Update app.py

app.py

@@ -7,8 +7,8 @@ from threading import Thread
 
 # Loading the tokenizer and model from Hugging Face's model hub.
 if torch.cuda.is_available():
-    tokenizer = AutoTokenizer.from_pretrained("0x7194633/fialka-13B-v3.1")
-    model = AutoModelForCausalLM.from_pretrained("0x7194633/fialka-13B-v3.1", load_in_8bit=True, device_map="auto")
+    tokenizer = AutoTokenizer.from_pretrained("0x7194633/fialka-13B-v4")
+    model = AutoModelForCausalLM.from_pretrained("0x7194633/fialka-13B-v4", load_in_8bit=True, device_map="auto")
 
 
 # Defining a custom stopping criteria class for the model's text generation.
@@ -28,7 +28,8 @@ def predict(message, history):
     stop = StopOnTokens()
 
     # Formatting the input for the model.
-    messages = "</s>".join(["</s>".join(["\n<|user|>" + item[0], "\n<|assistant|>" + item[1]])
+    messages = "<|system|>\nТы Фиалка - самый умный нейронный помощник, созданный 0x7o.</s>\n"
+    messages += "</s>".join(["</s>".join(["\n<|user|>" + item[0], "\n<|assistant|>" + item[1]])
                         for item in history_transformer_format])
     model_inputs = tokenizer([messages], return_tensors="pt").to("cuda")
     streamer = TextIteratorStreamer(tokenizer, timeout=10., skip_prompt=True, skip_special_tokens=True)