fix(generate:app.py): completion creation

app.py

@@ -9,7 +9,7 @@ from llama_cpp import Llama
 from huggingface_hub import hf_hub_download
 
 hf_hub_download(repo_id="TheBloke/dolphin-2_6-phi-2-GGUF", filename="dolphin-2_6-phi-2.Q5_K_S.gguf", local_dir=".")
-llm = Llama(model_path="./dolphin-2_6-phi-2.Q5_K_S.gguf")
+llm = Llama(model_path="./dolphin-2_6-phi-2.Q5_K_S.gguf", n_ctx=512, last_n_tokens_size=256, n_threads=4, n_gpu_layers=0)
 
 ins = '''<|im_start|>user
 {question}<|im_end|>
@@ -66,9 +66,12 @@ custom_theme = BlueTheme()
 
 def generate(instruction):
 	prompt = ins.format(question=instruction)
-	response = llm(prompt, stop=['<|im_start|>user', '<|im_end|>'])
-	result = response['choices'][0]['text']
-	return result
+	response = llm.create_completion(prompt, stream=True, stop=['<|im_start|>user', '<|im_end|>'], repeat_penalty=1.1)
+
+	result = ""
+	for output in response:
+		result += output['choices'][0]['text']
+		yield result
 
 examples = [
 	"How do dogs bark?",