app.py

import gradio as gr
from huggingface_hub import InferenceClient

client = InferenceClient()

# Llama 3 - Text
model_llama_3_text = "meta-llama/Llama-3.2-3B-Instruct"

def fn_llama_3_text(
    prompt,
    history,
    system_prompt,
    max_tokens,
    temperature,
    top_p,
):
    
    # With System Prompt
    messages = [{"role": "system", "content": [{"type": "text", "text": system_prompt}]}]
    history.append(messages[0])
    messages.append({"role": "user", "content": [{"type": "text", "text": prompt}]})
    history.append(messages[1])
    
    stream = client.chat.completions.create(
        model = model_llama_3_text,
        messages = history,
        max_tokens = max_tokens,
        temperature = temperature,
        top_p = top_p,
        stream = True,
    )
    
    chunks = []
    for chunk in stream:
        chunks.append(chunk.choices[0].delta.content or "")
        yield "".join(chunks)

app_llama_3_text = gr.ChatInterface(
    fn = fn_llama_3_text,
    type = "messages",
    additional_inputs = [
        gr.Textbox(value="You are a helpful assistant.", label="System Prompt"),
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max Tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-P"),
    ],
    title = "Meta Llama 3",
    description = model_llama_3_text,
)

# Llama 3 - Vision
model_llama_3_vision = "meta-llama/Llama-3.2-11B-Vision-Instruct"

def fn_llama_3_vision(
    prompt,
    image_url,
    #system_prompt,
    max_tokens,
    temperature,
    top_p,
):
    
    # Without System Prompt
    messages = [{"role": "user", "content": [{"type": "text", "text": prompt}]}]
    if image_url:
        messages[0]["content"].append({"type": "image_url", "image_url": {"url": image_url}})
    
    stream = client.chat.completions.create(
        model = model_llama_3_vision,
        messages = messages,
        max_tokens = max_tokens,
        temperature = temperature,
        top_p = top_p,
        stream = True,
    )
    
    chunks = []
    for chunk in stream:
        chunks.append(chunk.choices[0].delta.content or "")
        yield "".join(chunks)

app_llama_3_vision = gr.Interface(
    fn = fn_llama_3_vision,
    inputs = [
        gr.Textbox(label="Prompt"),
        gr.Textbox(label="Image URL")
    ],
    outputs = [
        gr.Textbox(label="Output")
    ],
    additional_inputs = [
        #gr.Textbox(value="You are a helpful assistant.", label="System Prompt"),
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max Tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(minimum=0.1, maximum=1.0, value=0.95, step=0.05, label="Top-P"),
    ],
    title = "Meta Llama 3",
    description = model_llama_3_vision,
)

app = gr.TabbedInterface(
    [app_llama_3_text, app_llama_3_vision],
    ["Llama 3 - Text", "Llama 3 - Vision"]
).launch()