Update app.py

app.py

@@ -1,4 +1,6 @@
 import os
+os.environ["VLLM_ENABLE_CHUNKED_PREFILL"] = "False"  # Disable chunked prefill as a workaround
+
 import re
 import uuid
 import json
@@ -93,9 +95,7 @@ SYSTEM_PROMPT = load_system_prompt(MODEL_ID, "SYSTEM_PROMPT.txt")
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
 # Initialize the Mistral LLM via vllm.
-# Note: Running this model on GPU may require very high VRAM.
-# The 'enforce_eager=True' parameter disables asynchronous output,
-# which avoids the NotImplementedError on platforms that do not support it.
+# The 'enforce_eager=True' parameter disables asynchronous output.
 llm = LLM(model=MODEL_ID, tokenizer_mode="mistral", device=device, enforce_eager=True)
 
 # -----------------------------------------------------------------------------