Upload folder using huggingface_hub

api.py

@@ -11,7 +11,7 @@ app = Flask(__name__)
 
 # Parse command line arguments
 parser = argparse.ArgumentParser(description='Start the Flask server with specified model and device.')
-parser.add_argument('--model-path', type=str, required=True, help='Path to the pretrained model')
+parser.add_argument('--model-path', type=str, default="models/Florence-2-base-ft", help='Path to the pretrained model')
 parser.add_argument('--device', type=str, choices=['cpu', 'gpu'], default='auto', help='Device to use: "cpu", "gpu", or "auto"')
 args = parser.parse_args()
 

app.py

@@ -8,6 +8,7 @@ import threading
 from datetime import datetime
 import paho.mqtt.client as mqtt
 import gradio as gr
+from api import predict_image
 
 # Constants and configuration
 IMAGE_PATH = "received_image.jpg"
@@ -97,12 +98,7 @@ def predict_image_json(image, task, prompt):
     msgid = str(datetime.now().timestamp())
     if task == "<OD>":
         prompt = ""
-    buffered = BytesIO()
-    image.save(buffered, format="JPEG")
-    image_base64 = base64.b64encode(buffered.getvalue()).decode('utf-8')
-    json_data = json.dumps({"msgid": msgid, "task": task, "prompt": prompt, "image": f"data:image/jpeg;base64,{image_base64}"})
-    response = requests.post("http://localhost:5000/predict", headers={"Content-Type": "application/json"}, data=json_data)
-    prediction = response.json().get("prediction", {})
+    prediction = predict_image(image, task, prompt)
     if task == "<OPEN_VOCABULARY_DETECTION>":
         prediction[task] = convert_to_od_format(prediction[task])
     return prediction

requirements.txt

@@ -1 +1 @@
-paho-mqtt
+paho