Switch nous research

app.py

@@ -10,12 +10,13 @@ from datetime import datetime
 checkpoint = "CohereForAI/c4ai-command-r-v01"
 checkpoint = "mistralai/Mistral-7B-Instruct-v0.1"
 checkpoint = "google/gemma-1.1-7b-it"
+checkpoint = "NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO"
 path_to_log = "FlaggedFalse.txt"
 
 # Inference client with the model (And HF-token if needed)
 client = InferenceClient(checkpoint)
 tokenizer = AutoTokenizer.from_pretrained(checkpoint)
-if checkpoint == "mistralai/Mistral-7B-Instruct-v0.1":
+if checkpoint == "NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO":
     # Tokenizer chat template correction(Only works for mistral models)
     chat_template = open("mistral-instruct.jinja").read()
     chat_template = chat_template.replace('    ', '').replace('\n', '')
@@ -58,7 +59,7 @@ def inference(message, history, systemPrompt=SYSTEM_PROMPT+SYSTEM_PROMPT_PLUS, t
         seed=random.randint(0, 999999999),
     )
     # Generating the response by passing the prompt in right format plus the client settings
-    stream = client.text_generation(format_prompt_gemma(message, history, systemPrompt),
+    stream = client.text_generation(format_prompt(message, history, systemPrompt),
                                     **client_settings)
     # Reading the stream
     partial_response = ""