Feature(MInference): add pycuda==2023.1

app.py

@@ -5,6 +5,10 @@ subprocess.run(
     env={"FLASH_ATTENTION_SKIP_CUDA_BUILD": "TRUE"},
     shell=True,
 )
+subprocess.run(
+    "pip install pycuda==2023.1",
+    shell=True,
+)
 
 import gradio as gr
 import os

requirements.txt

@@ -1,4 +1,3 @@
 triton==2.1.0
-pycuda==2023.1
 accelerate
 transformers