Update app.py

app.py

@@ -38,9 +38,6 @@ import langchain_community.vectorstores  # Vectorstore for publications
 import langchain_huggingface  # Embeddings
 import transformers
 
-import subprocess
-subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
-
 # The number of publications to retrieve for the prompt
 PUBLICATIONS_TO_RETRIEVE = 5
 
@@ -71,7 +68,7 @@ publication_vectorstore = langchain_community.vectorstores.FAISS.load_local(
 
 # Create the callable LLM
 llm = transformers.pipeline(
-    task="text-generation", model="Qwen/Qwen2.5-7B-Instruct-AWQ", device="cuda", attn_implementation="flash_attention_2"
+    task="text-generation", model="Qwen/Qwen2.5-7B-Instruct-AWQ", device="cuda"
 )