Update app.py

app.py

@@ -7,6 +7,14 @@ import os
 """
 For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
 """
+
+
+subprocess.run(
+    "pip install flash-attn --no-build-isolation",
+    env={"FLASH_ATTENTION_SKIP_CUDA_BUILD": "TRUE"},
+    shell=True,
+)
+
 client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
 # pip install 'git+https://github.com/huggingface/transformers.git'