Update app.py

app.py

@@ -12,9 +12,9 @@ DEFAULT_MAX_NEW_TOKENS = 1024
 MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "4096"))
 
 DESCRIPTION = """\
-# shisa-v2-llama3.1-8b
+# shisa-v2-unphi4-14b
 
-This Space demonstrates the [shisa-v2-llama3.1-8b](https://huggingface.co/shisa-ai/shisa-v2-llama3.1-8b) bilingual (JA/EN) chat model."""
+This Space demonstrates the [shisa-v2-unphi4-14b](https://huggingface.co/shisa-ai/shisa-v2-unphi4-14b) bilingual (JA/EN) chat model."""
 
 LICENSE = """
 """
@@ -24,7 +24,7 @@ if not torch.cuda.is_available():
 
 
 if torch.cuda.is_available():
-    model_id = "shisa-ai/shisa-v2-llama3.1-8b"
+    model_id = "shisa-ai/shisa-v2-unphi4-14b"
     model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.float16, device_map="auto")
     tokenizer = AutoTokenizer.from_pretrained(model_id)
     tokenizer.use_default_system_prompt = False
@@ -38,8 +38,7 @@ def generate(
     max_new_tokens: int = 1024,
     temperature: float = 0.6,
     top_p: float = 0.9,
-    top_k: int = 50,
-    repetition_penalty: float = 1.2,
+    repetition_penalty: float = 1.1,
 ) -> Iterator[str]:
     conversation = []
     if system_prompt:
@@ -111,7 +110,7 @@ chat_interface = gr.ChatInterface(
             minimum=1.0,
             maximum=2.0,
             step=0.05,
-            value=1.2,
+            value=1.1,
         ),
     ],
     stop_btn=None,