trying ZeroGPU again with flash-attn

https://huggingface.co/spaces/zero-gpu-explorers/README/discussions/75

app.py

@@ -7,9 +7,16 @@ import gradio as gr
 import cv2
 import tempfile
 from ultralytics import YOLO
-# import spaces
+import spaces, subprocess
 
-# @spaces.GPU
+# Install flash attention, skipping CUDA build if necessary
+subprocess.run(
+    "pip install flash-attn --no-build-isolation",
+    env={"FLASH_ATTENTION_SKIP_CUDA_BUILD": "TRUE"},
+    shell=True,
+)
+
+@spaces.GPU
 def yolov12_inference(image, video, model_id, image_size, conf_threshold):
     model = YOLO(model_id)
     if image: