Bad Improved Prompt Model

The Bad Improved Prompt Model is designed to enhance user inputs for image generation models. By refining vague or unclear prompts, it helps produce more accurate and visually appealing images.

Model

The fine-tuned model is available on Hugging Face:

• Model Name: bad-improved-prompt-model

It is fine-tuned from the model unsloth/llama-3.2-3b-instruct-unsloth-bnb-4bit.

Dataset

The model was trained on the following dataset:

• Dataset Name: bad-improved-prompt-pair-sharegpt

This dataset contains pairs of bad and improved prompts, facilitating the model's learning process to enhance prompt quality. It is available in Parquet format and consists of approximately 11,500 rows.

Fine-Tuning

The fine-tuning process is documented in the following Jupyter Notebook:

• Fine-Tuning Notebook: finetuning.ipynb

This notebook provides a step-by-step guide on how the model was fine-tuned using the dataset mentioned above.

Testing the Model

To evaluate the performance of the fine-tuned model, refer to the following Jupyter Notebook:

• Model Testing Notebook: finetune_model_test.ipynb

This notebook demonstrates how to test the model's ability to improve user prompts.

Usage

To utilize the fine-tuned model, you can use the following code:

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

model_id = "milliyin/bad-improved-prompt-model"

# Load the tokenizer and model
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto", load_in_8bit=True)

from unsloth.chat_templates import get_chat_template
from unsloth import FastLanguageModel

tokenizer = get_chat_template(
    tokenizer,
    chat_template="llama-3.1",
)
FastLanguageModel.for_inference(model)  # Enable native 2x faster inference

messages = [
    {"role": "system", "content": "You are the assistant who converts bad sentences into grammatically correct, well-structured, detailed, and lengthy sentences."},
    {"role": "user", "content": "Model wears red dress"},
]

inputs = tokenizer.apply_chat_template(
    messages,
    tokenize=True,
    add_generation_prompt=True,  # Must add for generation
    return_tensors="pt",
).to("cuda")

outputs = model.generate(
    input_ids=inputs,
    max_new_tokens=128,
    use_cache=True,
    temperature=1.5,
    min_p=0.1
)

print(tokenizer.batch_decode(outputs))

Replace "Model wears red dress" with the prompt you wish to improve. The model will generate an enhanced version of the input prompt.

Note: Ensure that the necessary libraries (transformers, torch, and unsloth) are installed in your environment. Additionally, verify that your system has a compatible GPU for model inference.

License

The model and dataset are licensed under the Apache-2.0 license.

Made with ❤️ by milliyin