Update app.py

app.py

@@ -71,7 +71,7 @@ def qwen_inference(model_name, media_input, text_input=None):
 
     if isinstance(media_input, str):
         media_path = media_input
-        if media_path.endswith(tuple([i for i in image_extensions.keys()])):
+        if media_path.endswith(tuple(image_extensions.keys())):
             media_type = "image"
         else:
             try:
@@ -83,10 +83,7 @@ def qwen_inference(model_name, media_input, text_input=None):
         {
             "role": "user",
             "content": [
-                {
-                    "type": media_type,
-                    media_type: media_path
-                },
+                { "type": media_type, media_type: media_path },
                 {"type": "text", "text": text_input},
             ],
         }
@@ -114,7 +111,6 @@ def qwen_inference(model_name, media_input, text_input=None):
     buffer = ""
     for new_text in streamer:
         buffer += new_text
-        # Remove <|im_end|> or similar tokens from the output
         buffer = buffer.replace("<|im_end|>", "")
         yield buffer
 
@@ -146,33 +142,19 @@ def generate_pdf(media_path, plain_text, font_choice, font_size, line_spacing, a
     styles["Normal"].fontName = font_choice
     styles["Normal"].fontSize = int(font_size)
     styles["Normal"].leading = int(font_size) * line_spacing
-    styles["Normal"].alignment = {
-        "Left": 0,
-        "Center": 1,
-        "Right": 2,
-        "Justified": 4
-    }[alignment]
+    styles["Normal"].alignment = {"Left": 0, "Center": 1, "Right": 2, "Justified": 4}[alignment]
 
     # Register font
     font_path = f"font/{font_choice}"
     pdfmetrics.registerFont(TTFont(font_choice, font_path))
 
     story = []
-
-    # Add image with size adjustment
-    image_sizes = {
-        "Small": (200, 200),
-        "Medium": (400, 400),
-        "Large": (600, 600)
-    }
+    image_sizes = {"Small": (200, 200), "Medium": (400, 400), "Large": (600, 600)}
     img = RLImage(media_path, width=image_sizes[image_size][0], height=image_sizes[image_size][1])
     story.append(img)
     story.append(Spacer(1, 12))
-
-    # Add plain text output
     text = Paragraph(plain_text, styles["Normal"])
     story.append(text)
-
     doc.build(story)
     return filename
 
@@ -180,17 +162,9 @@ def generate_docx(media_path, plain_text, font_choice, font_size, line_spacing,
     """Generates a DOCX document."""
     filename = f"output_{uuid.uuid4()}.docx"
     doc = docx.Document()
-
-    # Add image with size adjustment
-    image_sizes = {
-        "Small": docx.shared.Inches(2),
-        "Medium": docx.shared.Inches(4),
-        "Large": docx.shared.Inches(6)
-    }
+    image_sizes = {"Small": docx.shared.Inches(2), "Medium": docx.shared.Inches(4), "Large": docx.shared.Inches(6)}
     doc.add_picture(media_path, width=image_sizes[image_size])
     doc.add_paragraph()
-
-    # Add plain text output
     paragraph = doc.add_paragraph()
     paragraph.paragraph_format.line_spacing = line_spacing
     paragraph.paragraph_format.alignment = {
@@ -202,138 +176,84 @@ def generate_docx(media_path, plain_text, font_choice, font_size, line_spacing,
     run = paragraph.add_run(plain_text)
     run.font.name = font_choice
     run.font.size = docx.shared.Pt(int(font_size))
-
     doc.save(filename)
     return filename
 
-# Updated CSS for output styling
+# Minimal CSS for compact output
 css = """
-#output_text, #plain_text_output {
-    height: 200px;
-    overflow: auto;
-    border: 1px solid #ccc;
-}
-.submit-btn {
-    background-color: #cf3434 !important;
-    color: white !important;
-}
-.submit-btn:hover {
-    background-color: #ff2323 !important;
-}
-.download-btn {
-    background-color: #35a6d6 !important;
-    color: white !important;
-}
-.download-btn:hover {
-    background-color: #22bcff !important;
-}
+  #output { height: 400px; overflow: auto; border: 1px solid #ccc; }
+  .submit-btn { background-color: #cf3434 !important; color: white !important; }
+  .submit-btn:hover { background-color: #ff2323 !important; }
+  .download-btn { background-color: #35a6d6 !important; color: white !important; }
+  .download-btn:hover { background-color: #22bcff !important; }
 """
 
-# Gradio app setup with optimized UI
 with gr.Blocks(css=css) as demo:
-    gr.Markdown("# Qwen2VL Models: Vision and Language Processing")
+    gr.Markdown("# Qwen2VL: Vision & Language Processing")
 
-    with gr.Tab(label="Image Input"):
-        with gr.Row():
-            with gr.Column():
-                model_choice = gr.Dropdown(
-                    label="Model Selection",
-                    choices=list(MODEL_OPTIONS.keys()),
-                    value="Latex OCR"
-                )
-                input_media = gr.File(
-                    label="Upload Image", type="filepath"
-                )
-                text_input = gr.Textbox(label="Question", placeholder="Ask a question about the image...")
-                submit_btn = gr.Button(value="Submit", elem_classes="submit-btn")
-
-            with gr.Column():
-                output_text = gr.Textbox(label="Output Text", lines=5, elem_id="output_text")
-                plain_text_output = gr.Textbox(label="Standardized Plain Text", lines=5, elem_id="plain_text_output")
-
-        with gr.Row():
-            gr.Examples(
-                examples=[
-                    ["examples/1.png", "summarize the letter", "Text Analogy Ocrtest"],
-                    ["examples/2.jpg", "Summarize the full image in detail", "Latex OCR"],
-                    ["examples/3.png", "Describe the photo", "Qwen2VL Base"],
-                    ["examples/4.png", "summarize and solve the problem", "Math Prase"],
-                ],
-                inputs=[input_media, text_input, model_choice],
-                outputs=[output_text, plain_text_output],
-                fn=lambda img, question, model: qwen_inference(model, img, question),
-                cache_examples=False,
-            )
-
-        with gr.Accordion("Document Generation Options", open=False):
-            line_spacing = gr.Dropdown(
-                choices=[0.5, 1.0, 1.15, 1.5, 2.0, 2.5, 3.0],
-                value=1.5,
-                label="Line Spacing"
-            )
-            font_size = gr.Dropdown(
-                choices=["8", "10", "12", "14", "16", "18", "20", "22", "24"],
-                value="18",
-                label="Font Size"
-            )
-            font_choice = gr.Dropdown(
-                choices=[
-                    "DejaVuMathTeXGyre.ttf",
-                    "FiraCode-Medium.ttf",
-                    "InputMono-Light.ttf",
-                    "JetBrainsMono-Thin.ttf",
-                    "ProggyCrossed Regular Mac.ttf",
-                    "SourceCodePro-Black.ttf",
-                    "arial.ttf",
-                    "calibri.ttf",
-                    "mukta-malar-extralight.ttf",
-                    "noto-sans-arabic-medium.ttf",
-                    "times new roman.ttf",
-                    "ANGSA.ttf",
-                    "Book-Antiqua.ttf",
-                    "CONSOLA.TTF",
-                    "COOPBL.TTF",
-                    "Rockwell-Bold.ttf",
-                    "Candara Light.TTF",
-                    "Carlito-Regular.ttf Carlito-Regular.ttf",
-                    "Castellar.ttf",
-                    "Courier New.ttf",
-                    "LSANS.TTF",
-                    "Lucida Bright Regular.ttf",
-                    "TRTempusSansITC.ttf",
-                    "Verdana.ttf",
-                    "bell-mt.ttf",
-                    "eras-itc-light.ttf",
-                    "fonnts.com-aptos-light.ttf",
-                    "georgia.ttf",
-                    "segoeuithis.ttf",
-                    "youyuan.TTF",
-                    "TfPonetoneExpanded-7BJZA.ttf",
-                ],
-                value="youyuan.TTF",
-                label="Font Choice"
-            )
-            alignment = gr.Dropdown(
-                choices=["Left", "Center", "Right", "Justified"],
-                value="Justified",
-                label="Text Alignment"
-            )
-            image_size = gr.Dropdown(
-                choices=["Small", "Medium", "Large"],
-                value="Small",
-                label="Image Size"
-            )
-            file_format = gr.Radio(["pdf", "docx"], label="File Format", value="pdf")
-            get_document_btn = gr.Button(value="Get Document", elem_classes="download-btn")
-
-        submit_btn.click(
-            qwen_inference, [model_choice, input_media, text_input], [output_text]
-        ).then(
-            lambda output_text: format_plain_text(output_text), [output_text], [plain_text_output]
-        )
+    # Single column layout for compact UI
+    model_choice = gr.Dropdown(
+        label="Model Selection",
+        choices=list(MODEL_OPTIONS.keys()),
+        value="Latex OCR"
+    )
+    input_media = gr.File(label="Upload Image", type="filepath")
+    text_input = gr.Textbox(label="Question", placeholder="Ask a question about the image...")
+    submit_btn = gr.Button(value="Submit", elem_classes="submit-btn")
+    output_text = gr.Textbox(label="Output Text", lines=8)
+    plain_text_output = gr.Textbox(label="Standardized Plain Text", lines=8)
+
+    submit_btn.click(
+        qwen_inference, [model_choice, input_media, text_input], [output_text]
+    ).then(
+        lambda output_text: format_plain_text(output_text),
+        [output_text],
+        [plain_text_output]
+    )
 
-        get_document_btn.click(
-            generate_document, [input_media, output_text, file_format, font_choice, font_size, line_spacing, alignment, image_size], gr.File(label="Download Document")
-        )
+    gr.Markdown("### Document Generation Options")
+    line_spacing = gr.Dropdown(
+        choices=[0.5, 1.0, 1.15, 1.5, 2.0, 2.5, 3.0],
+        value=1.5,
+        label="Line Spacing"
+    )
+    font_size = gr.Dropdown(
+        choices=["8", "10", "12", "14", "16", "18", "20", "22", "24"],
+        value="18",
+        label="Font Size"
+    )
+    font_choice = gr.Dropdown(
+        choices=[
+            "DejaVuMathTeXGyre.ttf", "FiraCode-Medium.ttf", "InputMono-Light.ttf",
+            "JetBrainsMono-Thin.ttf", "ProggyCrossed Regular Mac.ttf", "SourceCodePro-Black.ttf",
+            "arial.ttf", "calibri.ttf", "mukta-malar-extralight.ttf", "noto-sans-arabic-medium.ttf",
+            "times new roman.ttf", "ANGSA.ttf", "Book-Antiqua.ttf", "CONSOLA.TTF", "COOPBL.TTF",
+            "Rockwell-Bold.ttf", "Candara Light.TTF", "Carlito-Regular.ttf", "Castellar.ttf",
+            "Courier New.ttf", "LSANS.TTF", "Lucida Bright Regular.ttf", "TRTempusSansITC.ttf",
+            "Verdana.ttf", "bell-mt.ttf", "eras-itc-light.ttf", "fonnts.com-aptos-light.ttf",
+            "georgia.ttf", "segoeuithis.ttf", "youyuan.TTF", "TfPonetoneExpanded-7BJZA.ttf"
+        ],
+        value="youyuan.TTF",
+        label="Font Choice"
+    )
+    alignment = gr.Dropdown(
+        choices=["Left", "Center", "Right", "Justified"],
+        value="Justified",
+        label="Text Alignment"
+    )
+    image_size = gr.Dropdown(
+        choices=["Small", "Medium", "Large"],
+        value="Small",
+        label="Image Size"
+    )
+    file_format = gr.Radio(["pdf", "docx"], label="File Format", value="pdf")
+    get_document_btn = gr.Button(value="Get Document", elem_classes="download-btn")
+    document_output = gr.File(label="Download Document")
+
+    get_document_btn.click(
+        generate_document,
+        [input_media, output_text, file_format, font_choice, font_size, line_spacing, alignment, image_size],
+        document_output
+    )
 
 demo.launch(debug=True)