Update app.py

app.py

@@ -28,12 +28,20 @@ def respond(
 ):
     torch.set_default_device("cuda")
 
-    model = AutoModelForCausalLM.from_pretrained("mistralai/Mistral-7B-Instruct-v0.2", torch_dtype="auto",load_in_4bit=True,trust_remote_code=True)
-    tokenizer = AutoTokenizer.from_pretrained("mistralai/Mistral-7B-Instruct-v0.2", trust_remote_code=True)
- 
+    tokenizer = AutoTokenizer.from_pretrained(
+        "cognitivecomputations/dolphin-2.8-mistral-7b-v02",
+        trust_remote_code=True
+    )
+    model = AutoModelForCausalLM.from_pretrained(
+        "cognitivecomputations/dolphin-2.8-mistral-7b-v02",
+        torch_dtype="auto",
+        load_in_4bit=True,
+        trust_remote_code=True
+    )
     history_transformer_format = history + [[message, ""]]
 
-    messages = system_prompt + "".join(["".join(["\n[INST]" + item[0], "[/INST]\n" + item[1] + "</s>"]) for item in history_transformer_format])
+    system_prompt = "<|im_start|>system\nYou are Dolphin, a helpful AI assistant.<|im_end|>"
+    messages = system_prompt + "".join(["".join(["\n<|im_start|>user\n" + item[0], "<|im_end|>\n<|im_start|>assistant\n" + item[1]]) for item in history_transformer_format])
     input_ids = tokenizer([messages], return_tensors="pt").to('cuda')
     streamer = TextIteratorStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
     generate_kwargs = dict(