Create app.py

app.py

@@ -0,0 +1,68 @@
+import os
+
+os.chdir("LLaVA_Med")
+os.system('pip install -q -e .')
+
+import warnings
+warnings.filterwarnings('ignore')
+
+import io
+from contextlib import redirect_stdout
+import gradio as gr
+from transformers import AutoTokenizer
+from llava.model.builder import load_pretrained_model
+from llava.mm_utils import get_model_name_from_path
+from llava.eval.run_llava import eval_model
+
+# Define the model path
+model_path = "Veda0718/llava-med-v1.5-mistral-7b-finetuned"
+
+# Load the model
+tokenizer, model, image_processor, context_len = load_pretrained_model(
+    model_path=model_path,
+    model_base=None,
+    model_name=get_model_name_from_path(model_path)
+)
+
+# Define the inference function
+def run_inference(image, question):
+    args = type('Args', (), {
+        "model_path": model_path,
+        "model_base": None,
+        "image_file": image,
+        "query": question,
+        "conv_mode": None,
+        "sep": ",",
+        "temperature": 0,
+        "top_p": None,
+        "num_beams": 1,
+        "max_new_tokens": 512
+    })()
+
+    # Capture the printed output of eval_model
+    f = io.StringIO()
+    with redirect_stdout(f):
+        eval_model(args)
+    output = f.getvalue()
+    return output
+
+# Create the Gradio interface
+with gr.Blocks(theme=gr.themes.Monochrome()) as app:
+    with gr.Column(scale=1):
+        gr.Markdown("<center><h1>LLaVA-Med</h1></center>")
+
+        with gr.Row():
+            image = gr.Image(type="filepath", scale=2)
+            question = gr.Textbox(placeholder="Enter a question", scale=3)
+
+        with gr.Row():
+            answer = gr.Textbox(placeholder="Answer pops up here", scale=1)
+
+        with gr.Row():
+            btn = gr.Button("Run Inference", scale=1)
+
+        btn.click(fn=run_inference, inputs=[image, question], outputs=answer)
+
+# Launch the app
+if __name__ == "__main__":
+    app.queue().launch()