update

app.py

@@ -11,7 +11,8 @@ if not huggingface_token:
     print("no HUGGINGFACE_TOKEN if you need set secret ")
     #raise ValueError("HUGGINGFACE_TOKEN environment variable is not set")
 
-model_id = "Qwen/Qwen1.5-0.5B-Chat"
+#model_id = "Qwen/Qwen1.5-0.5B-Chat"
+model_id = "Kendamarron/Tokara-0.5B-Chat-v0.1"
 
 device = "auto" # torch.device("cuda" if torch.cuda.is_available() else "cpu")
 dtype = torch.bfloat16
@@ -79,7 +80,7 @@ def generate_text(messages):
         
 
     text_generator = pipeline("text-generation", model=model, tokenizer=tokenizer,torch_dtype=dtype,device_map=device) #pipeline has not to(device)
-    result = text_generator(messages, max_new_tokens=256, do_sample=True, temperature=0.7)
+    result = text_generator(messages, max_new_tokens=256, do_sample=True, temperature=0.7,repetition_penalty=1.1,top_p=0.95,top_k=40)
 
     generated_output = result[0]["generated_text"]
     if isinstance(generated_output, list):