File size: 2,624 Bytes
647a796
 
e213acb
647a796
 
 
 
a344135
647a796
a344135
647a796
a344135
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
647a796
a344135
647a796
 
52bab4e
 
 
 
a344135
 
 
 
 
 
 
52bab4e
a344135
647a796
52bab4e
a344135
 
647a796
 
3fb9b0c
647a796
ba4cb28
1310cf1
79a19e9
3fb9b0c
ba4cb28
 
1310cf1
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
import gradio as gr
from transformers import T5Tokenizer, T5ForConditionalGeneration

# Load the base T5 model and tokenizer
model = T5ForConditionalGeneration.from_pretrained('t5-small')
tokenizer = T5Tokenizer.from_pretrained('t5-small')

def generate_clinical_report(input_text, max_length=256, num_beams=4, no_repeat_ngram_size=3, length_penalty=2.0, early_stopping=True):
    """
    Generate a clinical report from the input text using the T5 model with configurable parameters.
    """
    try:
        # Prepare input text
        input_ids = tokenizer.encode("summarize: " + input_text, return_tensors="pt", max_length=512, truncation=True)
        
        # Generate report with provided parameters
        outputs = model.generate(
            input_ids,
            max_length=max_length,
            num_beams=num_beams,
            no_repeat_ngram_size=no_repeat_ngram_size,
            length_penalty=length_penalty,
            early_stopping=early_stopping,
            bad_words_ids=[[tokenizer.encode(word, add_special_tokens=False)[0]] 
                          for word in ['http', 'www', '.com', '.org']]
        )
        
        # Decode and return the generated report
        return tokenizer.decode(outputs[0], skip_special_tokens=True)
    except Exception as e:
        print(f"Error generating report: {str(e)}")
        return f"Error: {str(e)}"

# Create Gradio interface with API configuration
demo = gr.Interface(
    fn=generate_clinical_report,
    inputs=[
        gr.Textbox(
            lines=8,
            placeholder="Enter clinical notes here...",
            label="Clinical Notes"
        ),
        gr.Slider(minimum=50, maximum=500, value=256, step=1, label="Max Length"),
        gr.Slider(minimum=1, maximum=8, value=4, step=1, label="Num Beams"),
        gr.Slider(minimum=1, maximum=5, value=3, step=1, label="No Repeat Ngram Size"),
        gr.Slider(minimum=0.1, maximum=5.0, value=2.0, step=0.1, label="Length Penalty"),
        gr.Checkbox(value=True, label="Early Stopping")
    ],
    outputs=gr.Textbox(lines=8, label="Generated Clinical Report"),
    title="Clinical Report Generator",
    description="Generate professional clinical reports from clinical notes using a T5 model.",
    allow_flagging="never",
    analytics_enabled=False
)

# Launch the app with optimized configuration for Hugging Face Spaces
if __name__ == "__main__":
    demo.queue()  # Enable queue with default settings
    demo.launch(
        server_name="0.0.0.0",
        share=True,
        show_error=True,
        max_threads=40  # Increase max threads for better performance
    )