Update flask_app.py

flask_app.py

@@ -8,7 +8,7 @@ import torch
 # Load your tokenizer and model
 tokenizer = AutoTokenizer.from_pretrained("NLPGenius/GovGPT-llama3")
 model = AutoModelForCausalLM.from_pretrained("NLPGenius/GovGPT-llama3", 
-                                             torch_dtype=torch.bfloat16, 
+                                             #torch_dtype=torch.bfloat16, 
                                              device_map="auto",)
 
 
@@ -21,7 +21,7 @@ llm_chain = transformers.pipeline(
     # we pass model parameters here too
     temperature=0.3,  # 'randomness' of outputs, 0.0 is the min and 1.0 the max
     do_sample=True,
-    max_length=2000,  # mex number of tokens to generate in the output
+    max_length=2000,  # max number of tokens to generate in the output
     truncation=True,
     repetition_penalty=1.1  # without this output begins repeating
 )
@@ -106,8 +106,8 @@ def model_inference(retriever, question):
         You are a helpful assistant. Answer the query accurately by focusing solely on the most relevant parts of the provided context.
 
         1. Identify and use only the sections of the context directly related to the query, ignoring unrelated or extraneous information.
-        2. If the query cannot be explicitly answered using the relevant parts of the context, respond with: "The answer is not found in the context provided."
-        3. If you are unsure or do not know the answer, respond with: "I don't know."
+        2. If the query cannot be explicitly answered using the relevant parts of the context, respond with: "The answer is not found."
+        3. If you are unsure or do not know the answer, respond with: "I don't know.", don't try to make up an answer.
         4. Do not add information, interpretations, or assumptions beyond what is explicitly stated in the context.
         5. Ensure the response is concise, avoids redundancy, and directly addresses the query.