README.md

---
base_model: llava-hf/llava-1.5-7b-hf
library_name: transformers
model_name: llava-lora-12-06-rpo-0.1
tags:
- generated_from_trainer
- trl
- dpo
licence: license
---

# Model Card for llava-lora-12-06-rpo-0.1

This model is a fine-tuned version of [llava-hf/llava-1.5-7b-hf](https://huggingface.co/llava-hf/llava-1.5-7b-hf).
It has been trained using [TRL](https://github.com/huggingface/trl).

## Quick start

```python
from transformers import pipeline

question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
generator = pipeline("text-generation", model="abshetty/llava-lora-12-06-rpo-0.1", device="cuda")
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
print(output["generated_text"])
```

## Training procedure

[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/ashetty21-university-of-california-berkeley/huggingface/runs/ck7mmct6)

This model was trained with DPO, a method introduced in [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://huggingface.co/papers/2305.18290).

### Framework versions

- TRL: 0.12.2
- Transformers: 4.46.3
- Pytorch: 2.5.1+cu121
- Datasets: 3.1.0
- Tokenizers: 0.20.3

## Citations

Cite DPO as:

```bibtex
@inproceedings{rafailov2023direct,
    title        = {{Direct Preference Optimization: Your Language Model is Secretly a Reward Model}},
    author       = {Rafael Rafailov and Archit Sharma and Eric Mitchell and Christopher D. Manning and Stefano Ermon and Chelsea Finn},
    year         = 2023,
    booktitle    = {Advances in Neural Information Processing Systems 36: Annual Conference on Neural Information Processing Systems 2023, NeurIPS 2023, New Orleans, LA, USA, December 10 - 16, 2023},
    url          = {http://papers.nips.cc/paper_files/paper/2023/hash/a85b405ed65c6477a4fe8302b5e06ce7-Abstract-Conference.html},
    editor       = {Alice Oh and Tristan Naumann and Amir Globerson and Kate Saenko and Moritz Hardt and Sergey Levine},
}
```

Cite TRL as:
    
```bibtex
@misc{vonwerra2022trl,
	title        = {{TRL: Transformer Reinforcement Learning}},
	author       = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouédec},
	year         = 2020,
	journal      = {GitHub repository},
	publisher    = {GitHub},
	howpublished = {\url{https://github.com/huggingface/trl}}
}
```

#Train the model
training_args = DPOConfig(
    output_dir="llava-lora-12-06-rpo-0.1",
    bf16=True,
    gradient_checkpointing=True,
    per_device_train_batch_size=8,
    per_device_eval_batch_size=4,
    gradient_accumulation_steps=32,
    evaluation_strategy="steps",
    eval_steps=1,
    learning_rate=1e-5,
    beta=0.1,
    warmup_ratio=0.1,
    lr_scheduler_type="cosine",
    num_train_epochs=2,
    rpo_alpha=0.1,
    dataset_num_proc=32,  # tokenization will use 32 processes
    dataloader_num_workers=32,  # data loading will use 32 workers
    logging_steps=1,
)

#Define LoRA configuration with specified rank
lora_config = LoraConfig(
    r=64,  # Set rank to 64
    lora_alpha=128,  # Set scaling factor to 128
    target_modules="all-linear",  # Target all linear layers
    lora_dropout=0.1,
)

trainer = DPOTrainer(
    model,
    ref_model=None,  # not needed when using peft
    args=training_args,
    train_dataset=train_dataset,
    eval_dataset=eval_dataset,
    tokenizer=processor,
    peft_config=lora_config,
)