Hugging Face's logo

Spaces:
selamw
/
BirdWatcher
Runtime error

App Files Community
BirdWatcher
File size: 3,282 Bytes 
01ba43f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
 
import gradio as gr
from PIL import Image
from transformers import BitsAndBytesConfig, PaliGemmaForConditionalGeneration, PaliGemmaProcessor
import spaces
import torch
import os

access_token = os.getenv('HF_token')

model_id = "selamw/BirdWatcher-AI"

bnb_config = BitsAndBytesConfig(load_in_8bit=True)


def convert_to_markdown(input_text):
    # Split the input text into sections based on the '**' delimiter
    sections = input_text.split("**")
    
    # Initialize the formatted output with the bird name
    formatted_output = f"**{sections[0].strip()}**\n"
    
    # Process each section to format it
    for i in range(1, len(sections), 2):
        if i + 1 < len(sections):
            # Use '##' for subheadings and clean up the text
            header = sections[i].strip() + "** "
            content = sections[i + 1].strip()
            formatted_output += f"\n**{header}{content}\n"
    
    # Return the formatted output
    return formatted_output.strip()


@spaces.GPU
def infer_fin_pali(image, question):
    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    
    model = PaliGemmaForConditionalGeneration.from_pretrained(model_id, quantization_config=bnb_config, token=access_token)
    processor = PaliGemmaProcessor.from_pretrained(model_id, token=access_token)

    inputs = processor(images=image, text=question, return_tensors="pt").to(device)

    predictions = model.generate(**inputs, max_new_tokens=512)
    decoded_output = processor.decode(predictions[0], skip_special_tokens=True)[len(question):].lstrip("\n")

    # Ensure proper Markdown formatting
    formatted_output = convert_to_markdown(decoded_output)

    return formatted_output


css = """
    #mkd {
        height: 500px; 
        overflow: auto; 
        border: 1px solid #ccc; 
    }
    h1 {
        text-align: center;
    }
    h3 {
        text-align: center;
    }
    h2 {
        text-align: left;
    }
    span.gray-text {
        color: gray;
    }
"""

with gr.Blocks(css=css) as demo:
    gr.HTML("<h1>🦩 Bird Identifier: Powered by Fine-tuned PaliGemma 🦜</h1>")
    gr.HTML("<h3>Upload an image of a bird, and the model will generate a detailed description of its species.</h3>")
    
    with gr.Tab(label="Bird Identification"):
        with gr.Row():
            input_img = gr.Image(label="Input Bird Image") 
            with gr.Column():
                with gr.Row():
                    question = gr.Text(label="Default Prompt", value="Describe this bird", elem_id="default-prompt")
                with gr.Row():
                    submit_btn = gr.Button(value="Run")
                with gr.Row():
                    output = gr.Markdown(label="Response")  # Use Markdown component to display output
                    # output = gr.Text(label="Response")  # Use Markdown component to display output
        
        submit_btn.click(infer_fin_pali, [input_img, question], [output])
        
        gr.Examples(
            [["020.jpg", "Describe this bird"],
             ["5.jpg", "Describe this bird"],
             ["1.jpg", "Describe this bird"]],
            inputs=[input_img, question],
            outputs=[output],
            fn=infer_fin_pali,
            label='Examples 👇'
        )

demo.launch(debug=True)