Sunshine279/gammaPO-gemma-2-9b-it

Model Details

Model Description

We fine-tuned google/gemma-2-9b-it on princeton-nlp/gemma2-ultrafeedback-armorm with the gamma-SimPO objective.

• Developed by: Jie Sun, Junkang Wu, Jiancan Wu, Zhibo Zhu, Xingyu Lu, Jun Zhou, Lintao Ma, Xiang Wang

• Model type: Causal Language Model

• License: gemma

• Finetuned from model: google/gemma-2-9b-it

Model Sources

• Repository: https://github.com/sunjie279/gammaPO

• Paper: https://arxiv.org/pdf/2506.03690

How to Get Started with the Model

import torch
from transformers import pipeline

model_id = "Sunshine279/gammaPO-gemma-2-9b-it"

generator = pipeline(
    "text-generation",
    model=model_id,
    model_kwargs={"torch_dtype": torch.bfloat16},
    device="cuda",
)
outputs = generator([{"role": "user", "content": "What's the difference between llamas and alpacas?"}],
                      do_sample=False,
                      eos_token_id=[generator.tokenizer.convert_tokens_to_ids("<end_of_turn>"), generator.tokenizer.eos_token_id],
                      max_new_tokens=200)
print(outputs[0]['generated_text'])

Training details

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 8e-07
• train_batch_size: 1
• eval_batch_size: 2
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 16
• total_train_batch_size: 128
• total_eval_batch_size: 16
• 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
2.5863	0.8594	400	2.5622	-18.1350	-23.0307	0.7828	4.8958	-2.3031	-1.8135	-15.8316	-15.8114

Framework versions

• Transformers 4.42.4
• Pytorch 2.3.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1