Spaces:
Sleeping
Sleeping
import gradio as gr | |
import requests | |
import os | |
import base64 | |
from PIL import Image | |
import io | |
import json | |
def resize_image(image_path, max_size=(800, 800), quality=85): | |
with Image.open(image_path) as img: | |
img.thumbnail(max_size, Image.Resampling.LANCZOS) | |
buffer = io.BytesIO() | |
img.save(buffer, format="JPEG", quality=quality) | |
return buffer.getvalue() | |
def filepath_to_base64(image_path): | |
img_bytes = resize_image(image_path) | |
img_base64 = base64.b64encode(img_bytes) | |
return f"data:image/jpeg;base64,{img_base64.decode('utf-8')}" | |
api_key = os.getenv('API_KEY') | |
def call_fuyu_8b_api(image_path, content, temperature=0.2, top_p=0.7, max_tokens=1024): | |
print(f"Caminho da imagem recebida: {image_path}") | |
print(f"Conteúdo: {content}") | |
print(f"Temperatura: {temperature}") | |
print(f"Top P: {top_p}") | |
print(f"Max Tokens: {max_tokens}") | |
image_base64 = filepath_to_base64(image_path) | |
invoke_url = "https://api.nvcf.nvidia.com/v2/nvcf/pexec/functions/9f757064-657f-4c85-abd7-37a7a9b6ee11" | |
headers = { | |
"Authorization": f"Bearer {api_key}", | |
"accept": "text/event-stream", | |
"content-type": "application/json", | |
} | |
payload = { | |
"messages": [ | |
{ | |
"content": f"{content} <img src=\"{image_base64}\" />", | |
"role": "user" | |
} | |
], | |
"temperature": temperature, | |
"top_p": top_p, | |
"max_tokens": max_tokens, | |
"stream": True | |
} | |
response = requests.post(invoke_url, headers=headers, json=payload, stream=True) | |
if response.status_code != 200: | |
print(f"Erro na requisição: {response.status_code}") | |
try: | |
error_details = response.json() | |
print(error_details) | |
except ValueError: | |
print(response.text) | |
else: | |
response_text = "" | |
for line in response.iter_lines(): | |
if line: | |
try: | |
# Decode the line from bytes to string | |
decoded_line = line.decode('utf-8') | |
# Remove the "data: " prefix | |
if decoded_line.startswith('data: '): | |
json_str = decoded_line[6:] # Remove the first 6 characters ('data: ') | |
json_line = json.loads(json_str) | |
# Assuming the structure is consistent with the examples you provided. | |
content_parts = json_line.get("choices", [{}])[0].get("delta", {}).get("content", "") | |
response_text += content_parts | |
else: | |
print(f"Unexpected line format: {decoded_line}") | |
except json.JSONDecodeError as e: | |
print(f"Error decoding JSON from response line: {e}") | |
print(f"Faulty line: {line}") | |
return response_text | |
content_input = gr.Textbox(lines=2, placeholder="Enter your content here...", label="Content") | |
image_input = gr.Image(type="filepath", label="Upload Image") | |
temperature_input = gr.Slider(minimum=0, maximum=1, step=0.01, value=0.2, label="Temperature") | |
top_p_input = gr.Slider(minimum=0, maximum=1, step=0.01, value=0.7, label="Top P") | |
max_tokens_input = gr.Slider(minimum=1, maximum=1024, step=1, value=1024, label="Max Tokens") | |
iface = gr.Interface(fn=call_fuyu_8b_api, | |
inputs=[image_input, content_input, temperature_input, top_p_input, max_tokens_input], | |
outputs="text", | |
title="Fuyu-8B API Explorer", | |
description=""" | |
<div style="text-align: center; font-size: 1.5em; margin-bottom: 20px;"> | |
<strong>Explore the Synergy of Text and Image with Fuyu-8B Multi-Modal Transformer</strong> | |
</div> | |
<p> | |
Embark on a journey into the future of AI with Fuyu-8B, a cutting-edge multi-modal transformer developed by Adept. This revolutionary model leverages an 8 billion parameter architecture to seamlessly integrate image understanding and text generation, offering a unique platform for exploring the intersection of visual and linguistic data. | |
</p> | |
<p> | |
<strong>How to Use:</strong> | |
</p> | |
<ol> | |
<li>Upload an <strong>image</strong> to provide a visual context for the model.</li> | |
<li>Type your <strong>content</strong> into the textbox, with your question.</li> | |
<li>Adjust <strong>Temperature</strong> and <strong>Top P</strong> sliders to fine-tune the creativity and variability of the output.</li> | |
<li>Set the <strong>Max Tokens</strong> to limit the length of the generated text.</li> | |
<li>Click <strong>Submit</strong> to witness the model's ability to generate insightful and relevant content based on the combined input of text and image.</li> | |
</ol> | |
<p> | |
<strong>Powered by NVIDIA's advanced AI technologies, Fuyu-8B API Explorer is your gateway to discovering new possibilities in multi-modal AI interactions, completely free to use.</strong> | |
</p> | |
<p> | |
<strong>Created by:</strong> @artificialguybr (<a href="https://twitter.com/artificialguybr">Twitter</a>) | |
</p> | |
<p> | |
<strong>Discover more:</strong> <a href="https://artificialguy.com">artificialguy.com</a> | |
</p> | |
""" | |
) | |
iface.launch() |