Update app.py

app.py

@@ -4,37 +4,16 @@ from PIL import Image
 from transformers import PaliGemmaForConditionalGeneration, PaliGemmaProcessor
 import spaces
 
-@spaces.GPU
-def infer_diagram(image, question):
-    model = PaliGemmaForConditionalGeneration.from_pretrained("google/paligemma-3b-ft-ai2d-448").to("cuda")
-    processor = PaliGemmaProcessor.from_pretrained("google/paligemma-3b-ft-ai2d-448")
-
-    inputs = processor(images=image, text=question, return_tensors="pt").to("cuda")
-
-    predictions = model.generate(**inputs, max_new_tokens=100)
-    return processor.decode(predictions[0], skip_special_tokens=True)[len(question):].lstrip("\n")
 
 @spaces.GPU
 def infer_ocrvqa(image, question):
     model = PaliGemmaForConditionalGeneration.from_pretrained("google/paligemma-3b-ft-ocrvqa-896").to("cuda")
     processor = PaliGemmaProcessor.from_pretrained("google/paligemma-3b-ft-ocrvqa-896")
-
     inputs = processor(images=image,text=question, return_tensors="pt").to("cuda")
-
-    predictions = model.generate(**inputs, max_new_tokens=200)
-    
+    predictions = model.generate(**inputs, max_new_tokens=100)
     return processor.decode(predictions[0], skip_special_tokens=True)[len(question):].lstrip("\n")
 
 @spaces.GPU
-def infer_infographics(image, question):
-  model = PaliGemmaForConditionalGeneration.from_pretrained("google/paligemma-3b-ft-infovqa-896").to("cuda")
-  processor = PaliGemmaProcessor.from_pretrained("google/paligemma-3b-ft-infovqa-896")
-
-  inputs = processor(images=image, text=question, return_tensors="pt").to("cuda")
-
-  predictions = model.generate(**inputs, max_new_tokens=100)
-  return processor.decode(predictions[0], skip_special_tokens=True)[len(question):].lstrip("\n")
-@spaces.GPU
 def infer_doc(image, question):
   model = PaliGemmaForConditionalGeneration.from_pretrained("google/paligemma-3b-ft-docvqa-896").to("cuda")
 
@@ -52,45 +31,22 @@ css = """
 """
 
 with gr.Blocks(css=css) as demo:
-  gr.HTML("<h1><center>PaliGemma Fine-tuned on Documents 📄<center><h1>")
-  gr.HTML("<h3><center>This Space is built for you to compare different PaliGemma models fine-tuned on document tasks. ⚡</h3>")
-  gr.HTML("<h3><center>Each tab in this app demonstrates PaliGemma models fine-tuned on document question answering, infographics question answering, diagram understanding, and reading comprehension from images. 📄📕📊<h3>")
-  gr.HTML("<h3><center>Models are downloaded on the go, so first inference in each tab might take time if it's not already downloaded.<h3>")
+  gr.HTML("<h1><center>PaliGemma для VQA/OCR 📄<center><h1>")
+  gr.HTML("<h3><center>Использование модели "как есть" без файнтюнинга на документах. ⚡</h3>")
+  
 
-  with gr.Tab(label="Visual Question Answering over Documents"):
+  with gr.Tab(label="Ответы на вопросы по документам"):
     with gr.Row():
       with gr.Column():
         input_img = gr.Image(label="Input Document")
         question = gr.Text(label="Question")
         submit_btn = gr.Button(value="Submit")
       output = gr.Text(label="Answer")
-    gr.Examples(
-    [["assets/docvqa_example.png", "How many items are sold?"]],
-    inputs = [input_img, question],
-    outputs = [output],
-    fn=infer_doc,
-    label='Click on any Examples below to get Document Question Answering results quickly 👇'
-    )
 
     submit_btn.click(infer_doc, [input_img, question], [output])
 
-  with gr.Tab(label="Visual Question Answering over Infographics"):
-    with gr.Row():
-      with gr.Column():
-        input_img = gr.Image(label="Input Image")
-        question = gr.Text(label="Question")
-        submit_btn = gr.Button(value="Submit")
-      output = gr.Text(label="Answer")
-    gr.Examples(
-    [["assets/infographics_example (1).jpeg", "What is this infographic about?"]],
-    inputs = [input_img, question],
-    outputs = [output],
-    fn=infer_infographics,
-    label='Click on any Examples below to get Infographics QA results quickly 👇'
-    )
-
-    submit_btn.click(infer_infographics, [input_img, question], [output])
-  with gr.Tab(label="Reading from Images"):
+  
+  with gr.Tab(label="Чтение текста со сканов"):
     with gr.Row():
       with gr.Column():
         input_img = gr.Image(label="Input Document")
@@ -98,27 +54,5 @@ with gr.Blocks(css=css) as demo:
         submit_btn = gr.Button(value="Submit")
       output = gr.Text(label="Infer")
     submit_btn.click(infer_ocrvqa, [input_img, question], [output])
-    gr.Examples(
-    [["assets/ocrvqa.jpg", "Who is the author of this book?"]],
-    inputs = [input_img, question],
-    outputs = [output],
-    fn=infer_doc,
-    label='Click on any Examples below to get image reading comprehension results quickly 👇'
-    )
-  with gr.Tab(label="Diagram Understanding"):
-    with gr.Row():
-      with gr.Column():
-        input_img = gr.Image(label="Input Diagram")
-        question = gr.Text(label="Question")
-        submit_btn = gr.Button(value="Submit")
-      output = gr.Text(label="Infer")
-    submit_btn.click(infer_diagram, [input_img, question], [output])
-    gr.Examples(
-    [["assets/diagram.png", "What is the diagram showing?"]],
-    inputs = [input_img, question],
-    outputs = [output],
-    fn=infer_doc,
-    label='Click on any Examples below to get diagram understanding results quickly 👇'
-    )
 
 demo.launch(debug=True)