gemma-4-finetune

Lora version

https://huggingface.co/rahul7star/gemma_4_lora

Quant Models

rahul7star/Gemma-Quant-Series

GGUF Test Link

https://huggingface.co/spaces/rahul7star/Gemma-4-E4B-Uncensored-HauhauCS-Aggressive-Q5_K_P

Fine Tune Model Use case

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="rahul7star/gemma-4-finetune")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

output
[{'input_text': [{'role': 'user',
    'content': [{'type': 'image',
      'url': 'https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG'},
     {'type': 'text', 'text': 'What animal is on the candy?'}]}],
  'generated_text': [{'role': 'user',
    'content': [{'type': 'image',
      'url': 'https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG'},
     {'type': 'text', 'text': 'What animal is on the candy?'}]},
   {'content': "Based on the image, the candies appear to be **chocolate-coated candies** with designs on them.\n\nThe designs visible on the candies are **animals**. Specifically, the green and blue candies seem to have a design that resembles a **bee** or some kind of **insect/animal**. The orange candy also has a design that looks like an **insect** or perhaps a stylized **animal**.\n\nWithout a clearer, closer view of the specific details on each candy, it's difficult to name the exact animal with certainty, but they are clearly **animal-themed candies**.",
    'role': 'assistant'}]}]

Gradio or Kaggle for Image and Text

import gradio as gr
from transformers import pipeline
from PIL import Image
import requests
from io import BytesIO

# -----------------------------
# Load model (cached once)
# -----------------------------
pipe = pipeline(
    "image-text-to-text",
    model="rahul7star/gemma-4-finetune"
)

# -----------------------------
# Helper: load image safely
# -----------------------------
def load_image(img):
    if isinstance(img, str):  # URL case
        response = requests.get(img)
        return Image.open(BytesIO(response.content)).convert("RGB")
    return img.convert("RGB")

# -----------------------------
# Inference function
# -----------------------------
def chat(image, text):
    if image is None or text is None:
        return "Please provide both image and text."

    image = load_image(image)

    messages = [
        {
            "role": "user",
            "content": [
                {"type": "image", "image": image},
                {"type": "text", "text": text}
            ]
        }
    ]

    try:
        output = pipe(text=messages)
        return output
    except Exception as e:
        return f"Error: {str(e)}"

# -----------------------------
# Gradio UI
# -----------------------------
with gr.Blocks(title="Gemma 4 Vision Chat") as demo:
    gr.Markdown("# 🧠 Gemma-4 Image + Text Chat (Kaggle)")

    with gr.Row():
        image_input = gr.Image(type="pil", label="Upload Image")
        text_input = gr.Textbox(label="Prompt", placeholder="Ask something about the image...")

    btn = gr.Button("Run")
    output = gr.Textbox(label="Model Output")

    btn.click(fn=chat, inputs=[image_input, text_input], outputs=output)

# -----------------------------
# Launch
# -----------------------------
demo.launch()

Gradio or Kaggle for text base chat only

import gradio as gr
from transformers import pipeline

# -----------------------------
# Load model (cached once)
# -----------------------------
pipe = pipeline(
    "text-generation",   # 👈 switched to text chat mode
    model="rahul7star/gemma-4-finetune"
)

# -----------------------------
# Chat function
# -----------------------------
def chat(user_message, history):
    if history is None:
        history = []

    # Convert Gradio history → messages format
    messages = []

    for h in history:
        messages.append({"role": "user", "content": h[0]})
        messages.append({"role": "assistant", "content": h[1]})

    messages.append({"role": "user", "content": user_message})

    try:
        response = pipe(
            messages,
            max_new_tokens=256,
            do_sample=True,
            temperature=0.7
        )

        # Extract output safely
        if isinstance(response, list):
            output = response[0]["generated_text"]
        else:
            output = response

        # Append to chat history
        history.append((user_message, output))
        return history, ""

    except Exception as e:
        history.append((user_message, f"Error: {str(e)}"))
        return history, ""

# -----------------------------
# Gradio UI
# -----------------------------
with gr.Blocks(title="Gemma 4 Text Chat") as demo:
    gr.Markdown("# 💬 Gemma-4 Text Chat (Kaggle / HF Model)")

    chatbot = gr.Chatbot()
    msg = gr.Textbox(placeholder="Type your message here...")
    clear = gr.Button("Clear")

    state = gr.State([])

    msg.submit(chat, inputs=[msg, state], outputs=[chatbot, msg])
    msg.submit(lambda h: h, inputs=[state], outputs=[state])

    clear.click(lambda: ([], []), outputs=[chatbot, state])

# -----------------------------
# Launch
# -----------------------------
demo.launch()

Uploaded finetuned model

• Developed by: rahul7star
• License: apache-2.0
• Finetuned from model : unsloth/gemma-4-E2B-it

This gemma4 model was trained 2x faster with Unsloth and Huggingface's TRL library.

<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>