.5398
Rewards/rejected: -3.5311
Rewards/accuracies: 0.7406
Rewards/margins: 0.9913
Logps/rejected: -214.9349
Logps/chosen: -206.5847
Logits/rejected: -2.0624
Logits/chosen: -2.1620

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: 4
eval_batch_size: 8
seed: 42
distributed_type: multi-GPU
gradient_accumulation_steps: 2
total_train_batch_size: 8
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: cosine
lr_scheduler_warmup_ratio: 0.1
num_epochs: 1

Training results

Training Loss	Epoch	Step	Validation Loss	Rewards/chosen	Rewards/rejected	Rewards/accuracies	Rewards/margins	Logps/rejected	Logps/chosen	Logits/rejected	Logits/chosen
0.652	0.2	200	0.6595	0.0498	-0.0320	0.6415	0.0818	-98.2975	-120.2629	-2.0236	-2.1231
0.4905	0.39	400	0.5551	-1.6581	-2.1527	0.6958	0.4946	-168.9884	-177.1946	-2.0950	-2.1951
0.4249	0.59	600	0.5327	-3.4554	-4.3247	0.7241	0.8693	-241.3867	-237.1045	-2.0782	-2.1773
0.5858	0.79	800	0.5207	-2.5072	-3.4512	0.7335	0.9440	-212.2718	-205.4982	-2.0586	-2.1591
0.6128	0.98	1000	0.5212	-2.5398	-3.5311	0.7406	0.9913	-214.9349	-206.5847	-2.0624	-2.1620

Framework versions

PEFT 0.7.1
Transformers 4.36.2
Pytorch 2.0.1+cu118
Datasets 2.14.6
Tokenizers 0.15.0