metadata

license: apache-2.0
library_name: peft
tags:
  - trl
  - dpo
  - generated_from_trainer
base_model: alignment-handbook/zephyr-7b-sft-full
model-index:
  - name: zephyr-7b-dpo-lora-pubmedqa-ultrafeedback
    results: []

zephyr-7b-dpo-lora-pubmedqa-ultrafeedback

This model is a fine-tuned version of alignment-handbook/zephyr-7b-sft-full on the None dataset. It achieves the following results on the evaluation set:

Loss: 0.5835
Rewards/chosen: -0.1486
Rewards/rejected: -0.4853
Rewards/accuracies: 0.7105
Rewards/margins: 0.3368
Logps/rejected: -314.4243
Logps/chosen: -302.0460
Logits/rejected: -2.5375
Logits/chosen: -2.5889

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
0.5703	0.92	7000	0.5835	-0.1500	-0.4872	0.7140	0.3372	-314.6089	-302.1864	-2.5236	-2.5765

Framework versions

PEFT 0.7.1
Transformers 4.36.2
Pytorch 2.1.1+cu121
Datasets 2.14.6
Tokenizers 0.15.0