initial commit

app.py

@@ -0,0 +1,101 @@
+import gradio as gr
+from transformers import AutoProcessor, AutoModelForCausalLM
+import spaces
+from PIL import Image
+import io
+import subprocess
+subprocess.run("pip install flash-attn --no-build-isolation", env={"FLASH_ATTENTION_SKIP_CUDA_BUILD": "TRUE"}, shell=True)
+
+models = {
+    "maxiw/Florence-2-ScreenQA-base": AutoModelForCausalLM.from_pretrained("maxiw/Florence-2-ScreenQA-base", trust_remote_code=True).to("cuda").eval(),
+}
+
+processors = {
+    "maxiw/Florence-2-ScreenQA-base": AutoProcessor.from_pretrained("maxiw/Florence-2-ScreenQA-base", trust_remote_code=True),
+}
+
+
+DESCRIPTION = "# [Florence-2-ScreenQA Demo](https://huggingface.co/maxiw/Florence-2-ScreenQA-base)"
+
+
+@spaces.GPU
+def run_example(task_prompt, image, text_input=None, model_id="maxiw/Florence-2-ScreenQA-base"):
+    model = models[model_id]
+    processor = processors[model_id]
+    if text_input is None:
+        prompt = task_prompt
+    else:
+        prompt = task_prompt + text_input
+    inputs = processor(text=prompt, images=image, return_tensors="pt").to("cuda")
+    generated_ids = model.generate(
+        input_ids=inputs["input_ids"],
+        pixel_values=inputs["pixel_values"],
+        max_new_tokens=1024,
+        early_stopping=False,
+        do_sample=False,
+        num_beams=3,
+    )
+    generated_text = processor.batch_decode(generated_ids, skip_special_tokens=False)[0]
+    parsed_answer = processor.post_process_generation(
+        generated_text,
+        task=task_prompt,
+        image_size=(image.width, image.height)
+    )
+    if "<SQA>" in parsed_answer:
+        parsed_answer = parsed_answer["<SQA>"]
+    return parsed_answer
+
+
+def process_image(image, task_prompt, text_input=None, model_id="maxiw/Florence-2-ScreenQA-base"):
+    image = Image.fromarray(image)  # Convert NumPy array to PIL Image
+    if task_prompt == "ScreenQA":
+        task_prompt = "<SQA>"
+        results = run_example(task_prompt, image, text_input, model_id=model_id)
+        return results
+    else:
+        print("Unknown task prompt")
+        return "", None  # Return empty string and None for unknown task prompts
+
+css = """
+  #output {
+    height: 500px; 
+    overflow: auto; 
+    border: 1px solid #ccc; 
+  }
+"""
+
+
+single_task_list =[
+    "ScreenQA"
+]
+
+
+with gr.Blocks(css=css) as demo:
+    gr.Markdown(DESCRIPTION)
+    with gr.Tab(label="Florence-2 Input"):
+        with gr.Row():
+            with gr.Column():
+                input_img = gr.Image(label="Input Picture")
+                model_selector = gr.Dropdown(choices=list(models.keys()), label="Model", value="maxiw/Florence-2-ScreenQA-base")
+                task_prompt = gr.Dropdown(choices=single_task_list, label="Task Prompt", value="ScreenQA")
+                text_input = gr.Textbox(label="Question")
+                submit_btn = gr.Button(value="Submit")
+            with gr.Column():
+                output_text = gr.Textbox(label="Output Text")
+
+        gr.Examples(
+            examples=[
+                ["image1.jpg", "ScreenQA", "What is the version of the settings?"],
+                ["image1.jpg", "ScreenQA", "What is the state of use lower resolution images?"],
+                ["image2.jpg", "ScreenQA", "How much is the discount for the product?"]
+            ],
+            inputs=[input_img, task_prompt, text_input],
+            outputs=[output_text],
+            fn=process_image,
+            cache_examples=True,
+            label="Try examples"
+        )
+
+        submit_btn.click(process_image, [input_img, task_prompt, text_input, model_selector], [output_text])
+
+demo.launch(debug=True)

pre-requirements.txt

@@ -0,0 +1 @@
+pip>=23.0.0

requirements.txt

@@ -0,0 +1,3 @@
+spaces
+transformers
+timm