fix causal

app.py

@@ -48,7 +48,7 @@ def load_llamaindex_stack(model_id: str, max_new_tokens: int, temperature: float
     # Tiny, fast sentence-transformers model for embeddings
     embed = HuggingFaceEmbedding(model_name="sentence-transformers/all-MiniLM-L6-v2")
 
-    
+    """
     tok = AutoTokenizer.from_pretrained(model_id)
     mdl = AutoModelForSeq2SeqLM.from_pretrained(model_id)
     text2text = pipeline(
@@ -58,20 +58,20 @@ def load_llamaindex_stack(model_id: str, max_new_tokens: int, temperature: float
         max_new_tokens=max_new_tokens,
         temperature=float(temperature)
     )
-    
+    """
 
     # Wrap the same tiny HF model for LlamaIndex
-
-    """
+    
     llm = HuggingFaceLLM(
         model_name=model_id,
-        tokenizer_name=model_id,        
+        tokenizer_name=model_id,
+        model_cls=AutoModelForSeq2SeqLM,
         context_window=2048,
         generate_kwargs={"max_new_tokens": max_new_tokens, "temperature": temperature},
         device_map="cpu",
     )
-    """    
-    llm = HuggingFaceLLM(pipeline=text2text)
+    
+    #llm = HuggingFaceLLM(pipeline=text2text)
 
     Settings.embed_model = embed
     Settings.llm = llm