Spaces:
Sleeping
Sleeping
File size: 4,099 Bytes
b472dbe 6d28548 b472dbe c021dcb 225f539 b472dbe 207e26b b472dbe c6232d6 a7e8129 4d9d123 a7e8129 c6232d6 a7e8129 e0ebd29 a7e8129 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 |
import PIL.Image
import gradio as gr
import base64
import time
import os
import google.generativeai as genai
import pathlib
txt_model = genai.GenerativeModel('gemini-pro')
vis_model = genai.GenerativeModel('gemini-pro-vision')
import os
GOOGLE_API_KEY=os.getenv('GOOGLE_API_KEY')
genai.configure(api_key=GOOGLE_API_KEY)
# Image to Base 64 Converter
def image_to_base64(image_path):
with open(image_path, 'rb') as img:
encoded_string = base64.b64encode(img.read())
return encoded_string.decode('utf-8')
# Function that takes User Inputs and displays it on ChatUI
def query_message(history,txt,img):
if not img:
history += [(txt,None)]
return history
base64 = image_to_base64(img)
data_url = f"data:image/jpeg;base64,{base64}"
history += [(f"{txt} ", None)]
return history
# Function that takes User Inputs, generates Response and displays on Chat UI
def llm_response(history,text,img):
if not img:
response = txt_model.generate_content(text)
history += [(None,response.text)]
return history
else:
img = PIL.Image.open(img)
response = vis_model.generate_content([text,img])
history += [(None,response.text)]
return history
# Interface Code- Selector method
def sentence_builder(animal, place):
return f"""how many {animal}s from the {place} are shown in the picture?"""
descript1 = gr.Markdown("""
## Multimodal Descript ##
<h5 align="center"><i>"Imagine learning XXXX."</i></h5>
Multimodal-CoT incorporates vision features in a decoupled training framework. The framework consists of two training stages: (i) rationale generation and (ii) answer inference. Both stages share the same model architecture but differ in the input and output.
""")
# gradio block
with gr.Blocks(theme='freddyaboulton/dracula_revamped') as app1:
title ="-COT-"
with gr.Row():
image_box = gr.Image(type="filepath")
chatbot = gr.Chatbot(
scale = 2,
height=750
)
text_box = gr.Dropdown(
["what is in the image", "provide alternative title for the image", "how many birds can be seen in the picture?"], label="Animal", info="Will add more animals later!"
)
btn = gr.Button("Submit")
clicked = btn.click(query_message,
[chatbot,text_box,image_box],
chatbot
).then(llm_response,
[chatbot,text_box,image_box],
chatbot
)
gr.Markdown("""
# Multimodal Chain-of-Thought Reasoning in Language Models
<h5 align="center"><i>"Imagine learning a textbook without figures or tables."</i></h5>
Multimodal-CoT incorporates vision features in a decoupled training framework. The framework consists of two training stages: (i) rationale generation and (ii) answer inference. Both stages share the same model architecture but differ in the input and output.
""")
with gr.Blocks(theme='snehilsanyal/scikit-learn') as app2:
gr.Markdown("## MM 2 ##")
title = descript1
with gr.Row():
image_box = gr.Image(type="filepath")
chatbot = gr.Chatbot(
scale = 2,
height=750
)
text_box = gr.Dropdown(
["what is in the image", "provide alternative title for the image", "how many birds can be seen in the picture?"], label="Animal", info="Will add more animals later!"
)
btn = gr.Button("Submit")
clicked = btn.click(query_message,
[chatbot,text_box,image_box],
chatbot
).then(llm_response,
[chatbot,text_box,image_box],
chatbot
)
with gr.Blocks(theme='snehilsanyal/scikit-learn') as demo:
gr.Markdown("# DEMO #")
gr.TabbedInterface([app1, app2], ["APP #1", "APP #2"])
demo.queue()
demo.launch() |