Update utils.py

utils.py

@@ -15,6 +15,8 @@ from transformers import AutoModelForCausalLM, AutoTokenizer
 from scipy.ndimage import binary_dilation
 import cv2
 import openai
+import subprocess
+subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
 from tenacity import retry, wait_exponential, stop_after_attempt, retry_if_exception_type